Article
To read the full-text of this research, you can request a copy directly from the authors.

Abstract

Today, intelligent machines \emph{interact and collaborate} with humans in a way that demands a greater level of trust between human and machine. A first step towards building intelligent machines that are capable of building and maintaining trust with humans is the design of a sensor that will enable machines to estimate human trust level in real-time. In this paper, two approaches for developing classifier-based empirical trust sensor models are presented that specifically use electroencephalography (EEG) and galvanic skin response (GSR) measurements. Human subject data collected from 45 participants is used for feature extraction, feature selection, classifier training, and model validation. The first approach considers a general set of psychophysiological features across all participants as the input variables and trains a classifier-based model for each participant, resulting in a trust sensor model based on the general feature set (i.e., a "general trust sensor model"). The second approach considers a customized feature set for each individual and trains a classifier-based model using that feature set, resulting in improved mean accuracy but at the expense of an increase in training time. This work represents the first use of real-time psychophysiological measurements for the development of a human trust sensor. Implications of the work, in the context of trust management algorithm design for intelligent machines, are also discussed.

No full-text available

Request Full-text Paper PDF

To read the full-text of this research,
you can request a copy directly from the authors.

... However, there is still much to learn about real-time trust estimation techniques as the current approaches have various limitations. Current approaches fail to provide trust measurements in scales traditionally used for trust in automation [1], or require prohibitive sophisticated sensing and perception methods [1,26]. These sophisticated methods include the processing of psychophysiological signals (e.g.: galvanic skin response), that are not practical for the vehicular environments, where driver-ADS interactions are likely to take place. ...
... However, there is still much to learn about real-time trust estimation techniques as the current approaches have various limitations. Current approaches fail to provide trust measurements in scales traditionally used for trust in automation [1], or require prohibitive sophisticated sensing and perception methods [1,26]. These sophisticated methods include the processing of psychophysiological signals (e.g.: galvanic skin response), that are not practical for the vehicular environments, where driver-ADS interactions are likely to take place. ...
... The framework is based on observable measures of drivers' behaviors and trust dynamic models. Although different trust estimation approaches have been previously reported in the literature [1,26], our method is simpler to implement. Those previous approaches represented trust as conditional probabilities. ...
... The analysis of human movement (HM) is undergoing changes because of access to technologies that use different methods of artificial intelligence (AI). Thus, the development of solutions that can learn to analyze, evaluate, diagnose, and prescribe appropriate highly personalized movement is feasible (Akash et al. 2018;Bastawrous et al. 2018;Cust et al. 2019;Kermany et al. 2018;Montoye et al. 2016;Mostafa et al. 2018;Sarowar 2018). Human performance is typically subjective, and subject to errors and bias. ...
... Human performance is typically subjective, and subject to errors and bias. The use of AI in the recognition of HM has the potential to improve both the efficiency and accuracy of the analysis of characteristics evaluated using data inputs that are obtained through biomechanical analysis equipment, such as inertial measurement units (IMUs) (accelerometry), view (cinemetry) and electromyography (EMG) (Akash et al. 2018;Bastawrous et al. 2018;Cust et al. 2019;Kermany et al. 2018;Montoye et al. 2016;Mostafa et al. 2018;Sarowar 2018). New technologies improve communication between devices that have such a purpose, thereby enabling greater interaction with human users. ...
... New technologies improve communication between devices that have such a purpose, thereby enabling greater interaction with human users. Small devices with good precision are easy to carry, which can make wearable technologies part of the routine for people who have access to this technology (Akash et al. 2018;Bastawrous et al. 2018;Kermany et al. 2018;Montoye et al. 2016;Mostafa et al. 2018;Sarowar 2018). Improving the accuracy and speed of decision making for assessing tasks in various types of environments is an important purpose of AI (Cippitelli et al. 2016;Cust et al. 2019;Manoj and Thyagaraju 2018;Montoye et al. 2016;Mostafa et al. 2018). ...
Article
Full-text available
Technological advances that involve computing and artificial intelligence (AI) have led to advances in analysis methods. Fuzzy logic (FL) serves as a qualitative interpretation tool for AI. The objective of this systematic review is to investigate the methods of human movement (HM) analysis using AI through FL to understand the characteristics of the movement of healthy people. To identify relevant studies published up to April 19, 2019, we conducted a study of the PubMed, Scopus, ScienceDirect, and IEEE Xplore databases. We included studies that evaluated HM through AI using FL in healthy people. A total of 951 articles were examined, of which six were selected because they met the criteria presented in the methods. The protocols had high heterogeneity, yet all articles selected presented statistically satisfactory results, in addition to low errors or a false positive index. Only one selected article presented protocol applicability within the free-living model. Generally, AI using FL is a good tool to help assess HM in healthy people, but the model still needs new data acquisition entries to make it applicability within the free-living model.
... Previous research has obtained promising results regarding trust estimation in AVs using drivers' behaviors and actions [7]- [9]. The second challenge, however, has not received as much attention. ...
... Our framework integrates a Kalman filter-based trust estimator developed in previous work [9] and an unprecedented real-time trust calibrator. We draw inspiration from recent approaches that have provided valuable insights for the development of trust estimators [7], [8]. These approaches, however, fall short on presenting strategies for adapting the behavior of the AV and manipulating the driver's trust to, ultimately, improve the driver-AV team performance through trust calibration. ...
... When using these scales, the measurement procedure relies on users' self-reports, which have clear practical limitations when researchers are interested in tracking trust levels for real-time applications. Given these limitations, techniques for trust estimation that can take advantage of models for trust dynamics have been investigated [7], [17]- [21]. ...
Article
Full-text available
Automated vehicles (AVs) that intelligently interact with drivers must build a trustworthy relationship with them. A calibrated level of trust is fundamental for the AV and the driver to collaborate as a team. Techniques that allow AVs to perceive drivers' trust from drivers' behaviors and react accordingly are, therefore, needed for context-aware systems designed to avoid trust miscalibrations. This letter proposes a framework for the management of drivers' trust in AVs. The framework is based on the identification of trust miscalibrations (when drivers' undertrust or overtrust the AV) and on the activation of different communication styles to encourage or warn the driver when deemed necessary. Our results show that the management framework is effective, increasing (decreasing) trust of undertrusting (overtrusting) drivers, and reducing the average trust miscalibration time periods by approximately 40%. The framework is applicable for the design of SAE Level 3 automated driving systems and has the potential to improve the performance and safety of driver-AV teams.
... However, there is still much to learn about real-time trust estimation techniques as the current approaches have various limitations. Current approaches fail to provide trust measurements in scales traditionally used for trust in automation [1], or require prohibitive sophisticated sensing and perception methods [1,26]. These sophisticated methods include the processing of psychophysiological signals (e.g.: galvanic skin response), that are not practical for the vehicular environments, where driver-ADS interactions are likely to take place. ...
... However, there is still much to learn about real-time trust estimation techniques as the current approaches have various limitations. Current approaches fail to provide trust measurements in scales traditionally used for trust in automation [1], or require prohibitive sophisticated sensing and perception methods [1,26]. These sophisticated methods include the processing of psychophysiological signals (e.g.: galvanic skin response), that are not practical for the vehicular environments, where driver-ADS interactions are likely to take place. ...
... The framework is based on observable measures of drivers' behaviors and trust dynamic models. Although different trust estimation approaches have been previously reported in the literature [1,26], our method is simpler to implement. Those previous approaches represented trust as conditional probabilities. ...
Article
Full-text available
Trust miscalibration issues, represented by undertrust and overtrust, hinder the interaction between drivers and self-driving vehicles. A modern challenge for automotive engineers is to avoid these trust miscalibration issues through the development of techniques for measuring drivers' trust in the automated driving system during real-time applications execution. One possible approach for measuring trust is through modeling its dynamics and subsequently applying classical state estimation methods. This paper proposes a in automated driving systems and also for estimating these varying trust levels. The estimation method integrates sensed behaviors (from the driver) through a Kalman filter-based approach. The sensed behaviors include eye-tracking signals, the usage time of the system, and drivers' performance on a non-driving-related task (NDRT). We conducted a study (n = 80) with a simulated SAE level 3 automated driving system, and analyzed the factors that impacted drivers' trust in the system. Data from the user study were also used for the identification of the trust model parameters. Results show that the proposed approach was successful in computing trust estimates over successive interactions between the driver and the automated driving system. These results encourage the use of strategies for modeling and estimating trust in automated driving systems. Such trust measurement technique paves a path for the design of trust-aware automated driving systems capable of changing their behaviors to control drivers' trust levels to mitigate both undertrust and overtrust.
... Progress with sensing technology has resulted in the development of inexpensive and efficient psychophysiological sensors and a shift away from self-reporting scales toward more objective methods for assessing trust correlates using psychophysiological signals. Measurements using physiological traits, such as employing heart-rate variability (HRV) measurement [55], evaluation of brain activity using fMRI [56], [57], and fNIRS [58], and electroencephalogram (EEG) [59]- [63], have been used to study and evaluate trust via its psychophysiological correlates. ...
... In [59], an empirical trust sensor model was proposed using data from GSR and EEG and showed that psychophysiological signals could be used for real-time human trust measurement inferences in agents. In particular, Akash et al. [59] developed a binary classification model for trust and distrust based on the data they collected in this manner. ...
... In [59], an empirical trust sensor model was proposed using data from GSR and EEG and showed that psychophysiological signals could be used for real-time human trust measurement inferences in agents. In particular, Akash et al. [59] developed a binary classification model for trust and distrust based on the data they collected in this manner. They further expanded their experiments [66] to dynamically vary and calibrate automation transparency to optimize HMIs. ...
... Several recent efforts have made progress in using combinations of physiological (e.g., neurological and psychophysiological) measures as potential triggers. For example, Akash et al. [9] developed a system aimed at operator trust based upon a machine learning algorithm with EEG and galvanic skin response (GSR, also referred to as electrodermal activity or EDA). However, the algorithm was trained to predict actions that the operator took, either trusting the autonomous system's suggestion vs. not trusting it. ...
... On the other hand, mental workload and trust are cognitive states of the operator and thus are not immediately observable. Despite this important difference, in the literature when physiological measures are used to estimate task load or trust actions, they are often incorrectly referred to as "workload" [10] or "trust" [9]. ...
... We will investigate two classes of physiological features (see EEG example in Fig. 2). The first class includes simple, traditional features, as shown in the EEG example to be average powers of different brain regions and frequency bands, that have been extensively studied in the HRI literature [9]. The second class includes those derived via network theory [15], [16]: a network is first constructed with brain regions as nodes and the functional interactions between these regions as edges and then network-based features such as transitivity and efficiency can be extracted. ...
Conference Paper
Full-text available
This paper advocates for adaptive autonomy for future spacecraft habitats through unobtrusive monitoring of human states. We propose to estimate states with models derived from human action and physiology and adapt the system or robot's level of autonomy to improve performance and safety within human-autonomy teams. We discuss the prior work in this area and our proposed methods to incorporate specific models associated with trust, workload, and situation awareness into AI decision-making algorithms.
... A phenomenon common to all these studies pertains to the use of experimental design or tasks that involved the modulation of reliability (or dependability) offered by an automated tool (Table I, column 2). Reliability refers to how consistent an automation can perform in providing accurate information and was largely manipulated through the programming and presentation of automated agents or advisories that can provide information with different probabilities of accuracy [26], [1], [27], [3] or risk-taking tendencies [28], [29]. ...
... More importantly, with respect to the brain regions that were activated during human-automation trust-related decision making, these recent EEG studies pinpointed anterior regions such as the (i) lateral prefrontal cortex [29], (ii) anterior cingulate cortex (ACC) [1], [3], and posterior regions such as the (iii) occipital cortex [29] and (iv) fusiform gyrus (also known as the occipitotemporal gyrus) [3]. Two studies further showed that power increases or variation in the beta frequency band (12 Hz -35 Hz) were associated with increased levels of trust [27] and making discriminatory judgments between trustworthy and untrustworthy stimuli [26]. Figure 1 shows the brain regions from which trust-related decision making EEG signals were recorded. ...
... Research Questions: The review of the recent neuroscience literature on human-automation trust showed that this form trust has been largely investigated in the context of automated tools that are not 100% reliable or trustworthy [26], [1], [27], [3], [28], [29], [30], [31]. The investigation of trust in this fashion presupposes uncertainty in automated tool use and this study follows this notion in the proposed fMRI task design. ...
Research Proposal
Full-text available
The current proposal aims at a neuroscientific investigation of the magnitudes of trust air traffic controllers (ATCOs) show when using short term conflict alert (STCA) systems with different levels of reliability that can elicit high and low levels of human-automation trust. STCA is an automated warning system used by all ATCOs for the purpose of conflict detection and designed for the primary purpose of ensuring safe separation between any pair of surveillance tracks. The operational use of STCA depends a great deal on the ATCO's trust in the system, and this trust is in term dependent on the perceived system reliability. As different levels of system reliability will engender different levels of uncertainty or mistrust in its use, this proposal adopts an operational definition of trust that involves decision-making under situations with uncertainty and vulnerability. This means that human-automation trust, in the context of STCA use, relates to how well the system can facilitate successful conflict detection under circumstances where uncertainty or unreliability lies in its use. [COPYRIGHT CC-BY-NC-ND 4.0, J. Y. ZHONG 2021, NANYANG TECHNOLOGICAL UNIVERSITY] {For discussion, please message the author directly on RG.}
... We model the operator as using a stochastic reliance model that determines ADA reliance during an interaction. Reliance upon, or trust of, an autonomous aid can change quickly and responds to certain conditions, thus the reliance model used in this work aims to capture the relevant behaviors [26], [27]. We base our operator reliance model on the well known decision field theory (DFT) models from [4], [27]. ...
... Note that the reliance model described above may not apply to all operators. As shown by [4], [26], there is significant variance in how an individual interacts with a decision aid that depends on many personal characteristics. As such the parameters (see Table I) are assumed to be randomly selected from a corresponding distribution. ...
... Myopic here meaning that the decision aid only suggests the numerically optimal task a opt at every iteration regardless of the indicator model. Comparing to the myopic case allows us to better quantify the potential benefits of the task predictor and indicator model system, as opposed to the widely used advice-only decision aid [2], [26], [34]. ...
Preprint
Full-text available
In this work, we develop a game-theoretic modeling of the interaction between a human operator and an autonomous decision aid when they collaborate in a multi-agent task allocation setting. In this setting, we propose a decision aid that is designed to calibrate the operator's reliance on the aid through a sequence of interactions to improve overall human-autonomy team performance. The autonomous decision aid employs a long short-term memory (LSTM) neural network for human action prediction and a Bayesian parameter filtering method to improve future interactions, resulting in an aid that can adapt to the dynamics of human reliance. The proposed method is then tested against a large set of simulated human operators from the choice prediction competition (CPC18) data set, and shown to significantly improve human-autonomy interactions when compared to a myopic decision aid that only suggests predicted human actions without an understanding of reliance.
... The proposed method overcomes the limitations of previously published trust estimation approaches. For instance, those approaches fail to provide trust estimates in scales traditionally used for trust in automation [1], or require prohibitive sophisticated sensing and perception methods [1,72]. Those approaches are also considered overly complex, as they include the processing of psychophysiological signals (e.g., galvanic skin response) that are not practical for the vehicular environments where driver-ADS interactions take place. ...
... The proposed method overcomes the limitations of previously published trust estimation approaches. For instance, those approaches fail to provide trust estimates in scales traditionally used for trust in automation [1], or require prohibitive sophisticated sensing and perception methods [1,72]. Those approaches are also considered overly complex, as they include the processing of psychophysiological signals (e.g., galvanic skin response) that are not practical for the vehicular environments where driver-ADS interactions take place. ...
... The proposed trust estimator is the second contribution of this dissertation, and is based on observable measures of drivers' behaviors and trust dynamic models. Although different trust estimation approaches have been previously reported in the literature [1,72], the proposed trust estimation method is simpler to implement. The proposed trust estimator represents trust in a continuous numerical scale, which is consistent with Muir's scale [83] and, therefore, also consistent with the theoretical background on trust in automation. ...
Thesis
Full-text available
Trust has gained attention in the Human-Robot Interaction (HRI) field, as it is considered an antecedent of people's reliance on machines. In general, people are likely to rely on and use machines they trust, and to refrain from using machines they do not trust. Recent advances in robotic perception technologies open paths for the development of machines that can be aware of people's trust by observing their human behaviors. This dissertation explores the role of trust in the interactions between humans and robots, particularly Automated Vehicles (AVs). Novel methods and models are proposed for perceiving and processing drivers' trust in AVs and for determining both humans' natural trust and robots' artificial trust. Two high-level problems are addressed in this dissertation: (1) the problem of avoiding or reducing miscalibrations of drivers' trust in AVs, and (2) the problem of how trust can be used to dynamically allocate tasks between a human and a robot that collaborate. A complete solution is proposed for the problem of avoiding or reducing trust miscalibrations. This solution combines methods for estimating and influencing drivers' trust through interactions with the AV. Three main contributions stem from that solution: (i) the characterization of risk factors that affect drivers’ trust in AVs, which provided theoretical evidence for the development of a linear model for driver trust in AVs; (ii) the development of a new method for real-time trust estimation, which leveraged the trust linear model mentioned above for the implementation of a Kalman-filter-based approach, able to provide numerical estimates from the processing of drivers' behavioral measurements; and (iii) the development of a new method for trust calibration, which identifies trust miscalibration instances from comparisons between drivers' trust in the AV and that AV's capabilities, and triggers messages from the AV to the driver. These messages are effective for encouraging or warning drivers that are undertrusting or overtrusting the AV capabilities respectively as shown by the obtained results. Although the development of a trust-based solution for dynamically allocating tasks between a human and a robot (i.e., the second high-level problem addressed in this dissertation) remains an open problem, we take a step forward in that direction. The fourth contribution of this dissertation is the development of a unified bi-directional model for predicting natural and artificial trust. This trust model is based on mathematical representations of both the trustee agent's capabilities and the required capabilities for the execution of a task. Trust emerges from comparisons between the agent capabilities and task requirements, roughly replicating the following logic: if a trustee agent's capabilities exceed the requirements for executing a certain task, then the agent can be highly trusted (to execute that task); conversely, if that trustee agent's capabilities fall short of that task requirements, trust should be low. In this trust model, the agent's capabilities are represented by random variables that are dynamically updated over interactions between the trustor and the trustee whenever the trustee is successful or fails in the execution of a task. These capability representations allow for the numerical computation of human's trust or robot's trust, which is represented by the probability of a given trustee agent to execute a given task successfully.
... Recently, researchers have introduced methods for assessing trust using physiological measurements in different fields such as human-automation interaction (HAI), human-machine interaction (HMI), and also human-robot interaction (Wang et al. 2018;Akash et al. 2018;Hald, Rehmn, and Moeslund 2020). Measuring trust using physiological measurements can provide us with real-time assessment of trust (Ajenaghughrure et al. 2019;Khalid et al. 2016). ...
... A physiological signal that many researchers have recently used to assess the trust between humans and robots is the EEG signal. EEG had been used alone or in combination with other measurements for as assessing trust in many studies (Hu et al. 2016;Akash et al. 2018;Park, Shahrdar, and Nojoumian 2018;Huang and Nam 2020;Hald, Rehmn, and Moeslund 2020). ...
Preprint
Full-text available
In recent years a modern conceptualization of trust in human-robot interaction (HRI) was introduced by Ullman et al.\cite{ullman2018does}. This new conceptualization of trust suggested that trust between humans and robots is multidimensional, incorporating both performance aspects (i.e., similar to the trust in human-automation interaction) and moral aspects (i.e., similar to the trust in human-human interaction). But how does a robot violating each of these different aspects of trust affect human trust in a robot? How does trust in robots change when a robot commits a moral-trust violation compared to a performance-trust violation? And whether physiological signals have the potential to be used for assessing gain/loss of each of these two trust aspects in a human. We aim to design an experiment to study the effects of performance-trust violation and moral-trust violation separately in a search and rescue task. We want to see whether two failures of a robot with equal magnitudes would affect human trust differently if one failure is due to a performance-trust violation and the other is a moral-trust violation.
... For example, Khawaji, Zhou, Chen and Marcus (2015) investigated EDA as a possible indicator of interpersonal trust, showing that the EDA of users of a text-chat environment was strongly affected by trust and cognitive load. The link between trust and EDA was investigated further by Akash, Hu, Jain and Reid (2018). In this study, ...
... The authors measured participants' EEG response signals and EDA during each trial, and built a trust model based on these values. They concluded that physiological measures were promising real-time indicators of participants' trust in automation (Akash et al., 2018). In a similar study, Morris, Erno and Pilcher (2017) used a simulated automated vehicle to test participants' trust and EDA during risky and safe driving situations. ...
Thesis
Full-text available
Trust predicts the disuse and misuse of automated vehicles. While a lack of trust may induce drivers to not use the automated vehicle's functionalities, excessive trust can lead to dangerous outcomes, with drivers using the system in ways that were not originally intended by its designers. This dissertation explores new ways through which trust can be reliably measured and aligned with the true capabilities of the automated driving system.
... A phenomenon common to all these studies pertains to the use of experimental design or tasks that involved the modulation of reliability (or dependability) offered by an automated tool (Table II, column 2). Reliability refers to how consistent an automation can perform in providing accurate information and was largely manipulated through the programming and presentation of automated agents or advisories that can provide information with different probabilities of accuracy [35], [1], [36], [3] or risk-taking tendencies [37], [38]. ...
... More importantly, with respect to the brain regions that were activated during human-automation trust-related decision making, these recent EEG studies pinpointed anterior regions such as the (i) lateral prefrontal cortex [38], (ii) anterior cingulate cortex (ACC) [1], [3], and posterior regions such as the (iii) occipital cortex [38] and (iv) fusiform gyrus (also known as the occipitotemporal gyrus) [3]. Two studies further showed that power increases or variation in the beta frequency band (12 Hz -35 Hz) were associated with increased levels of trust [36] and making discriminatory judgments between trustworthy and untrustworthy stimuli [35]. Figure 2 shows the brain regions from which trust-related decision making EEG signals were recorded. ...
Technical Report
Full-text available
The current review addresses emerging issues that arise from the creation of safe, beneficial, and trusted artificial intelligence–air traffic controller (AI-ATCO) systems for air traffic management (ATM). These issues concern trust between the human user and automated or AI tools of interest, resilience, safety, and transparency. To tackle these issues, we advocate the development of practical AI-ATCO teaming frameworks by bringing together concepts and theories from neuroscience and explainable AI (XAI). By pooling together knowledge from both ATCO and AI perspectives, we seek to establish confidence in AI-enabled technologies for ATCOs. In this review, we present an overview of the extant studies that shed light on the research and development of trusted human-AI systems, and discuss the prospects of extending such works to building better trusted ATCO-AI systems. This paper contains three sections elucidating trust-related human performance, AI and explainable AI (XAI), and human-AI teaming. [COPYRIGHT CC BY-NC-ND 4.0, J. Y. ZHONG 2021, NANYANG TECHNOLOGICAL UNIVERSITY]
... Hence, trust is an essential factor in human-computer interaction because it helps users cope with scenarios of uncertainty and decision making [14,47,72]. Thus, it is essential to ensure a hitch-free interaction between users and technologies [7]. For example, prior research has demonstrated the need to foster the trust of users in technologies such as unmanned aerial vehicles (UAV), intelligent personal assistants, online shopping, and recommender agents [11,19,25,55]. ...
... The EDA features only had a couple of features for selection. This same scenario was reported in prior trust research that combined GSR and EEG signals [7]. Therefore, future research could consider the performance variation between ensemble trust classifier models trained with features from single psychophysiological signals and ensemble trust classifier models trained and tested with datasets containing features from multiple psychophysiological signals. ...
Article
Full-text available
Trust as a precursor for users' acceptance of artificial intelligence (AI) technologies that operate as a conceptual extension of humans (e.g., autonomous vehicles (AVs)) is highly influenced by users' risk perception amongst other factors. Prior studies that investigated the interplay between risk and trust perception recommended the development of real-time tools for monitoring cognitive states (e.g., trust). The primary objective of this study was to investigate a feature selection method that yields feature sets that can help develop a highly optimized and stable ensemble trust classifier model. The secondary objective of this study was to investigate how varying levels of risk perception influence users' trust and overall reliance on technology. A within-subject four-condition experiment was implemented with an AV driving game. This experiment involved 25 participants, and their electroencephalogram, electrodermal activity, and facial electromyogram psychophysiological signals were acquired. We applied wrapper, filter, and hybrid feature selection methods on the 82 features extracted from the psychophysiological signals. We trained and tested five voting-based ensemble trust classifier models using training and testing datasets containing only the features identified by the feature selection methods. The results indicate the superiority of the hybrid feature selection method over other methods in terms of model performance. In addition, the self-reported trust measurement and overall reliance of participants on the technology (AV) measured with joystick movements throughout the game reveals that a reduction in risk results in an increase in trust and overall reliance on technology.
... The fMRI study (Goodyear et al., 2016) also identified the left and right pre supplementary motor area, located around FCz, which overlaps with the regions identified in the aforementioned EEG-based trust studies. Time-domain features were predominately found significant for EEG features in the frontocentral region locations that are centered around FCz (Akash et al., 2018;Eun-Soo et al., 2019), similar to that identified in a literature review of EEG brain-computer-interface features (Lotte et al., 2007). For studies that did not manipulate reliability, there was more variance in the locations that correlate with trust conditions. ...
... neural activity was with EEG, where the remaining papers utilized fMRI. The majority of the studies included looked into artificial intelligence or computer algorithms with correct and incorrect responses at different reliability rates(Dong et al., 2015;Goodyear et al., 2016;Akash et al., 2018;de Visser et al., 2018;Wang et al., 2018;Ajenaghughrure et al., 2019;Eun-Soo et al., 2019;Pushparaj et al., 2019;Sanders et al., 2019) ...
Article
Full-text available
Investigations into physiological or neurological correlates of trust has increased in popularity due to the need for a continuous measure of trust, including for trust-sensitive or adaptive systems, measurements of trustworthiness or pain points of technology, or for human-in-the-loop cyber intrusion detection. Understanding the limitations and generalizability of the physiological responses between technology domains is important as the usefulness and relevance of results is impacted by fundamental characteristics of the technology domains, corresponding use cases, and socially acceptable behaviors of the technologies. While investigations into the neural correlates of trust in automation has grown in popularity, there is limited understanding of the neural correlates of trust, where the vast majority of current investigations are in cyber or decision aid technologies. Thus, the relevance of these correlates as a deployable measure for other domains and the robustness of the measures to varying use cases is unknown. As such, this manuscript discusses the current-state-of-knowledge in trust perceptions, factors that influence trust, and corresponding neural correlates of trust as generalizable between domains.
... The settings used to record the brainwaves were as follows: the sampling frequency was set at 256 Hz, the high-frequency filter was set at 60 Hz, and the lowfrequency filter was set at 0.1 Hz. During this EEG experiment, only the alpha (8-13 Hz), beta (13)(14)(15)(16)(17)(18)(19)(20)(21)(22)(23)(24)(25)(26)(27)(28)(29)(30), and gamma (30-60 Hz) waves, which are brainwaves related to decision-making under situations of trust and mistrust, were analyzed when the participants were asked to choose between manual or automatic control of a simulated driving game. Delta (0.2-4 Hz) and theta (4-8 Hz) waves occur when people are asleep, and were therefore not analyzed in this experiment. ...
... It is therefore necessary to research these factors further. Akash et al. [27] developed trust-sensor models, in which machines can detect psychophysiological features of trust in humans using EEG signals and the galvanic skin response. Whereas they examined how intelligent machines can detect trust in humans, the present paper examined how human trust in automation can be measured. ...
Article
Full-text available
In modem society, automation is sufficiently complex to conduct advanced tasks. The role of the human operator in controlling a complex automation is crucial for avoiding failures, reducing risk, and preventing unpredictable situations. Measuring the level of trust of human operators is vital in predicting their acceptance and reliance on automation. In this study, an electroencephalogram (EEG) is used to identify specific brainwaves under trusted and mistrusted cases of automation. A power spectrum analysis was used for a brainwave analysis. The results indicate that the power of the alpha and beta waves is stronger for a trusted situation, whereas the power of gamma waves was stronger for a mistrusted situation. When the level of human trust in automation increases, the use of automatic control increases. Therefore, the findings of this research will contribute to utilizing a neurological technology to measure the level of trust of the human operator, which can affect the decision-making and the overall performance of automation used in industries. © 2020. The Korean Institute of Intelligent Systems. All Rights Reserved.
... While we do implicate behavioral performance with communication between cognitive conflict and attentional processing mechanisms, it is not surprising that theta synchrony index this coherence. This is in light of prior work that implicates theta-band synchrony in the brain with joint/social attention, social cooperation, and interaction (Kawasaki and Yamaguchi, 2013;Dai et al., 2017;Akash et al., 2018;Wass et al., 2018;see Liu et al., 2017 for a review). While some of these studies utilized hyperscanning techniques, they still implicate inter-brain theta synchronization to these effects (Wass et al., 2018), which can shed light to the nature of the theta synchrony that we observe here. ...
... More generally, our findings follow prior work that support the idea that implicit measures like coherence between brain regions can be used to design, implement and control agents using adaptive automation (Akash et al., 2018;Kohn et al., 2021;Krueger and Wiese, 2021;Eloy et al., 2022). For example, Akash and colleagues have shown that data from the electroencephalogram (EEG) can assist in developing human trust sensors that can implicitly predict an operator's trust levels. ...
Article
Full-text available
As technological advances progress, we find ourselves in situations where we need to collaborate with artificial agents (e.g., robots, autonomous machines and virtual agents). For example, autonomous machines will be part of search and rescue missions, space exploration and decision aids during monitoring tasks (e.g., baggage-screening at the airport). Efficient communication in these scenarios would be crucial to interact fluently. While studies examined the positive and engaging effect of social signals (i.e., gaze communication) on human-robot interaction, little is known about the effects of conflicting robot signals on the human actor's cognitive load. Moreover, it is unclear from a social neuroergonomics perspective how different brain regions synchronize or communicate with one another to deal with the cognitive load induced by conflicting signals in social situations with robots. The present study asked if neural oscillations that correlate with conflict processing are observed between brain regions when participants view conflicting robot signals. Participants classified different objects based on their color after a robot (i.e., iCub), presented on a screen, simulated handing over the object to them. The robot proceeded to cue participants (with a head shift) to the correct or incorrect target location. Since prior work has shown that unexpected cues can interfere with oculomotor planning and induces conflict, we expected that conflicting robot social signals which would interfere with the execution of actions. Indeed, we found that conflicting social signals elicited neural correlates of cognitive conflict as measured by mid-brain theta oscillations. More importantly, we found higher coherence values between mid-frontal electrode locations and posterior occipital electrode locations in the theta-frequency band for incongruent vs. congruent cues, which suggests that theta-band synchronization between these two regions allows for communication between cognitive control systems and gaze-related attentional mechanisms. We also find correlations between coherence values and behavioral performance (Reaction Times), which are moderated by the congruency of the robot signal. In sum, the influence of irrelevant social signals during goal-oriented tasks can be indexed by behavioral, neural Abubshait et al. Connectivity Between Cognitive-Control and Attention in HRI oscillation and brain connectivity patterns. These data provide insights about a new measure for cognitive load, which can also be used in predicting human interaction with autonomous machines.
... As such, EDA has been used as a proxy for TiA. EDA corresponds with increased engagement with robots (Bethel et al., 2007), and the levels of EDA have been shown to be affected by level of trust (Khawaji et al., 2015), which provides support for their result use as a trust measure (e.g., Akash et al., 2018). Placing electrodes on the hands does limit the interactions that can be performed with this method, but minor task modifications or using the participant's non-dominant hand reduce those limitations. ...
... These methods capture activity in brain regions and networks for regions that correlate with trust and are therefore used as a proxy for TiA. EEG signals have indicated surprise and violation of expectations while monitoring imperfect automation or algorithms (Akash et al., 2018;de Visser et al., 2018;Somon et al., 2019) and can differentiate between human-like agents (Dong et al., 2015;Wang et al., 2018;Jung et al., 2019). fMRI signals have differentiated brain regions and networks associated with observation of errors with machines (Desmet et al., 2014), the tendency to comply with automation (Goodyear et al., 2016, and differences between human-human trust and humanmachine trust (Riedl et al., 2011(Riedl et al., , 2014. ...
Article
Full-text available
With the rise of automated and autonomous agents, research examining Trust in Automation (TiA) has attracted considerable attention over the last few decades. Trust is a rich and complex construct which has sparked a multitude of measures and approaches to study and understand it. This comprehensive narrative review addresses known methods that have been used to capture TiA. We examined measurements deployed in existing empirical works, categorized those measures into self-report, behavioral, and physiological indices, and examined them within the context of an existing model of trust. The resulting work provides a reference guide for researchers, providing a list of available TiA measurement methods along with the model-derived constructs that they capture including judgments of trustworthiness, trust attitudes, and trusting behaviors. The article concludes with recommendations on how to improve the current state of TiA measurement.
... Furthermore, researchers investigating users trust assessment in real-time (i.e., during interaction) using single (electroencephalogram (EEG), functional near infrared spectroscopy (FNIRS), and electrodermal activity (EDA)) and multi-modal ( EEG+EDA, audio/speech+ photoplethysmography + video ) psychophysiological signals has developed fairly accurate classifier models that are capable of detecting users trust state from psychophysiological signals during interaction with AI-based systems. (Ajenaghughrure et al., 2019;Hirshfield et al., 2011;Shafiei et al., 2018;Lochner et al., 2019;Akash et al., 2018;Hu et al., 2016;). ...
... For instance, Hu et al., (2016), despite extracting 108 features from the psychophysiological signals (EEG 105, EDA 3), the model utilized more EEG features (8) and less EDA features (2). Also, Akash et al., (2018), despite extracting 147 EEG features and 2 EDA features, both models (general and customized) used more EEG features (11 and 10) than EDA features (1 and 2). Furthermore, though the resulting model developed by Khalid et al., (2018) utilized features extracted (facial action code units, photoplethysmography (video-heart rate), audio/speech) from video and audio/speech psychophysiological signals, no details of the numbers of selected features per signal was provided. ...
... Researchers have proposed to develop paradigms that anticipate human interaction behaviors-such as trust in automation-and influence humans to make optimal choices about automation use [1,17,29,38]. Pre-requisites for such an approach involve the capability to quantitatively predict human behavior and an algorithm for determining the optimal intervention to influence human behavior. Chen et al. [11] has modeled human trust dynamics on a table-clearing task, and adjusted manipulation robot's control behavior considering human trust status. ...
... In this setting, we believe trust calibration is especially important because of two reasons: 1) the system reliability (at least user's perceived system reliability) is affected by the scene complexity 2) user trust to the system might change frequently depending on system reliability and traffic condition given that most users are not trained experts. We develop a probabilistic model of the user trust and workload dynamics using human 1 Following those previous studies we define transparency as "the descriptive quality of an interface pertaining to its abilities to afford an operator's comprehension about an intelligent agent's intent, performance, future plans, and reasoning process" [10]. subject data collected using a driving simulator for urban driving scenes. ...
... Galvanic skin response captures physiological arousal levels based on the conductivity of the surface of the skin (Akash, Hu, Jain, & Reid, 2018). GSR is significantly affected by both trust and cognitive load in a text-chat environment (Khawaji, Zhou, Chen, & Marcus, 2015), and the phasic component of GSR was also a significant predictor of trust levels (Akash et al., 2018), but the results were not clear about the corresponding GSR readings for a high level of trust. ...
... Galvanic skin response captures physiological arousal levels based on the conductivity of the surface of the skin (Akash, Hu, Jain, & Reid, 2018). GSR is significantly affected by both trust and cognitive load in a text-chat environment (Khawaji, Zhou, Chen, & Marcus, 2015), and the phasic component of GSR was also a significant predictor of trust levels (Akash et al., 2018), but the results were not clear about the corresponding GSR readings for a high level of trust. ...
Chapter
Any functional human-AI-robot team consists of multiple stakeholders, as well as one or more artificial agents (e.g., AI agents and embodied robotic agents). Each stakeholder's trust in the artificial agent matters because it not only impacts their performance on tasks with human teammates and artificial agents but also influences their trust in other stakeholders and how other stakeholders trust the artificial agents. Interpersonal trust and human-agent trust mutually influence each other. Traditional measures of trust in human-robot interactions have been focused on one end-user’s trust in one artificial agent rather than investigating the team level of trust that involves all relevant stakeholders and the interactions among these entities. Also, traditional measures of trust have been mostly static, unable to capture the distributed trust dynamics at a team level. To fill this gap, this chapter proposes a distributed dynamic team trust (D2T2) framework and potential measures for its applications in human-AI-robot teaming.
... Researchers have proposed to develop paradigms that anticipate human interaction behaviors-such as trust in automation-and influence humans to make optimal choices about automation use [1,17,29,38]. Pre-requisites for such an approach involve the capability to quantitatively predict human behavior and an algorithm for determining the optimal intervention to influence human behavior. Chen et al. [11] has modeled human trust dynamics on a table-clearing task, and adjusted manipulation robot's control behavior considering human trust status. ...
... In this setting, we believe trust calibration is especially important because of two reasons: 1) the system reliability (at least user's perceived system reliability) is affected by the scene complexity 2) user trust to the system might change frequently depending on system reliability and traffic condition given that most users are not trained experts. We develop a probabilistic model of the user trust and workload dynamics using human subject data collected using a driving simulator for urban driving 1 Following those previous studies we define transparency as "the descriptive quality of an interface pertaining to its abilities to afford an operator's comprehension about an intelligent agent's intent, performance, future plans, and reasoning process" [10]. scenes. ...
Preprint
Full-text available
Properly calibrated human trust is essential for successful interaction between humans and automation. However, while human trust calibration can be improved by increased automation transparency, too much transparency can overwhelm human workload. To address this tradeoff, we present a probabilistic framework using a partially observable Markov decision process (POMDP) for modeling the coupled trust-workload dynamics of human behavior in an action-automation context. We specifically consider hands-off Level 2 driving automation in a city environment involving multiple intersections where the human chooses whether or not to rely on the automation. We consider automation reliability, automation transparency, and scene complexity, along with human reliance and eye-gaze behavior, to model the dynamics of human trust and workload. We demonstrate that our model framework can appropriately vary automation transparency based on real-time human trust and workload belief estimates to achieve trust calibration.
... More recently, there has been a trend to move away from self-report measures towards more "objective" measures of trust [16,17,[86][87][88]. These include physiological measures such as eye-tracking [89,90], social-cues extracted from video feed/cameras [91,92], audio [93][94][95], skin response [96,97], and neural measures [96][97][98][99], as well as play behavior in behavioral economic games [37,92,100]. ...
... More recently, there has been a trend to move away from self-report measures towards more "objective" measures of trust [16,17,[86][87][88]. These include physiological measures such as eye-tracking [89,90], social-cues extracted from video feed/cameras [91,92], audio [93][94][95], skin response [96,97], and neural measures [96][97][98][99], as well as play behavior in behavioral economic games [37,92,100]. ...
Article
Full-text available
Purpose of Review To assess the state-of-the-art in research on trust in robots and to examine if recent methodological advances can aid in the development of trustworthy robots. Recent Findings While traditional work in trustworthy robotics has focused on studying the antecedents and consequences of trust in robots, recent work has gravitated towards the development of strategies for robots to actively gain, calibrate, and maintain the human user’s trust. Among these works, there is emphasis on endowing robotic agents with reasoning capabilities (e.g., via probabilistic modeling). Summary The state-of-the-art in trust research provides roboticists with a large trove of tools to develop trustworthy robots. However, challenges remain when it comes to trust in real-world human-robot interaction (HRI) settings: there exist outstanding issues in trust measurement, guarantees on robot behavior (e.g., with respect to user privacy), and handling rich multidimensional data. We examine how recent advances in psychometrics, trustworthy systems, robot-ethics, and deep learning can provide resolution to each of these issues. In conclusion, we are of the opinion that these methodological advances could pave the way for the creation of truly autonomous, trustworthy social robots.
... For instance, self-reporting survey instruments, such as those developed in [10][11][12], that are predominantly used for measuring trust are subjective and not objective. Thus, it is not possible to use self-reporting instruments to assess trust in real-time because only after completing one or more tasks/interactions are such instruments applied [13,14]. ...
... This development has aided researchers in achieving groundbreaking discoveries. For example, in the context of human-computer interaction (HCI), researchers have developed intelligent interfaces that enable the adaptation of machines to a user's trust state using one or more physiological signals [13,30]. ...
Article
Full-text available
Trust plays an essential role in all human relationships. However, measuring trust remains a challenge for researchers exploring psychophysiological signals. Therefore, this article aims to systematically map the approaches used in studies assessing trust with psychophysiological signals. In particular, we examine the numbers and frequency of combined psychophysiological signals, the primary outcomes of previous studies, and the types and most commonly used data analysis techniques for analyzing psychophysiological data to infer a trust state. For this purpose, we employ a systematic mapping review method, through which we analyze 51 carefully selected articles (studies focused on trust using psychophysiology). Two significant findings are as follows: (1) Psychophysiological signals from EEG(electroencephalogram) and ECG(electrocardiogram) for monitoring peripheral and central nervous systems are the most frequently used to measure trust, while audio and EOG(electro-oculography) psychophysiological signals are the least commonly used. Moreover, the maximum number of psychophysiological signals ever combined so far is three (3). Most of which are peripheral nervous system monitoring psychophysiological signals that are low in spatial resolution. (2) Regarding outcomes: there is only one tool proposed for assessing trust in an interpersonal context, excluding trust in a technology context. Moreover, there are no stable and accurate ensemble models that have been developed to assess trust; all prior attempts led to unstable but fairly accurate models or did not satisfy the conditions for combining several algorithms (ensemble). In conclusion, the extent to which trust can be assessed using psychophysiological measures during user interactions (real-time) remains unknown, as there several issues, such as the lack of a stable and accurate ensemble trust classifier model, among others, that require urgent research attention. Although this topic is relatively new, much work has been done. However, more remains to be done to provide clarity on this topic.
... 2) Trial-Based Trust Ratings: After each junction trial, participants were asked "To what extent do you trust the autonomous vehicle?". Similar single-item trust measures have been used in past research [26]. To avoid lengthiness of questions during the simulation, before the study, it was clarified that trust ratings are to be based on the current trial, and not on the accumulative AV experience as a whole. ...
... Third, the study focused on drivers' trust-in-AV based on current trial experience, so as to study brain signals which are modulated experimentally as a function of trial conditions (i.e. normal vs malfunction, in CAD and FAD); this experimental design that compares normal vs malfunctions is also why similar frequencies of these trial types were incorporated in the study (see [26]). Having shed light on the relevant brain metrics (e.g. ...
Article
Trust in autonomous vehicles (AV) is of critical importance and demands comprehensive interdisciplinary research. While most studies utilize subjective measures, we employ electroencephalography (EEG) to study in a more objective manner the cognitive states associated with trust during AV driving. Subjects drove a simulated AV in Conditional Automation Driving (SAE L3) and Full Automation Driving (SAE L5) modes. In the experimental design, malfunctions were induced at both automation levels. Self-reported trust in the AV was reduced immediately after Full Automation malfunctions, but not after Conditional Automation malfunctions when subjects were able to take over vehicle control to avoid danger. EEG analyses reveal that during Full Automation malfunctions, there was a significant enhancement in approach motivation (i.e. desire to re-engage) and a disruption of right frontal functional clustering that supports executive cognition (i.e. planning and decision-making). No neurocognitive disruptions were observed during Conditional Automation malfunctions. Our results demonstrate that it is not automation malfunctions per se (e.g. failure to decelerate) that deteriorate trust, but rather the inability for human drivers to adaptively mitigate the risk of negative outcomes (e.g. risk of crashing) resulting from those malfunctions. This is reflected in changes in brain activity associated with motivational state and action planning. Keeping the human driver on-the-loop protects against trust loss. Frontal alpha EEG is a neural correlate of trust-in-automation, with potential for trust monitoring using wearable technology to support driver-vehicle adaptivity.
... Physiological sensing technology has the potential to detect and monitor aforementioned issues that affect human performance. Although work demonstrating the link between physiological metrics and constructs such as trust (Akash et al. 2018;Dong et al. 2015;Fu et al. 2018;M. Wang et al. 2018) and situation awareness (Catherwood et al. 2014;Kaur et al. 2019;Yeo et al. 2017;Zhang et al. 2021) is still developing, the ability of physiological sensors to measure workload has been well-established. ...
Article
Full-text available
Remotely piloted aircraft (RPA) operations are often characterized as highly taxing and dynamic. Physiological sensing technology can enhance personnel monitoring and training for these high-stress environments; however, work assessing the effectiveness of physiological sensors during RPA operations is limited. The proposed work tests two hypotheses: (1) physiological sensors can distinguish operator workload between scenario difficulty levels, and (2) the sensors can quantify the impact of RPA events on the operator workload. Twelve pilots completed RPA simulations at all three difficulty levels while physiological sensors collected electroencephalogram (EEG) and heart rate activity. Hypotheses were tested using mixed-effects models. Observed heart rate variability metrics did not differ among the three scenario difficulty levels except for LF/HF ratio. A 47% and 57% reduction in alpha band power was observed between easy and hard difficulty levels for the frontal and parietal channels, respectively. Abort and Reach objective events resulted in 0.2–0.3 dB lower beta activity and 66 ms increased heart rate, while losing sight of the objective (e.g., fog) had 0.72 dB increased high beta activity. Different physiological modalities (EEG and ECG) had varying effectiveness in distinguishing scenario difficulty and RPA events, suggesting a hybrid sensing approach may provide more insight than just using one modality. In conclusion, physiological sensing can distinguish operator response to scenario difficulty and events in high-fidelity RPA simulations.
... Although some work had been done in the problem of estimating human's trust in a robot [1,18], no method was entirely appropriate for the driver-AV context. The challenge was to combine sensors to monitor the driver's behaviors that were adequate to be used in the vehicular environment with mathematical models representing the dynamics of trust over the interactions between the driver and the AV. ...
Preprint
Full-text available
Advances in perception and artificial intelligence technology are expected to lead to seamless interaction between humans and robots. Trust in robots has been evolving from the theory on trust in automation, with a fundamental difference: unlike traditional automation, robots could adjust their behaviors depending on how their human counterparts appear to be trusting them or how humans appear to be trustworthy. In this extended abstract I present my research on methods for processing trust in the particular context of interactions between a driver and an automated vehicle, which has the goal of achieving higher safety and performance standards for the team formed by those human and robotic agents.
... On their simulated trust dataset, they obtained 96.61% accuracy. Akash et al. (2018) developed an empirical trust model of object detection in AVs and they used a quadratic discriminant classifier and psychophysiological measurements, such as electroencephalography (EEG) and galvanic skin response (GSR). Their model's best accuracy was 78.55%. ...
Article
Full-text available
Technological advances in the automotive industry are bringing automated driving closer to road use. However, one of the most important factors affecting public acceptance of automated vehicles (AVs) is the public’s trust in AVs. Many factors can influence people’s trust, including perception of risks and benefits, feelings, and knowledge of AVs. This study aims to use these factors to predict people’s dispositional and initial learned trust in AVs using a survey study conducted with 1175 participants. For each participant, 23 features were extracted from the survey questions to capture his/her knowledge, perception, experience, behavioral assessment, and feelings about AVs. These features were then used as input to train an eXtreme Gradient Boosting (XGBoost) model to predict trust in AVs. With the help of SHapley Additive exPlanations (SHAP), we were able to interpret the trust predictions of XGBoost to further improve the explainability of the XGBoost model. Compared to traditional regression models and black-box machine learning models, our findings show that this approach was powerful in providing a high level of explainability and predictability of trust in AVs, simultaneously.
... On their simulated trust dataset, they obtained 96.61% accuracy. Akash et al. (2018) developed an empirical trust model of object detection in AVs and they used a quadratic discriminant classifier and psychophysiological measurements, such as electroencephalography (EEG) and galvanic skin response (GSR). Their model's best accuracy was 78.55%. ...
Preprint
Full-text available
Technological advances in the automotive industry are bringing automated driving closer to road use. However, one of the most important factors affecting public acceptance of automated vehicles (AVs) is the public's trust in AVs. Many factors can influence people's trust, including perception of risks and benefits, feelings, and knowledge of AVs. This study aims to use these factors to predict people's dispositional and initial learned trust in AVs using a survey study conducted with 1175 participants. For each participant, 23 features were extracted from the survey questions to capture his or her knowledge, perception, experience, behavioral assessment, and feelings about AVs. These features were then used as input to train an eXtreme Gradient Boosting (XGBoost) model to predict trust in AVs. With the help of SHapley Additive exPlanations (SHAP), we were able to interpret the trust predictions of XGBoost to further improve the explainability of the XGBoost model. Compared to traditional regression models and black-box machine learning models, our findings show that this approach was powerful in providing a high level of explainability and predictability of trust in AVs, simultaneously.
... The contents of this section were previously published by Akash, Hu, Jain, and Reid in ACM Transactions on Interactive Intelligent Systems [102] and are reported here with minor modifications. ...
Thesis
Full-text available
Intelligent machines, and more broadly, intelligent systems, are becoming increasingly common in the everyday lives of humans. Nonetheless, despite significant advancements in automation, human supervision and intervention are still essential in almost all sectors, ranging from manufacturing and transportation to disaster-management and healthcare. These intelligent machines interact and collaborate with humans in a way that demands a greater level of trust between human and machine. While a lack of trust can lead to a human's disuse of automation, over-trust can result in a human trusting a faulty autonomous system which could have negative consequences for the human. Therefore, human trust should be calibrated to optimize these human-machine interactions. This calibration can be achieved by designing human-aware automation that can infer human behavior and respond accordingly in real-time. In this dissertation, I present a probabilistic framework to model and calibrate a human's trust and workload dynamics during his/her interaction with an intelligent decision-aid system. More specifically, I develop multiple quantitative models of human trust, ranging from a classical state-space model to a classification model based on machine learning techniques. Both models are parameterized using data collected through human-subject experiments. Thereafter, I present a probabilistic dynamic model to capture the dynamics of human trust along with human workload. This model is used to synthesize optimal control policies aimed at improving context-specific performance objectives that vary automation transparency based on human state estimation. I also analyze the coupled interactions between human trust and workload to strengthen the model framework. Finally, I validate the optimal control policies using closed-loop human subject experiments. The proposed framework provides a foundation toward widespread design and implementation of real-time adaptive automation based on human states for use in human-machine interactions.
... Walker et al., 2019;Lu & Sarter, 2019) or physiological measures such as EEG and GSR (e.g. Akash et al., 2018;Wang et al., 2018;Gupta et al., 2020). Although these measures could be used for dynamic tracking they require special hardware, and might suffer from ambiguities and confounders due to the indirect measurement and in many use cases cannot be used for life-long calibration of trust. ...
Preprint
Explainability, interpretability and how much they affect human trust in AI systems are ultimately problems of human cognition as much as machine learning, yet the effectiveness of AI recommendations and the trust afforded by end-users are typically not evaluated quantitatively. We developed and validated a general purpose Human-AI interaction paradigm which quantifies the impact of AI recommendations on human decisions. In our paradigm we confronted human users with quantitative prediction tasks: asking them for a first response, before confronting them with an AI's recommendations (and explanation), and then asking the human user to provide an updated final response. The difference between final and first responses constitutes the shift or sway in the human decision which we use as metric of the AI's recommendation impact on the human, representing the trust they place on the AI. We evaluated this paradigm on hundreds of users through Amazon Mechanical Turk using a multi-branched experiment confronting users with good/poor AI systems that had good, poor or no explainability. Our proof-of-principle paradigm allows one to quantitatively compare the rapidly growing set of XAI/IAI approaches in terms of their effect on the end-user and opens up the possibility of (machine) learning trust.
...  evaluation of trust in autonomy (Akash et al., 2018);  detection and correction of robot errors and misunderstanding of human gestures (Kim et al., 2017;Krol & Zander, 2017);  prediction of pilot auditory error for adaptive cockpits (Dehais et al., 2019);  detection of severity and type of system errors perceived by the human (Wirth et al., 2019). ...
Conference Paper
Full-text available
A new dawn of intelligent machines has re-energised the concept of human-machine teaming (HMT) whereby humans, and autonomous systems, collaborate towards a shared operational goal. Across Defence, Human Factors specialists will be challenged to integrate human-autonomy teams into already complex systems for which knowing the functional state of human teammates will be critical to system optimisation. Presently, innovation in machine learning and data collection methods is making human cognition more available to operational settings than ever before. This paper overviews the state of the art in techniques for estimating human functional state from the perspective of designing complex military systems involving artificially intelligent (AI) agents. Considerations are provided for designers seeking to quantify variables such as mental workload, situation awareness (SA) or the level of demand upon particular communication modes, whether for system operation or design and evaluation. Finally, some examples of methods used in HMT research are presented along with a speculative look at future influencesupon the specification of human functional statefor use with autonomy in Defence.
... Physiological sensing technology has the potential to detect and monitor aforementioned issues that affect human perfor-mance. Although work demonstrating the link between phys-iological metrics and constructs such as trust (Akash et al. 2018;Dong et al. 2015;Fu et al. 2018;M. Wang et al. 2018) and situation awareness (Catherwood et al. 2014;Kaur et al. 2019;Yeo et al. 2017;Zhang et al. 2021) is still developing, the ability of physiological sensors to measure workload has been well-established. ...
Article
Full-text available
Remotely piloted aircraft (RPA) operations are often characterized as highly taxing and dynamic. Physiological sensing tech- nology can enhance personnel monitoring and training for these high-stress environments; however, work assessing the effec- tiveness of physiological sensors during RPA operations is limited. The proposed work tests two hypotheses: (1) physiological sensors can distinguish operator workload between scenario difficulty levels, and (2) the sensors can quantify the impact of RPA events on the operator workload. Twelve pilots completed RPA simulations at all three difficulty levels while physiological sensors collected electroencephalogram (EEG) and heart rate activity. Hypotheses were tested using mixed-effects models. Observed heart rate variability metrics did not differ among the three scenario difficulty levels except for LF/HF ratio. A 47% and 57% reduction in alpha band power was observed between easy and hard difficulty levels for the frontal and parietal channels, respectively. Abort and Reach objective events resulted in 0.2–0.3 dB lower beta activity and 66 ms increased heart rate, while losing sight of the objective (e.g., fog) had 0.72 dB increased high beta activity. Different physiological modalities (EEG and ECG) had varying effectiveness in distinguishing scenario difficulty and RPA events, suggesting a hybrid sensing approach may provide more insight than just using one modality. In conclusion, physiological sensing can distinguish operator response to scenario difficulty and events in high-fidelity RPA simulations.
... The experiment was carried out using virtual reality as a human-machine interface to convey situational information, which is shown to improve trust in vehicle automation. Akash et al. [28] proposed a customized set of psychophysiological features for each individual to build a classifierbased trust-sensor model, which has investigated the trust estimation in real-time. Nevertheless, the psychophysiological measurements are intrusive and impractical in real-world implementation. ...
Preprint
Full-text available
Trust calibration is necessary to ensure appropriate user acceptance in advanced automation technologies. A significant challenge to achieve trust calibration is to quantitatively estimate human trust in real-time. Although multiple trust models exist, these models have limited predictive performance partly due to individual differences in trust dynamics. A personalized model for each person can address this issue, but it requires a significant amount of data for each user. We present a methodology to develop customized model by clustering humans based on their trust dynamics. The clustering-based method addresses the individual differences in trust dynamics while requiring significantly less data than personalized model. We show that our clustering-based customized models not only outperform the general model based on entire population, but also outperform simple demographic factor-based customized models. Specifically, we propose that two models based on ``confident'' and ``skeptical'' group of participants, respectively, can represent the trust behavior of the population. The ``confident'' participants, as compared to the ``skeptical'' participants, have higher initial trust levels, lose trust slower when they encounter low reliability operations, and have higher trust levels during trust-repair after the low reliability operations. In summary, clustering-based customized models improve trust prediction performance for further trust calibration considerations.
... The experiment was carried out using virtual reality as a human-machine interface to convey situational information, which is shown to improve trust in vehicle automation. Akash et al. [28] proposed a customized set of psychophysiological features for each individual to build a classifierbased trust-sensor model, which has investigated the trust estimation in real-time. Nevertheless, the psychophysiological measurements are intrusive and impractical in real-world implementation. ...
... • evaluation of trust in autonomy (Akash et al., 2018); • detection and correction of robot errors and misunderstanding of human gestures (Kim et al., 2017;Krol & Zander, 2017); • prediction of pilot auditory error for adaptive cockpits (Dehais et al., 2019); • detection of severity and type of system errors perceived by the human (Wirth et al., 2019). ...
... However, little work to date has allowed for real-time monitoring of human trust. The approaches that would allow for real-time monitoring rely on expensive and cumbersome physiological monitor equipment (Akash et al., 2018;de Visser et al., 2018;Hergeth et al., 2016;Jung et al., 2019;Tenhundfeld et al., 2019;Walker et al., 2018). ...
Preprint
Full-text available
The main goal of this study is to empirically ascertain team communication behaviors and their relationship with team performance in Human-Autonomy teams (HAT)s and human-human teams in a simulated dynamic task environment. To do so, we address the following research questions: (1) how strongly each team within each condition is representable by their communication behavior, and (2) which of those communication behaviors within each condition are associated with team performance similarly or differently? There were three heterogeneous and interdependent teammates: navigator, pilot, and photographer. Each teammate had a unique role relevant to the team’s objective of efficiently taking good photographs of target waypoints. We applied the following data mining methods: K-means clustering to determine clustered communication behaviors of pre-specified conditions; confusion matrix to evaluate the accuracy of the clusters of communication behaviors; stepwise regression to identify a subset of communication behaviors within each condition that was associated with team performance. The findings from this study are: (1) some team communication behaviors were classified in the three conditions, and the accuracy of this classification was strong; (2) condition membership moderated the relationship between team communication behaviors and performance.
... Feature extraction has been carried out on the raw sensor signals from the accelerometer and gyroscope. The time-domain features, frequency domain features, and descriptive features were extracted from each axis x, y, and z of both sensors based on similar work carried out in earlier studies [3,14,35,42,57]. The sensor data is of streaming nature and an appropriate method was required through which features can be extracted in real-time. ...
Article
Full-text available
Smoking cessation efforts can be greatly influenced by providing just-in-time intervention to individuals who are trying to quit smoking. Detecting smoking activity accurately among the confounding activities of daily living (ADLs) being monitored by the wearable device is a challenging and intriguing research problem. This study aims to develop a machine learning based modeling framework to identify the smoking activity among the confounding ADLs in real-time using the streaming data from the wrist-wearable IMU (6-axis inertial measurement unit) sensor. A low-cost wrist-wearable device has been designed and developed to collect raw sensor data from subjects for the activities. A sliding window mechanism has been used to process the streaming raw sensor data and extract several time-domain, frequency-domain, and descriptive features. Hyperparameter tuning and feature selection have been done to identify best hyperparameters and features respectively. Subsequently, multi-class classification models are developed and validated using in-sample and out-of-sample testing. The developed models obtained predictive accuracy (area under receiver operating curve) up to 98.7% for predicting the smoking activity. The findings of this study will lead to a novel application of wearable devices to accurately detect smoking activity in real-time. It will further help the healthcare professionals in monitoring their patients who are smokers by providing just-in-time intervention to help them quit smoking. The application of this framework can be extended to more preventive healthcare use-cases and detection of other activities of interest. Supplementary information: The online version contains supplementary material available at 10.1007/s11042-022-12349-6.
... For example, human workers' trust in their robot colleagues influences their cooperation efficiency. DNN's potential to extract the human behavior pattern can also investigate the human trust model in machines (Akash et al., 2018). ...
Article
Retired electric-vehicle lithium-ion battery (EV-LIB) packs pose severe environmental hazards. Efficient recovery of these spent batteries is a significant way to achieve closed-loop lifecycle management and a green circular economy. It is crucial for carbon neutralization, and for coping with the environmental and resource challenges associated with the energy transition. EV-LIB disassembly is recognized as a critical bottleneck for mass-scale recycling. Automated disassembly of EV-LIBs is extremely challenging due to the large variety and uncertainty of retired EV-LIBs. Recent advances in artificial intelligence (AI) machine learning (ML) provide new ways for addressing these problems. This study aims to provide a systematic review and forward-looking perspective on how AI/ML methodology can significantly boost EV-LIB intelligent disassembly for achieving sustainable recovery. This work examines the key advances and research opportunities of emerging intelligent technologies for EV-LIB disassembly, and recycling and reuse of industrial products in general. We show that AI could benefit the whole disassembly process, particularly addressing the uncertainty and safety issues. Currently, EV-LIB state prognostics, disassembly decision-making as well as target detection are indicated as promising areas to realize intelligence. The challenges still exist for extensive autonomy due to present AI's inherent limitations, mechanical and chemical complexities, and sustainable benefits concerns. This paper provides the practical map to direct how to implement EV-LIB intelligent disassembly as well as forward-looking perspectives for addressing these challenges.
... Few studies have attempted to classify trust in automated agents using neurophysiological signals in naturalistic settings. Those that do exist employ very controlled but hyper-specific tasks that reduce the participants' potential decisions and freedom in approaching the task (Akash et al., 2018;Wang et al., 2018;Gupta et al., 2019). While promising, these studies are largely unable to draw any clear functional neuroergonomic conclusions from their results, highlighting the need for more robust, ecologically valid, and generalizable studies of neurophysiological modeling of human-agent trust. ...
Article
Full-text available
Intelligent agents are rapidly evolving from assistants into teammates as they perform increasingly complex tasks. Successful human-agent teams leverage the computational power and sensory capabilities of automated agents while keeping the human operator's expectation consistent with the agent's ability. This helps prevent over-reliance on and under-utilization of the agent to optimize its effectiveness. Research at the intersection of human-computer interaction, social psychology, and neuroergonomics has identified trust as a governing factor of human-agent interactions that can be modulated to maintain an appropriate expectation. To achieve this calibration, trust can be monitored continuously and unobtrusively using neurophysiological sensors. While prior studies have demonstrated the potential of functional near-infrared spectroscopy (fNIRS), a lightweight neuroimaging technology, in the prediction of social, cognitive, and affective states, few have successfully used it to measure complex social constructs like trust in artificial agents. Even fewer studies have examined the dynamics of hybrid teams of more than 1 human or 1 agent. We address this gap by developing a highly collaborative task that requires knowledge sharing within teams of 2 humans and 1 agent. Using brain data obtained with fNIRS sensors, we aim to identify brain regions sensitive to changes in agent behavior on a long- and short-term scale. We manipulated agent reliability and transparency while measuring trust, mental demand, team processes, and affect. Transparency and reliability levels are found to significantly affect trust in the agent, while transparency explanations do not impact mental demand. Reducing agent communication is shown to disrupt interpersonal trust and team cohesion, suggesting similar dynamics as human-human teams. Contrasts of General Linear Model analyses identify dorsal medial prefrontal cortex activation specific to assessing the agent's transparency explanations and characterize increases in mental demand as signaled by dorsal lateral prefrontal cortex and frontopolar activation. Short scale event-level data is analyzed to show that predicting whether an individual will trust the agent, with data from 15 s before their decision, is feasible with fNIRS data. Discussing our results, we identify targets and directions for future neuroergonomics research as a step toward building an intelligent trust-modulation system to optimize human-agent collaborations in real time.
Article
Proximate human-robot teaming (pxHRT) is a complex subspace within human-robot interaction. Studies in this space involve a range of equipment and methods, including the ability to sense people and robots precisely. Research in this area draws from a wide variety of other fields, from human-human interaction to control theory, making study design complex, particularly for those outside the field of HRI. In this paper, we introduce a framework that helps researchers consider tradeoffs across various task contexts, platforms, sensors, and analysis methods; metrics frequently used in the field; and common challenges researchers may face. We demonstrate the use of the framework via a case study which employs an autonomous mobile manipulator continuously engaging in shared workspace, handover, and co-manipulation tasks with people, and explores the effect of cognitive workload on pxHRT dynamics. We also demonstrate the utility of the framework in a case study with two groups of researchers new to pxHRT. With this framework, we hope to enable researchers, especially those outside HRI, to more thoroughly consider these complex components within their studies, more easily design experiments, and more fully explore research questions within the space of pxHRT.
Article
Background Herd immunity or community immunity refers to the reduced risk of infection among susceptible individuals in a population through the presence and proximity of immune individuals. Recent studies suggest that improving the understanding of community immunity may increase intentions to get vaccinated. Objective This study aims to design a web application about community immunity and optimize it based on users’ cognitive and emotional responses. Methods Our multidisciplinary team developed a web application about community immunity to communicate epidemiological evidence in a personalized way. In our application, people build their own community by creating an avatar representing themselves and 8 other avatars representing people around them, for example, their family or coworkers. The application integrates these avatars in a 2-min visualization showing how different parameters (eg, vaccine coverage, and contact within communities) influence community immunity. We predefined communication goals, created prototype visualizations, and tested four iterative versions of our visualization in a university-based human-computer interaction laboratory and community-based settings (a cafeteria, two shopping malls, and a public library). Data included psychophysiological measures (eye tracking, galvanic skin response, facial emotion recognition, and electroencephalogram) to assess participants’ cognitive and affective responses to the visualization and verbal feedback to assess their interpretations of the visualization’s content and messaging. Results Among 110 participants across all four cycles, 68 (61.8%) were women and 38 (34.5%) were men (4/110, 3.6%; not reported), with a mean age of 38 (SD 17) years. More than half (65/110, 59.0%) of participants reported having a university-level education. Iterative changes across the cycles included adding the ability for users to create their own avatars, specific signals about who was represented by the different avatars, using color and movement to indicate protection or lack of protection from infectious disease, and changes to terminology to ensure clarity for people with varying educational backgrounds. Overall, we observed 3 generalizable findings. First, visualization does indeed appear to be a promising medium for conveying what community immunity is and how it works. Second, by involving multiple users in an iterative design process, it is possible to create a short and simple visualization that clearly conveys a complex topic. Finally, evaluating users’ emotional responses during the design process, in addition to their cognitive responses, offers insights that help inform the final design of an intervention. Conclusions Visualization with personalized avatars may help people understand their individual roles in population health. Our app showed promise as a method of communicating the relationship between individual behavior and community health. The next steps will include assessing the effects of the application on risk perception, knowledge, and vaccination intentions in a randomized controlled trial. This study offers a potential road map for designing health communication materials for complex topics such as community immunity.
Article
Full-text available
: The present study aimed to compare the effect of metacognitive therapy-based group intervention and group acceptance-based behavioral therapy on psychophysiological signs of professional soccer players in the U-19 league in Tehran. Participants were professional soccer players occupied in professional soccer leagues in Tehran. The participants were entered into the assessment stage, and after obtaining informed consent, they were randomly assigned to one of the three experiment groups, namely MCT, MAC, and WL. The participants’ psychophysiological signs included EEG, EMG, HR, GSR, temperature, and RR, which were recorded using the ewave 8-channel neuro-biofeedback device. The data were analyzed using the eProbe7.8.3 software of Rubymind.us. The results demonstrated that MCT and MAC could make some changes in psychophysiological signs of anxious soccer players. MCT was shown to affect Fz Highbeta-Gama and RR, which had a correlation with anxiety. In addition, MAC was observed to affect the asymmetry of F3 alpha and F4 alpha as a remarkable EEG pattern of aggression. However, MCT and MAC did not show any effect on HR, EMG, temperature, and GSR. According to these findings, it can be inferred that those soccer players who are uncomfortable and suffer from anxiety and aggression-related problems may benefit from such interventions.
Article
Trust miscalibration remains a major challenge for human-machine interaction. It can lead to misuse or disuse of automated systems. To date, most trust research has relied on subjective ratings and behavioral or physiological data to assess trust. Those trust measurements are discrete, disruptive and quite difficult to implement. To better understand the process of trust calibration, we propose eye tracking as an unobtrusive method for inferring trust levels in real time. Using an Unmanned Aerial Vehicle simulation, participants were exposed to varying levels of reliability of an automated target detection system. Eye movement data were captured and labeled as high or low trust based on subjective trust ratings. Feature extraction and raw eye movement data were compared as input for multiple classification modeling methods. Accuracy rates of 92% and 80%, respectively, were achieved with individual-level and group-level modeling, suggesting that eye tracking is a promising technique for tracing trust levels.
Article
Full-text available
Effective human-multiagent teams will incorporate the cognitive skills of the human with the autonomous capabilities of the multiagent group to maximize task performance. However, producing a seamless fusion requires a greater understanding of the human’s cognitive state as it reacts to uncertainties in both the task environment and agent dynamics. This study examines external behaviors in concert with neurophysiological measures acquired via electroencephalography (EEG) to probe the interactions between cognitive processes, behaviors, and performance in a human-multiagent team task. We show that changes in the α (8-12Hz) and θ (4-8Hz) bands of EEG indicate a higher burden on the cognitive resources associated with visual-spatial reasoning required to estimate a more complex kinematic state of robotic agents. These results are reinforced by complementary behavioral shifts in gaze and pilot inputs. Additionally, higher performing participants tend to engage more actively in the task by utilizing greater amounts of visual-spatial reasoning. Finally, we show that features based on EEG dynamic-network-metrics provide discriminative information that distinguish gaze behaviors associated with the attention process.
Preprint
Full-text available
The work reported here addresses the capacity of psychophysiological sensors and measures using Electroencephalogram (EEG) and Galvanic Skin Response (GSR) to detect levels of trust for humans using AI-supported Human-Machine Interaction (HMI). Improvements to the analysis of EEG and GSR data may create models that perform as well, or better than, traditional tools. A challenge to analyzing the EEG and GSR data is the large amount of training data required due to a large number of variables in the measurements. Researchers have routinely used standard machine-learning classifiers like artificial neural networks (ANN), support vector machines (SVM), and K-nearest neighbors (KNN). Traditionally, these have provided few insights into which features of the EEG and GSR data facilitate the more and least accurate predictions - thus making it harder to improve the HMI and human-machine trust relationship. A key ingredient to applying trust-sensor research results to practical situations and monitoring trust in work environments is the understanding of which key features are contributing to trust and then reducing the amount of data needed for practical applications. We used the Local Interpretable Model-agnostic Explanations (LIME) model as a process to reduce the volume of data required to monitor and enhance trust in HMI systems - a technology that could be valuable for governmental and public sector applications. Explainable AI can make HMI systems transparent and promote trust. From customer service in government agencies and community-level non-profit public service organizations to national military and cybersecurity institutions, many public sector organizations are increasingly concerned to have effective and ethical HMI with services that are trustworthy, unbiased, and free of unintended negative consequences.
Chapter
Today, advances in robotics and autonomous systems enable construction workers to collaboratively work with robots, assigning physically demanding tasks to them. However, working alongside an industrial robot is a novel experience that may take a heavy toll on workers’ bodies and minds, particularly in dynamic and uncertain environments, such as construction sites. Given that trust is identified as a critical element for successful cooperation between humans and robots, effective measurement of workers’ trust in collaborative robots can lead to practical evaluation of human-robot partnership. In this context, most studies of trust have relied on self-reports. Nevertheless, questionnaires are unable to determine the trust promptly, impeding early preventive interventions. Furthermore, they are invasive, interfering with workers’ daily operations. To address these issues, this study proposes a procedure to non-invasively and continuously recognize workers’ trust in collaborative construction robots using electroencephalogram (EEG) signals. To that end, an experiment was conducted in which human workers performed a collaborative construction task (i.e., handling materials) with a virtual robot in an immersive environment. Meanwhile, workers’ EEG signals were recorded using a wearable sensor. Subsequently, the level of trust of the workers in collaborative robots was measured using the Trust Perception Scale-HRI. By analyzing acquired signals and applying different machine learning algorithms, it was found that EEG signals can be implemented to differentiate levels of trust of construction workers in their robotic counterparts. These findings suggest the feasibility of using workers’ EEG signals as a reliable, real-time indicator of trust in collaborative construction robots, which can be regarded as a practical approach for evaluating human-robot collaboration.
Article
Trust is essential for establishing and maintaining cooperative behaviors between individuals and institutions in a wide variety of social, economic, and political contexts. This book explores trust through the lens of neurobiology, focusing on empirical, methodological, and theoretical aspects. Written by a distinguished group of researchers from economics, psychology, human factors, neuroscience, and psychiatry, the chapters shed light on the neurobiological underpinnings of trust as applied in a variety of domains. Researchers and students will discover a refined understanding of trust by delving into the essential topics in this area of study outlined by leading experts.
Chapter
Trust is essential for establishing and maintaining cooperative behaviors between individuals and institutions in a wide variety of social, economic, and political contexts. This book explores trust through the lens of neurobiology, focusing on empirical, methodological, and theoretical aspects. Written by a distinguished group of researchers from economics, psychology, human factors, neuroscience, and psychiatry, the chapters shed light on the neurobiological underpinnings of trust as applied in a variety of domains. Researchers and students will discover a refined understanding of trust by delving into the essential topics in this area of study outlined by leading experts.
Book
Full-text available
This book provides an overview of recent research developments in the automation and control of robotic systems that collaborate with humans. A measure of human collaboration being necessary for the optimal operation of any robotic system, the contributors exploit a broad selection of such systems to demonstrate the importance of the subject, particularly where the environment is prone to uncertainty or complexity. They show how such human strengths as high-level decision-making, flexibility, and dexterity can be combined with robotic precision, and ability to perform task repetitively or in a dangerous environment. The book focuses on quantitative methods and control design for guaranteed robot performance and balanced human experience. Its contributions develop and expand upon material presented at various international conferences. They are organized into three parts covering: • one-human–one-robot collaboration; • one-human–multiple-robot collaboration; and • human–swarm collaboration. Individual topic areas include resource optimization (human and robotic), safety in collaboration, abstraction of swarm systems to make them suitable for human control, modeling and control of internal force interactions for collaborative manipulation, and the sharing of control between human and automated systems, etc. Control and decision algorithms feature prominently in the text, importantly within the context of human factors and the constraints they impose. Applications such as assistive technology, driverless vehicles, cooperative mobile robots, and swarm robots are considered. Illustrative figures and tables are provided throughout the book. Researchers and students working in controls, and the interaction of humans and robots will learn new methods for human–robot collaboration from this book and will find the cutting edge of the subject described in depth.
Conference Paper
Full-text available
In an increasingly automated world, trust between humans and autonomous systems is critical for successful integration of these systems into our daily lives. In particular, for autonomous systems to work cooperatively with humans, they must be able to sense and respond to the trust of the human. This inherently requires a control-oriented model of dynamic human trust behavior. In this paper, we describe a gray-box modeling approach for a linear third-order model that captures the dynamic variations of human trust in an obstacle detection sensor. The model is parameterized based on data collected from 581 human subjects, and the goodness of fit is approximately 80% for a general population. We also discuss the effect of demographics, such as national culture and gender, on trust behavior by re-parameterizing our model for subpopulations of data. These demographic-based models can be used to help autonomous systems further predict variations in human trust dynamics.
Article
Full-text available
Human trust in automation plays an important role in successful interactions between humans and machines. To design intelligent machines that can respond to changes in human trust, real-time sensing of trust level is needed. In this paper, we describe an empirical trust sensor model that maps psychophysiological measurements to human trust level. The use of psychophysiological measurements is motivated by their ability to capture a human's response in real time. An exhaustive feature set is considered, and a rigorous statistical approach is used to determine a reduced set of ten features. Multiple classification methods are considered for mapping the reduced feature set to the categorical trust level. The results show that psychophysiological measurements can be used to sense trust in real-time. Moreover, a mean accuracy of 71.57% is achieved using a combination of classifiers to model trust level in each human subject. Future work will consider the effect of human demographics on feature selection and modeling.
Conference Paper
Full-text available
Exchanging text messages via software on smart phones and computers has recently become one of the most popular ways for people to communicate and accomplish their tasks. However, there are negative aspects to using this kind of software, for example, it has been found that people communicating in the text-chat environment may experience a lack of trust and may face different levels of cognitive load [1, 11]. This study examines a novel way to measure interpersonal trust and cognitive load when they overlap with each other in the text-chat environment. We used Galvanic Skin Response (GSR), a physiological measurement, to collect data from twenty-eight subjects at four gradients and overlapping conditions between trust and cognitive load. The findings show that the GSR signals were significantly affected by both trust and cognitive load and provide promising evidence that GSR can be used as a tool for measuring interpersonal trust when cognitive load is low and also for measuring cognitive load when trust is high.
Article
Full-text available
In this paper we review classification algorithms used to design brain–computer interface (BCI) systems based on electroencephalography (EEG). We briefly present the commonly employed algorithms and describe their critical properties. Based on the literature, we compare them in terms of performance and provide guidelines to choose the suitable classification algorithm(s) for a specific BCI.
Article
Full-text available
Human-agent collectives (HAC) offers a new science for exploring the computational and human aspects of society. They are a new class of socio-technical systems in which humans and smart software (agents) engage in flexible relationships in order to achieve both their individual and collective goals. Sometimes the humans take the lead, sometimes the computer does and this relationship can vary dynamically. HACs are fundamentally socio-technical systems. Relationships between users and autonomous software systems will be driven as much by user-focused issues as technical ones. Humans and agents will form short-lived teams in HACs and coordinate their activities to achieve the various individual and joint goals present in the system before disbanding. This will be a continual process as new goals, opportunities and actors arrive. The novel approaches to HAC formation and operation must also address the needs of the humans within the system. Users will have to negotiate with software agents regarding the structure of the coalitions they will collectively form, and then coordinate their activities within the resulting coalition. The ways in which HACs operate requires us to reconsider some of the prevailing assumptions of provenance work. HACs need to understand and respond to the behavior of people and how this human activity is captured, processed, and managed raises significant ethical and privacy concerns.
Article
Full-text available
This article presents a framework of adaptive, measurable decision making for Multiple Attribute Decision Making (MADM) by varying decision factors in their types, numbers, and values. Under this framework, decision making is measured using physiological sensors such as Galvanic Skin Response (GSR) and eye-tracking while users are subjected to varying decision quality and difficulty levels. Following this quantifiable decision making, users are allowed to refine several decision factors in order to make decisions of high quality and with low difficulty levels. A case study of driving route selection is used to set up an experiment to test our hypotheses. In this study, GSR features exhibit the best performance in indexing decision quality. These results can be used to guide the design of intelligent user interfaces for decision-related applications in HCI that can adapt to user behavior and decision-making performance.
Article
Full-text available
Objective: We systematically review recent empirical research on factors that influence trust in automation to present a three-layered trust model that synthesizes existing knowledge. Background: Much of the existing research on factors that guide human-automation interaction is centered around trust, a variable that often determines the willingness of human operators to rely on automation. Studies have utilized a variety of different automated systems in diverse experimental paradigms to identify factors that impact operators’ trust. Method: We performed a systematic review of empirical research on trust in automation from January 2002 to June 2013. Papers were deemed eligible only if they reported the results of a human-subjects experiment in which humans interacted with an automated system in order to achieve a goal. Additionally, a relationship between trust (or a trust-related behavior) and another variable had to be measured. All together, 101 total papers, containing 127 eligible studies, were included in the review. Results: Our analysis revealed three layers of variability in human–automation trust (dispositional trust, situational trust, and learned trust), which we organize into a model. We propose design recommendations for creating trustworthy automation and identify environmental conditions that can affect the strength of the relationship between trust and reliance. Future research directions are also discussed for each layer of trust. Conclusion: Our three-layered trust model provides a new lens for conceptualizing the variability of trust in automation. Its structure can be applied to help guide future research and develop training interventions and design procedures that encourage appropriate trust.
Article
Full-text available
Promise is one of the most powerful tools producing trust and facilitating cooperation, and sticking to the promise is deemed as a key social norm in social interactions. The present study explored the extent to which promise would influence investors' decision-making in the trust game where promise had no predictive value regarding trustees' reciprocation. In addition, we examined the neural underpinnings of the investors' outcome processing related to the trustees' promise keeping and promise breaking. Consistent with our hypothesis, behavioral results indicated that promise could effectively increase the investment frequency of investors. Electrophysiological results showed that, promise induced larger differentiated -FRN responses to the reward and non-reward discrepancy. Taken together, these results suggested that promise would promote cooperative behavior, while breach of promise would be regarded as a violation of the social norm, corroborating the vital role of non-enforceable commitment in social decision making.
Article
Full-text available
High cognitive load arises from complex time and safety-critical tasks, for example, mapping out flight paths, monitoring traffic, or even managing nuclear reactors, causing stress, errors, and lowered performance. Over the last five years, our research has focused on using the multimodal interaction paradigm to detect fluctuations in cognitive load in user behavior during system interaction. Cognitive load variations have been found to impact interactive behavior: by monitoring variations in specific modal input features executed in tasks of varying complexity, we gain an understanding of the communicative changes that occur when cognitive load is high. So far, we have identified specific changes in: speech, namely acoustic, prosodic, and linguistic changes; interactive gesture; and digital pen input, both interactive and freeform. As ground-truth measurements, galvanic skin response, subjective, and performance ratings have been used to verify task complexity. The data suggest that it is feasible to use features extracted from behavioral changes in multiple modal inputs as indices of cognitive load. The speech-based indicators of load, based on data collected from user studies in a variety of domains, have shown considerable promise. Scenarios include single-user and team-based tasks; think-aloud and interactive speech; and single-word, reading, and conversational speech, among others. Pen-based cognitive load indices have also been tested with some success, specifically with pen-gesture, handwriting, and freeform pen input, including diagraming. After examining some of the properties of these measurements, we present a multimodal fusion model, which is illustrated with quantitative examples from a case study. The feasibility of employing user input and behavior patterns as indices of cognitive load is supported by experimental evidence. Moreover, symptomatic cues of cognitive load derived from user behavior such as acoustic speech signals, transcribed text, digital pen trajectories of handwriting, and shapes pen, can be supported by well-established theoretical frameworks, including O'Donnell and Eggemeier's workload measurement [1986] Sweller's Cognitive Load Theory [Chandler and Sweller 1991], and Baddeley's model of modal working memory [1992] as well as McKinstry et al.'s [2008] and Rosenbaum's [2005] action dynamics work. The benefit of using this approach to determine the user's cognitive load in real time is that the data can be collected implicitly that is, during day-to-day use of intelligent interactive systems, thus overcomes problems of intrusiveness and increases applicability in real-world environments, while adapting information selection and presentation in a dynamic computer interface with reference to load.
Article
Full-text available
As automated controllers supplant human intervention in controlling complex systems, the operators' role often changes from that of an active controller to that of a supervisory controller. Acting as supervisors, operators can choose between automatic and manual control. Improperly allocating function between automatic and manual control can have negative consequences for the performance of a system. Previous research suggests that the decision to perform the job manually or automatically depends, in part, upon the trust the operators invest in the automatic controllers. This paper reports an experiment to characterize the changes in operators' trust during an interaction with a semi-automatic pasteurization plant, and investigates the relationship between changes in operators' control strategies and trust. A regression model identifies the causes of changes in trust, and a ‘trust transfer function’ is developed using lime series analysis to describe the dynamics of trust. Based on a detailed analysis of operators' strategies in response to system faults we suggest a model for the choice between manual and automatic control, based on trust in automatic controllers and self-confidence in the ability to control the system manually.
Article
Full-text available
Automation does not mean humans are replaced; quite the opposite. Increasingly, humans are asked to interact with automation in complex and typically large-scale systems, including aircraft and air traffic control, nuclear power, manufacturing plants, military systems, homes, and hospitals. This is not an easy or error-free task for either the system designer or the human operator/automation supervisor, especially as computer technology becomes ever more sophisticated. This review outlines recent research and challenges in the area, including taxonomies and qualitative models of human-automation interaction; descriptions of automation-related accidents and studies of adaptive automation; and social, political, and ethical issues.
Article
Full-text available
Trust is among the most important factors in human life, as it pervades almost all domains of society. Although behavioral research has revealed a number of insights into the nature of trust, as well as its antecedents and consequences, an increasing number of scholars have begun to investigate the topic from a biological perspective to gain a deeper understanding. These biological investigations into trust have been carried out on three levels of analysis: genes, endocrinology, and the brain. Based on these three levels, we present a review of the literature on the biology of trust. Moreover, we integrate our findings into a conceptual framework which unifies the three levels of analysis, and we also link the biological levels to trust behavior. The results show that trust behavior is at least moderately genetically predetermined. Moreover, trust behavior is associated with specific hormones, in particular oxytocin, as well as specific brain structures, which are located in the basal ganglia, limbic system, and the frontal cortex. Based on these results, we discuss both methodological and thematic implications. (PsycINFO Database Record (c) 2012 APA, all rights reserved)
Article
Full-text available
Current inductive machine learning algorithms typically use greedy search with limited lookahead. This prevents them to detect significant conditional dependencies between the attributes that describe training objects. Instead of myopic impurity functions and lookahead, we propose to use RELIEFF, an extension of RELIEF developed by Kira and Rendell [10, 11], for heuristic guidance of inductive learning algorithms. We have reimplemented Assistant, a system for top down induction of decision trees, using RELIEFF as an estimator of attributes at each selection step. The algorithm is tested on several artificial and several real world problems and the results are compared with some other well known machine learning algorithms. Excellent results on artificial data sets and two real world problems show the advantage of the presented approach to inductive learning.
Article
Full-text available
In the feature subset selection problem, a learning algorithm is faced with the problem of selecting a relevant subset of features upon which to focus its attention, while ignoring the rest. To achieve the best possible performance with a particular learning algorithm on a particular training set, a feature subset selection method should consider how the algorithm and the training set interact. We explore the relation between optimal feature subset selection and relevance. Our wrapper method searches for an optimal feature subset tailored to a particular algorithm and a domain. We study the strengths and weaknesses of the wrapper approach and show a series of improved designs. We compare the wrapper approach to induction without feature subset selection and to Relief, a filter approach to feature subset selection. Significant improvement in accuracy is achieved for some datasets for the two families of induction algorithms used: decision trees and Naive-Bayes.
Article
Full-text available
Electrodermal activity is characterized by the superposition of what appear to be single distinct skin conductance responses (SCRs). Classic trough-to-peak analysis of these responses is impeded by their apparent superposition. A deconvolution approach is proposed, which separates SC data into continuous signals of tonic and phasic activity. The resulting phasic activity shows a zero baseline, and overlapping SCRs are represented by predominantly distinct, compact impulses showing an average duration of less than 2 s. A time integration of the continuous measure of phasic activity is proposed as a straightforward indicator of event-related sympathetic activity. The quality and benefit of the proposed measure is demonstrated in an experiment with short interstimulus intervals as well as by means of a simulation study. The advances compared to previous decomposition methods are discussed.
Article
Full-text available
In this paper we review classification algorithms used to design brain–computer interface (BCI) systems based on electroencephalography (EEG). We briefly present the commonly employed algorithms and describe their critical properties. Based on the literature, we compare them in terms of performance and provide guidelines to choose the suitable classification algorithm(s) for a specific BCI.
Article
Full-text available
To address the neurocognitive mechanisms that underlie choices made after receiving information from an anonymous individual, reaction times (Experiment 1) and event-related brain potentials (Experiment 2) were recorded as participants played three variants of the coin toss game. In this game, participants guess the outcomes of unseen coin tosses after a person in another room (dubbed 'the reporter') observes the coin toss outcomes and then sends reports (which may or may not be truthful) to participants about whether the coins landed on heads or tails. Participants knew that the reporter's interests were aligned with their own (common interests), opposed to their own (conflicting interests) or opposed to their own, but that the reporter was penalized every time he or she sent a false report about the coin toss outcome (penalty for lying). In the common interests and penalty for lying conditions, participants followed the reporter's reports over 90% of the time, in contrast to <59% of the time in the conflicting interests condition. Reaction time results indicated that participants took similar amounts of time to respond in the common interests and penalty for lying conditions and that they were reliably faster than in the conflicting interests condition. Event-related potentials timelocked to the reporter's reports revealed a larger P2, P3 and late positive complex response in the common interests condition than in the other two, suggesting that participants' brains processed the reporter's reports differently in the common interests condition relative to the other two conditions. Results suggest that even when people behave as if they trust information, they consider communicative efforts of individuals whose interests are aligned with their own to be slightly more informative than those of individuals who are made trustworthy by an institution, such as a penalty for lying.
Article
Full-text available
Previous research has indicated that the frequency of skin conductance responses without external stimulation or motor activity is a reliable indicator of psychophysiological states and traits. Some authors have suggested that cognitions elicit nonspecific skin conductance responses. These cognitions may resemble the stimuli that evoke a specific skin conductance response. In a within subjects design (n = 31 graduate students) the onset of nonspecific skin conductance responses triggered a signal for the subject to rate cognitions on several indices. These ratings ("absent" to "fully present") were compared with samples in the absence of phasic electrodermal activity. The subjects' current concerns, negative emotion, subjective arousal, and inner speech were rated to be significantly more intense at the time of nonspecific skin conductance responses compared to electrodermal nonresponding periods. Cognitive processes seem to be concomitants of nonspecific skin conductance responses.
Article
Full-text available
Two experiments were designed to examine the effects of attentional demands on the electroencephalogram during cognitive and emotional tasks. We found an interaction of task with hemisphere as well as more overall parietal alpha for tasks not requiring attention to the environment, such as mental arithmetic, than for those requiring such attention. Differential hemispheric activation for beta was found most strongly in the temporal areas for emotionally positive or negative tasks and in the parietal areas for cognitive tasks.
Article
Full-text available
We present an overview of our research into brain-computer interfacing (BCI). This comprises an offline study of the effect of motor imagery on EEG and an online study that uses pattern classifiers incorporating parameter uncertainty and temporal information to discriminate between different cognitive tasks in real-time.
Article
Full-text available
The ability to continuously and unobtrusively monitor levels of task engagement and mental workload in an operational environment could be useful in identifying more accurate and efficient methods for humans to interact with technology. This information could also be used to optimize the design of safer, more efficient work environments that increase motivation and productivity. The present study explored the feasibility of monitoring electroencephalo-graphic (EEG) indices of engagement and workload acquired unobtrusively and quantified during performance of cognitive tests. EEG was acquired from 80 healthy participants with a wireless sensor headset (F3-F4,C3-C4,Cz-POz,F3-Cz,Fz-C3,Fz-POz) during tasks including: multi-level forward/backward-digit-span, grid-recall, trails, mental-addition, 20-min 3-Choice Vigilance, and image-learning and memory tests. EEG metrics for engagement and workload were calculated for each 1 -s of EEG. Across participants, engagement but not workload decreased over the 20-min vigilance test. Engagement and workload were significantly increased during the encoding period of verbal and image-learning and memory tests when compared with the recognition/ recall period. Workload but not engagement increased linearly as level of difficulty increased in forward and backward-digit-span, grid-recall, and mental-addition tests. EEG measures correlated with both subjective and objective performance metrics. These data in combination with previous studies suggest that EEG engagement reflects information-gathering, visual processing, and allocation of attention. EEG workload increases with increasing working memory load and during problem solving, integration of information, analytical reasoning, and may be more reflective of executive functions. Inspection of EEG on a second-by-second timescale revealed associations between workload and engagement levels when aligned with specific task events providing preliminary evidence that second-by-second classifications reflect parameters of task performance.
Article
Full-text available
The overall aim of this research is to develop an EEG-based computer interface for use by people with severe physical disabilities. The work comprises an `offline' study and an `online' study, the offline study establishing principles of interface design and the online study putting those principles into practice. The work focuses on using EEG signals to drive one-dimensional cursor movements on a computer screen and our approach is characterised by our emphasis on pattern recognition methods rather than on biofeedback training. Two key technical features further define our approach: firstly, we use dynamic rather than static pattern recognition algorithms and, secondly, we infer not just the parameters of our classifier but also the uncertainty on those parameters. Both of these features result in more robust cursor control. 1 Introduction The ultimate aim of this research is to develop an EEG-based computer interface for use by people with severe physical disabilities. This would, ...
Conference Paper
Full-text available
. The aim of this paper is to analyse and formalise the dynamics of trust in the light of experiences. A formal framework is introduced for the analysis and specification of models for trust evolution and trust update. Different properties of these models are formally defined. 1 Introduction Trust is the attitude an agent has with respect to the dependability/capabilities of some other agent (maybe itself) or with respect to the turn of events. The agent might for example trust that the statements made by another agent are true. The agent might trust the commitment of another agent with respect to a certain (joint) goal. The agent might trust that another agent is capable of performing certain tasks. The agent might trust itself to be able to perform some tasks. The agent might trust that the current state of affairs will lead to a state of affairs that is agreeable to its own intentions, goals, commitments, or desires. In [1], [2] the importance of the notion trust is shown for agent...
Article
Sequential search methods characterized by a dynamically changing number of features included or eliminated at each step, henceforth “floating” methods, are presented. They are shown to give very good results and to be computationally more effective than the branch and bound method.
Book
During the past decade there has been an explosion in computation and information technology. With it have come vast amounts of data in a variety of fields such as medicine, biology, finance, and marketing. The challenge of understanding these data has led to the development of new tools in the field of statistics, and spawned new areas such as data mining, machine learning, and bioinformatics. Many of these tools have common underpinnings but are often expressed with different terminology. This book describes the important ideas in these areas in a common conceptual framework. While the approach is statistical, the emphasis is on concepts rather than mathematics. Many examples are given, with a liberal use of color graphics. It should be a valuable resource for statisticians and anyone interested in data mining in science or industry. The book's coverage is broad, from supervised learning (prediction) to unsupervised learning. The many topics include neural networks, support vector machines, classification trees and boosting---the first comprehensive treatment of this topic in any book. This major new edition features many topics not covered in the original, including graphical models, random forests, ensemble methods, least angle regression & path algorithms for the lasso, non-negative matrix factorization, and spectral clustering. There is also a chapter on methods for "wide" data (p bigger than n), including multiple testing and false discovery rates. Trevor Hastie, Robert Tibshirani, and Jerome Friedman are professors of statistics at Stanford University. They are prominent researchers in this area: Hastie and Tibshirani developed generalized additive models and wrote a popular book of that title. Hastie co-developed much of the statistical modeling software and environment in R/S-PLUS and invented principal curves and surfaces. Tibshirani proposed the lasso and is co-author of the very successful An Introduction to the Bootstrap. Friedman is the co-inventor of many data-mining tools including CART, MARS, projection pursuit and gradient boosting.
Book
Provides easy learning and understanding of DWT from a signal processing point of view. Presents DWT from a digital signal processing point of view, in contrast to the usual mathematical approach, making it highly accessible Offers a comprehensive coverage of related topics, including convolution and correlation, Fourier transform, FIR filter, orthogonal and biorthogonal filters Organized systematically, starting from the fundamentals of signal processing to the more advanced topics of DWT and Discrete Wavelet Packet Transform. Written in a clear and concise manner with abundant examples, figures and detailed explanations Features a companion website that has several MATLAB programs for the implementation of the DWT with commonly used filters.
Article
This paper describes a discrete wavelet transform-based feature extraction scheme for the classification of EEG signals. In this scheme, the discrete wavelet transform is applied on EEG signals and the relative wavelet energy is calculated in terms of detailed coefficients and the approximation coefficients of the last decomposition level. The extracted relative wavelet energy features are passed to classifiers for the classification purpose. The EEG dataset employed for the validation of the proposed method consisted of two classes: (1) the EEG signals recorded during the complex cognitive task-Raven's advance progressive metric test and (2) the EEG signals recorded in rest condition-eyes open. The performance of four different classifiers was evaluated with four performance measures, i.e., accuracy, sensitivity, specificity and precision values. The accuracy was achieved above 98 % by the support vector machine, multi-layer perceptron and the K-nearest neighbor classifiers with approximation (A4) and detailed coefficients (D4), which represent the frequency range of 0.53-3.06 and 3.06-6.12 Hz, respectively. The findings of this study demonstrated that the proposed feature extraction approach has the potential to classify the EEG signals recorded during a complex cognitive task by achieving a high accuracy rate.
Book
The study of event-related potentials (ERPs) -- signal-averaged EEG recordings that are time-locked to perceptual, cognitive, and motor events -- has increased dramatically in recent years, but until now there has been no comprehensive guide to ERP methodology comparable to those available for fMRI techniques. Event-Related Potentials meets the need for a practical and concise handbook of ERP methods that is suitable for both the novice user of an ERP system and a researcher more experienced in cognitive electrophysiology.The chapters in the first section discuss the design of ERP experiments, providing a practical foundation for understanding the design of ERP experiments and interpreting ERP data. Topics covered include quantification of ERP data and theoretical and practical aspects of ANOVAs as applied to ERP datasets. The second section presents a variety of approaches to ERP data analysis and includes chapters on digital filtering, artifact removal, source localization, and wavelet analysis. The chapters in the final section of the book cover the use of ERPs in relation to such specific participant populations as children and neuropsychological patients and the ways in which ERPs can be combined with related methodologies, including intracranial ERPs and hemodynamic imaging.