No full-text available
To read the full-text of this research,
you can request a copy directly from the authors.
Measuring human-computer trust
Abstract
In this study a psychometric instrument specifically designed to measure human-computer trust (HCT) was developed and tested. A rigorous method similar to that described by Moore and Benbasat (1991) was adopted. It was found that both cognitive and affective components of trust could be measured and that, in this study, the affective components were the strongest indicators of trust. The reliability of the instrument, measured as Cronbach's alpha, was 0.94. This instrument is the first of its kind to be specifically designed to measure HCT and shown empirically to be valid and reliable.
... Identifying such factors as well as the way they influence user behaviour and attitude towards a model has been an active research area for decades within the human factors and the AI communities, resulting in several behavioural theories describing the dynamics of the human-AI interaction [Lee and Moray, 1992, Linegang et al., 2006, Madsen and Gregor, 2000. A consistent point of convergence among these theories is that both model-related factors, such as the extent to which a model is perceived to be reliable and understandable, and user-related factors, such as their selfconfidence in their abilities to carry out a task, play a crucial role in the formation of the human-AI relationship. ...
... While this approach has the merit of providing a common ground upon which it is possible to compare the two, it reduces explanations to reliability indicators, even though their primarily function is to enhance understanding [Hoffman et al., 2018]. In addition, while prior research suggests that information regarding reliability and understanding have complementary functions [Zuboff, 1988, Sheridan, 1989, Lee and Moray, 1992, Madsen and Gregor, 2000, Kelly, 2003, the aforementioned approach fails to capture this aspect and provide relevant insights. For example, uncertainty estimates may help users decide the extent to which to rely on a model, but they provide no justifications in cases where a model makes incorrect predictions, hindering model acceptance [Ashoori and Weisz, 2019]. ...
... Moreover, another point that warrants further consideration is the way trust is operationalized in recent surveys. In particular, trust is almost exclusively assessed through the lens of agreement and switching percentages [Zhang et al., 2020], as opposed to using specialized trust measuring scales, such as those developed in [Madsen and Gregor, 2000, Jian et al., 2000, Adams et al., 2003, Cahour and Forzy, 2009. Nevertheless, it is well established that both of these percentages measure reliance, not trust, and that they may fail to account for confounding variables, such as time constraints, inherent application risks, or users' own self-confidence [Miller et al., 2016, Chancey et al., 2013. ...
AI and ML models have already found many applications in critical domains, such as healthcare and criminal justice. However, fully automating such high-stakes applications can raise ethical or fairness concerns. Instead, in such cases, humans should be assisted by automated systems so that the two parties reach a joint decision, stemming out of their interaction. In this work we conduct an empirical study to identify how uncertainty estimates and model explanations affect users' reliance, understanding, and trust towards a model, looking for potential benefits of bringing the two together. Moreover, we seek to assess how users' behaviour is affected by their own self-confidence in their abilities to perform a certain task, while we also discuss how the latter may distort the outcome of an analysis based on agreement and switching percentages.
... Trust is a state effectuated by the trustor in which the trustee has power over some subset of the trustor's goals that the trustor believes they could not accomplish with a better net outcome on their own. The state may be an attitude (Merritt & Ilgen, 2008; J. D. Lee & See, 2004), belief (Gefen et al., 2003), expectation (Muir & Moray, 1996), judgement (Merritt & Ilgen, 2008; J. D. Lee & See, 2004), willingness (Madsen & Gregor, 2000;Goillau, Kelly, Boardman, & Jeannot, 2003;McKnight & Chervany, 2001b;Körber, 2018), confidence (Madsen & Gregor, 2000;Goillau et al., 2003), or reliance . The perceived power of the trustee over the trustor's goals is often expressed in terms of the trustor's vulnerability, uncertainty, or risk (Schaefer, 2013). ...
... Trust is a state effectuated by the trustor in which the trustee has power over some subset of the trustor's goals that the trustor believes they could not accomplish with a better net outcome on their own. The state may be an attitude (Merritt & Ilgen, 2008; J. D. Lee & See, 2004), belief (Gefen et al., 2003), expectation (Muir & Moray, 1996), judgement (Merritt & Ilgen, 2008; J. D. Lee & See, 2004), willingness (Madsen & Gregor, 2000;Goillau, Kelly, Boardman, & Jeannot, 2003;McKnight & Chervany, 2001b;Körber, 2018), confidence (Madsen & Gregor, 2000;Goillau et al., 2003), or reliance . The perceived power of the trustee over the trustor's goals is often expressed in terms of the trustor's vulnerability, uncertainty, or risk (Schaefer, 2013). ...
... Beyond one-and two-item assessments, alternatives exist for classifying more complex trust survey instruments. Some try to create a single scale to just capture (dis)trust as opposed to those that posit multiple layers (vertical) (Merritt et al., 2019;Merritt, 2011a;McKnight et al., 2011;Goillau et al., 2003) or constructs (horizontal) (Muir & Moray, 1996;Madsen & Gregor, 2000; J. D. Lee & Moray, 1994;Lee & Liang, 2015). Among multi-dimensional survey creators, some are focused on performance-based trust (e.g., Madsen & Gregor, 2000;Wojton, Porter, T. Lane, Bieber, & Madhavan, 2020), and others are relation-based (sometimes termed affective) (e.g., Wechsung, Weiss, Kühnel, Ehrenbrink, & Möller, 2013;Rupp, Michaelis, McConnell, & Smither, 2016), with a recent trend towards more mixed approaches (e.g., McKnight et al., 2011;Gefen et al., 2003;Wang & Emurian, 2005;Park, 2020), as described in Law and Scheutz (2021). ...
A significant challenge to measuring human-automation trust is the amount of construct proliferation, models, and questionnaires with highly variable validation. However, all agree that trust is a crucial element of technological acceptance, continued usage, fluency, and teamwork. Herein, we synthesize a consensus model for trust in human-automation interaction by performing a meta-analysis of validated and reliable trust survey instruments. To accomplish this objective, this work identifies the most frequently cited and best-validated human-automation and human-robot trust questionnaires, as well as the most well-established factors, which form the dimensions and antecedents of such trust. To reduce both confusion and construct proliferation, we provide a detailed mapping of terminology between questionnaires. Furthermore, we perform a meta-analysis of the regression models that emerged from those experiments which used multi-factorial survey instruments. Based on this meta-analysis, we demonstrate a convergent experimentally validated model of human-automation trust. This convergent model establishes an integrated framework for future research. It identifies the current boundaries of trust measurement and where further investigation is necessary. We close by discussing choosing and designing an appropriate trust survey instrument. By comparing, mapping, and analyzing well-constructed trust survey instruments, a consensus structure of trust in human-automation interaction is identified. Doing so discloses a more complete basis for measuring trust emerges that is widely applicable. It integrates the academic idea of trust with the colloquial, common-sense one. Given the increasingly recognized importance of trust, especially in human-automation interaction, this work leaves us better positioned to understand and measure it.
... Furthermore, trust in the recommender agent is crucial for acceptance of decision-support outcomes since higher trust in an advisor is positively associated with accepting a recommendation (Sniezek & Van Swol, 2001;Van Swol & Sniezek, 2005). Trust encompasses a cognition-based and an affective-based component, with the former being grounded in knowledge and experience with a person or technology and the latter being derived from feelings and emotional responses and therefore going beyond a rational evaluation of the trustee (Madsen & Gregor, 2000;McAllister, 1995). Thus, even if a recommendation seems fair, low trust in the decision-support agent can still influence the acceptance of the recommendation and vice versa. ...
... Trust was measured with the subscales "perceived technical competence" and "faith" from the Human-Computer Trust Scale by Madsen and Gregor (2000). The scale was originally designed to measure human-computer trust. ...
... Consequentialist Table 1. Adapted Items of the Human-Computer Trust Scale (Madsen & Gregor, 2000). ...
Although algorithm-based systems are increasingly used as a decision-support for managers, there is still a lack of research on the effects of algorithm use and more specifically on potential algorithmic bias on decision-makers. To investigate how potential social bias in a recommendation outcome influences trust, fairness perceptions, and moral judgement, we used a moral dilemma scenario. Participants (N = 215) imagined being human resource managers responsible for personnel selection and receiving decision-support from either human colleagues or an algorithm-based system. They received an applicant preselection that was either gender-balanced or predominantly male. Although participants perceived algorithm-based support as less biased, they also perceived it as generally less fair and had less trust in it. This could be related to the finding that participants perceived algorithm-based systems as more consistent but also as less likely to uphold moral standards. Moreover, participants tended to reject algorithm-based preselection more often than human-based and were more likely to use utilitarian judgements when accepting it, which may indicate different underlying moral judgement processes.
... In 1996, Muir and Moray [14] proposed a scale to measure trust in automation. Since then, many studies have emerged with a similar purpose of developing scales those include empirically Derived (ED) [15], Human-Computer Trust (HTC) [16], SHAPE Automation Trust Index (SATI) [9,17], Trust Perceived Scale-HRI [18,19], and HRI Trust Scale [20]. Those are examples of scales empirically developed to understand how trust is perceived in technologically enhanced scenarios. ...
... Trust has since long been addressed in various disciplines (e.g., social psychology, economy, philosophy, and industrial organization). Each domain that explores trust, affects either an attitude, an intention, or a behavior [14,16]. ...
... Although there are tools capable of measuring trust in HRI, most of them have problems related to the trust assessment, since they are used in the context of automation in general and not applied to cobots in particular. Some of the existing scales that measure trust in the HRI need more robust statistical analysis [15][16][17]. ...
Recently there has been an increasing demand for technologies (automated and intelligent machines) that brings benefits to organizations and society. Similar to the widespread use of personal computers in the past, today's needs are towards facilitating human-machine technology appropriation, especially in highly risky and regulated industries like robotics, manufacturing, automation, military, finance, or healthcare. In this context, trust can be used as a critical element to instruct how human-machine interaction should occur. Considering the context-dependency and multidimensional trust, this study seeks to find a way to measure the effects of perceived trust in a collaborative robot (cobot), regardless of its literal credibility as a real person. This article aims at translating, adapting, and validating a Human-Computer Trust Scale (HCTM) in human-robot interaction (HRI) context and its application to cobots. The Human-Robot Interaction Trust Scale (HRITS) involved 239 participants and included eleven items. The 2nd order CFA with a general factor called "trust" have proven to be empirically robust (CFI = :94; TLI = :93; SRMR = :04; and RMSEA = :05) [CR = :84; AVE = :58, and MaxRðHÞ = :92]; results indicated a good measurement of the general factor trust, and the model satisfied the criteria for measure trust. An analysis of the differences in perceptions of trust by gender was conducted using a t-test. This analysis showed that statistical differences by gender exist (p = :04). This study's results allowed for a better understanding of trust in HRI, specifically regarding cobots. The validation of a Portuguese scale for trust assessment in HRI can give a valuable contribution to designing collaborative environments between humans and robots.
... According to Madsen and Gregor (2000), there are two main types of human-computer trust: cognitive and affective. When humans perceive machines as reliable, technically competent, and easily understandable, they trust the machines in cognitive ways. ...
... According to a review by Malle and Ullman (2021) of previous research on human-robot trust, most studies have focused on robots' performances and how humans' evaluations of such performances affect their trust which consists of reliability, consistency, and competence, to name a few. A scale by Madsen and Gregor (2000) encompasses all those commonly used trust dimensions while being relatively parsimonious (i.e., 25-item scale vs. 40-item scale by Schaffer, 2013). Motivated by the theoretical categorisation of human-computer trust into cognitive and affective dimensions and existing research findings, the current study focuses on how the emotional experiences of IVA users may impact their trust in machines. ...
... Participants' trust in Siri was measured using a 25-item, 5-point Likert-type scale (1 = strongly disagree, 5 = strongly agree) originally developed by Madsen and Gregor (2000) and modified for this research context. The questions included the following five constructs: perceived reliability (five items, e.g., 'Siri always provides the advice I require to make my decision'), perceived technical competence (five items, e.g., 'Siri uses appropriate methods to reach decisions'), perceived understandability (five items, e.g., 'I know what will happen the next time I use Siri because I understand how it behaves'), faith (five items, e.g., 'I believe advice from Siri even when I don't know for certain that it is correct'), and personal attachment (five items, e.g., 'I would feel a sense of loss if Siri was unavailable and I could no longer use it'). ...
A theoretical model of trust in human-machine communication (HMC) was tested and emotional experience and social presence were evaluated during an interaction with an intelligent virtual agent (IVA), Siri. A two (‘American female’ or ‘American male’ Siri) by two (functional or social task) experiment was conducted with 229 subjects with random assignments. According to multivariate analyses of covariances, participants reported higher levels of emotional significance when they interacted with Siri to inquire about functional tasks. No gender or interaction effects between gender and task were detected. Confirmatory factor analyses and structural equation modelling indicated that both social presence and emotional experience were directly and positively associated with five dimensions of trust (i.e. perceived reliability, technical competence, perceived understandability, faith, and personal attachment) in Siri. This direct effect model was significant after controlling for the effects of task type. Additionally, a test of the mediation model indicated full mediation between emotional experiences and trust by social presence.
... Previous trust-related works regarded trust as a multidimensional concept encompassing both the user's confidence in the system and their willingness to act on the system's decisions and advice (Madsen & Gregor, 2000). There have been widely used scales to measure the human-machine trust (Jian, Bisantz, Drury, & Llinas, 1998) and instruments to measure the dimensions of cognitive and affective trust in the intelligent decision aids (Madsen & Gregor, 2000). ...
... Previous trust-related works regarded trust as a multidimensional concept encompassing both the user's confidence in the system and their willingness to act on the system's decisions and advice (Madsen & Gregor, 2000). There have been widely used scales to measure the human-machine trust (Jian, Bisantz, Drury, & Llinas, 1998) and instruments to measure the dimensions of cognitive and affective trust in the intelligent decision aids (Madsen & Gregor, 2000). We gathered human trust in the machine teammate during the inflation process using a 7-point Likert scale ranging from 'strongly disagree' to 'strongly agree'. ...
... At the end of the experiment, we collected the subjective evaluations from the two team collaboration modes regarding the dimensions of perceived performance (task and machine), perceived decision authority (Wynne & Lyons, 2018), perceived task interdependence (Wynne & Lyons, 2018), perceived risk, responsibility attribution of loss (Dietvorst et al., 2015), decision style matching (Madsen & Gregor, 2000), pleasure of the experience, and willingness to collaborate (Madsen & Gregor, 2000). Table 4 shows the ten questions of the subjective evaluation. ...
Human-machine teaming has shown great potential in sequential risky decision-making (SRDM), and it is promising that machines will no longer work as subordinates. Technical advances prompt people to consider the contexts of humans and machines sharing decision authority. This study aims to compare task performance, human behaviors, and subjective perception of machines in a human-dominated team (machine as a subordinate) versus a human-machine joint team (machine as a partner). We modified the Balloon Analogue Risk Task (BART) experiment to include a highly accurate machine and accommodate two types of human-machine teams (HMTs). The results showed that both HMTs yielded comparable task performance and overperformed human or machine deciding alone. In the human-machine joint team, the machine as a partner entailed human decision-makers to cede power and coordinate, and their pumping decisions became more conservative and fluctuating. Moreover, human decision-makers reacted more sensitively to the different results of the last trial. Although subjects generally favored working with a machine subordinate, they exhibited similar trust levels in both HMTs after sufficient interaction. Our preliminary findings show that allocating a partial decision authority to highly accurate machines changed human behaviors without impairing task performance or trust in collaboration.
... Although many studies have examined trust in the interpersonal and societal domains, in different technologies, studies addressing trust in medical AI-assisted diagnosis and treatment are scarce. Madsen and Gregor (2000) defined HCT as "the extent to which a user is confident in, and willing to act on the basis of the recommendations, actions, and decisions of an artificially intelligent decision aid" [42], which enhances healthcare workers' adoption intention of AI-assisted diagnosis and treatment [27,33]. Theories of interpersonal relationships have established trust as a social glue in relationships, groups, and societies [21,43]. ...
... Although many studies have examined trust in the interpersonal and societal domains, in different technologies, studies addressing trust in medical AI-assisted diagnosis and treatment are scarce. Madsen and Gregor (2000) defined HCT as "the extent to which a user is confident in, and willing to act on the basis of the recommendations, actions, and decisions of an artificially intelligent decision aid" [42], which enhances healthcare workers' adoption intention of AI-assisted diagnosis and treatment [27,33]. Theories of interpersonal relationships have established trust as a social glue in relationships, groups, and societies [21,43]. ...
... The interaction between people and technology has special trust characteristics [47]. HCT is the degree to which people have confidence in AI systems and are willing to take action [42]. Trust is considered an attitude intention [47], which could directly influence acceptance and help people make cognitive judgments by decreasing risk perception [48] and enhancing benefit perception [49]. ...
Artificial intelligence (AI)-assisted diagnosis and treatment could expand the medical scenarios and augment work efficiency and accuracy. However, factors influencing healthcare workers’ adoption intention of AI-assisted diagnosis and treatment are not well-understood. This study conducted a cross-sectional study of 343 dental healthcare workers from tertiary hospitals and secondary hospitals in Anhui Province. The obtained data were analyzed using structural equation modeling. The results showed that performance expectancy and effort expectancy were both positively related to healthcare workers’ adoption intention of AI-assisted diagnosis and treatment. Social influence and human–computer trust, respectively, mediated the relationship between expectancy (performance expectancy and effort expectancy) and healthcare workers’ adoption intention of AI-assisted diagnosis and treatment. Furthermore, social influence and human–computer trust played a chain mediation role between expectancy and healthcare workers’ adoption intention of AI-assisted diagnosis and treatment. Our study provided novel insights into the path mechanism of healthcare workers’ adoption intention of AI-assisted diagnosis and treatment.
... One is the complex nature of trust. According to the human-computer trust (HCT) model (Madsen and Gregor, 2000), trust is formed in two dimensions: cognition-based trust and affect-based trust. Cognition-based trust is based on humans' intellectual perceptions of AI reasoning, whereas affect-based trust is based on humans' emotional responses to AI systems. ...
... The study workflow is described in Fig. 3. After completing a Human-AI task, participants were asked to complete cognitive-based trust scale (Madsen and Gregor, 2000). Each of the participants completed ten Human-AI tasks, which resulted in 410 completed tasks. ...
... For the scope of this study, we measure the contribution of an XAI class on trust calibration in two ways: subjective and objective. First, subjective measurements were used by following the cognitive-based trust scale proposed in (Madsen and Gregor, 2000) which quantifies the extent to which the XAI interface helps users to understand, rely on and perceive the technical competency of the AI. Specifically, self-reporting cognitive-based trust measures were used in this study to observe whether an XAI class helped in increasing or decreasing trust during the sessions (RQ1.1). ...
Machine learning has made rapid advances in safety-critical applications, such as traffic control, finance, and healthcare. With the criticality of decisions they support and the potential consequences of following their recommendations, it also became critical to provide users with explanations to interpret machine learning models in general, and black-box models in particular. However, despite the agreement on explainability as a necessity, there is little evidence on how recent advances in eXplainable Artificial Intelligence literature (XAI) can be applied in collaborative decision-making tasks, i.e., Human decision-maker and an AI system working together, to contribute to the process of trust calibration effectively. This research conducts an empirical study to evaluate four XAI classes for their impact on trust calibration. We take clinical decision support systems as a case study and adopt a within-subject design followed by semi-structured interviews. We gave participants clinical scenarios and XAI interfaces as a basis for decision-making and rating tasks. Our study involved 41 medical practitioners who use clinical decision support systems frequently. We found that users perceive the contribution of explanations to trust calibration differently according to the XAI class and to whether XAI interface design fits their job constraints and scope. We revealed additional requirements on how explanations shall be instantiated and designed to help a better trust calibration. Finally, we build on our findings and present guidelines for designing XAI interfaces.
... The take-over performance measures included accuracy (whether the participant safely avoided crashing in the TOR or system-malfunction conditions) and response time (RT; First response of at least 1 degree turn of steering wheel from event onset or braking action). The subjective dependent measures included a trust measure adapted from the Human-Computer Trust questionnaire by Madsen and Gregor (2000), the Automated Driving Opinion Survey (ADOS; Kyriakidis et al., 2015), and the Usefulness of Automation Survey (UAS; Gold et al. (2015). The subjective measures are further described in the Surveys section. ...
... Questionnaire (Madsen & Gregor, 2000) and told to rate their initial trust in the automated driving system they had just experienced. This measure of trust was to compare to the trust measure from the ADOS, serve as a baseline of trust in the system, and to examine the equality of the different experimental groups. ...
... Questionnaire (Madsen & Gregor, 2000). The scale included 24 questions that assess humanautomation trust based on five separate criteria to encompass the multi-dimensional aspects of trust. ...
The performance of a driving automation system (DAS) can influence the human drivers' trust in the system. This driving-simulator study examined how different types of DAS failures affected drivers' trust. The automation-failure type (no-failure, takeover-request, system-malfunction) was manipulated among 122 participants, when a critical hazard event occurred. The dependent measures included participants’ trust ratings after each of seven drives and their takeover performance following the hazard. Results showed that trust improved before any automation failure occurred, demonstrating proper trust calibration toward the errorless system. In the takeover-request and system-malfunction conditions, trust decreased similarly in response to the automation failures, although the takeover-request condition had better takeover performance. For the drives after the automation failure, trust was gradually repaired but did not recover to the original level. This study demonstrated how trust develops and responds to DAS failures, informing future research for trust repair interventions in designing DASs.
... Although these scales have advantages, each falls short in their creation process based on established psychometric principles for scale creation (Fabrigar et al., 1999) or from a lack of theoretical backing. The most notable scales that claim to assess system trustworthiness are the Trust in Automated Systems Survey (TASS; Jian et al., 2000), the Human-Computer Trust scale (HCT; Madsen & Gregor, 2000), the Multidimensional Measure of Trust (MDMT; Malle & Ullman, 2021), the Trust Perception Scale-Human Robot Interaction (HRI; Schaefer, 2016), and the Trust of Automated Systems Test (TOAST; Wojton et al., 2020). Despite their prevalence, the authors are not aware of any scales other than the TOAST that have assessed trustworthiness from the theoretical perspective suggested by Lee and See (2004). ...
... Similarly, researchers gave little to no information regarding their initial item pool and came up with too few items for a factor analysis (e.g., TOAST; Wojton et al., 2020). The authors of the TASS (Jian et al., 2000), HCT (Madsen & Gregor, 2000), and MDMT (Malle & Ullman, 2021) performed EFAs on data from small sample sizes for scale development, possibly leading to underfactoring. Lastly, only one of the researchers reported utilizing scree plots (Schaefer, 2016) or parallel analyses to estimate the number of factors. ...
Objective:
We created and validated a scale to measure perceptions of system trustworthiness.
Background:
Several scales exist in the literature that attempt to assess trustworthiness of system referents. However, existing measures suffer from limitations in their development and validation. The current study sought to develop a scale based on theory and methodological rigor.
Method:
We conducted exploratory and confirmatory factor analyses on data from two online studies to develop the System Trustworthiness Scale (STS). Additional analyses explored the manipulation of the factors and assessed convergent and divergent validity.
Results:
The exploratory factor analyses resulted in a three-factor solution that represented the theoretical constructs of trustworthiness: performance, purpose, and process. Confirmatory factor analyses confirmed the three-factor solution. In addition, correlation and regression analyses demonstrated the scale's divergent and predictive validity.
Conclusion:
The STS is a psychometrically valid and predictive scale for assessing trustworthiness perceptions of system referents.
Applications:
The STS assesses trustworthiness perceptions of systems. Importantly, the scale differentiates performance, purpose, and process constructs and is adaptable to a variety of system referents.
... We focused our investigation to contexts with well-defined end-user goals in which AI provides the end-user with decision-making support to reach those goals [31]. Trust is a complex social construct [18], so we narrowed our scope to an established view of human-AI trust [55], which defines trust as the extent to which the end-user is confident in and willing to act on the basis of AI recommendations [34,36]. Thus, we measure how much end-users rely on AI advice to make their decisions, instead of depending on subjective self-reported perceptions of AI trustworthiness [8]. ...
... Existing work used both qualitative methods to explain end-users' decision to trust AI and quantitative methods to measure the magnitude of such trust. Although end-users' self-reported measures of trust could be collected using surveys [34], end-user attitudes towards trust measured using such survey instruments could differ from the actual willingness of the end-user to act based on AI recommendations [8]. As an alternative, existing research often used trust-related behavioral measures [55]-objective measures of end-user behavior when interacting with AI. ...
Trustworthy Artificial Intelligence (AI) is characterized, among other things, by: 1) competence, 2) transparency, and 3) fairness. However, end-users may fail to recognize incompetent AI, allowing untrustworthy AI to exaggerate its competence under the guise of transparency to gain unfair advantage over other trustworthy AI. Here, we conducted an experiment with 120 participants to test if untrustworthy AI can deceive end-users to gain their trust. Participants interacted with two AI-based chess engines, trustworthy (competent, fair) and untrustworthy (incompetent, unfair), that coached participants by suggesting chess moves in three games against another engine opponent. We varied coaches' transparency about their competence (with the untrustworthy one always exaggerating its competence). We quantified and objectively measured participants' trust based on how often participants relied on coaches' move recommendations. Participants showed inability to assess AI competence by misplacing their trust with the untrustworthy AI, confirming its ability to deceive. Our work calls for design of interactions to help end-users assess AI trustworthiness.
... This is created by drawing key information from the complex algorithm and is often presented in an interface thru the set of rules, a summary of features used, relative examples, or supplementary information (Das & Rad, 2020;Jin et al., 2021). The running hypothesis is that these explanations calibrate trust cognitively (Madsen & Gregor, 2000), allowing users to think and eventually create conclusions. This results in a research direction focused on improving the cognitive resource, by developing new techniques, for mental model building. ...
... However, trust has long been known to work beyond cognition. Notably, many social sciences and human-computer interface (HCI) scholars have identified that trust from cognitive cues can also be developed via irrational factors like emotions (Lee & See, 2004;Madsen & Gregor, 2000;Riegelsberger et al., 2003). Previous studies with similar transparency utility such as for social robots (Gompei & Umemuro, 2018), warning alerts (Buck et al., 2018), intelligent personal assistance (Chen & Park, 2021), security seals (Bernardo & Tangsoc, 2021) had verified this, which profoundly changed how they are managed and used (e.g., focusing on design) to maximize its effectiveness in developing trust or reliance. ...
The rise of Explainable Artificial Intelligence (XAI) has been a game changer for the growth of Artificial Intelligence (AI) powered systems. By providing human-level explanations, it systematically solves the most significant issue that AI faces: the black-box paradox realized from the complex hidden layers of deep and machine learning that powers it. Fundamentally, it allows users to learn how the AI operates and comes to decisions, thus enabling cognitive calibration of trust and subsequent reliance on the system. This conclusion has been supported by various research under different contexts and has motivated the development of newer XAI techniques. However, as human-computer interaction and social science studies suggest, these findings might be limited as the emotional component, which is also established from the interaction, was not considered. Emotions have long been determined to play a dominant role in decision-making as they can rapidly and unconsciously be infused in judgments. This insinuates an idea that XAI might facilitate trust calibration not solely because of the cognitive information it provides but of the emotions developed on the explanations. Considering this idea has not been explored, this study aims to examine the effects of emotions associated with the interaction with XAI towards trust, reliance, and explanation satisfaction. One hundred twenty-three participants were invited to partake in an online experiment anchored in an image classification testbed. The premise was that they were hired to classify different species of animals and plants, with an XAI-equipped image classification AI available to give them recommendations. At the end of each trial, they were tasked to rate their emotions upon interaction with the XAI, trust in the system, and satisfaction with the explanation. Reliance was measured based on whether they accepted the recommendations of AI. Results show that users who felt surprisingly happy and trusting emotions reported high trust, reliance, and satisfaction. On the other hand, users that developed fearfully dismayed and anxiously suspicious emotions have a significant negative relationship with satisfaction. Essentially, as supported by the post-interview, the study surfaced three critical findings on the affective functionality of XAI. First, emotions developed are mainly attributed to the design and overall composition rather than the information it carries. Second, trust and reliance can only be developed from positive emotions. Users might not trust and rely on an AI system even if it has a meaningful explanation if it develops negative emotions to the user. Third, explanation satisfaction can be triggered by both positive and negative emotions. The former is mainly from the presentation of XAI, while the latter is because of understanding the limitation of the AI.
... Initially, we used three items to form the construct of ATT -one was adopted from Davis et al. (1992) and two from Higgins and his co-authors (Compeau et al., 1999;Thompson et al., 1991). As we derived ST from the perceived local explainability of an intelligent system decision's result visualization as well as the perceived global explainability of the intelligent system's decision process, we initially included five items from Madsen and Gregor (2000) to address the global component and two items from Cramer et al. (2008) Lastly, the measurement items for facilitating conditions and social influence were adapted from Taylor and Todd (1995), Thompson et al. (1991), Moore and Benbasat (1991), and Davis (1989). ...
... Our results also show that while being influenced by transparency, trust is not solely explained by it. In accordance with Madsen and Gregor (2000), the pre-existing propensity to trust that is reflected by TP requires extra treatment that goes beyond simply providing explanations. Thus, trust issues need to be addressed head-on by implementing guidelines for trustworthy AI (Thiebes et al., 2021). ...
Contemporary decision support systems are increasingly relying on artificial intelligence technology such as machine learning algorithms to form intelligent systems. These systems have human-like decision capacity for selected applications based on a decision rationale which cannot be looked-up conveniently and constitutes a black box. As a consequence, acceptance by end-users remains somewhat hesitant. While lacking transparency has been said to hinder trust and enforce aversion towards these systems, studies that connect user trust to transparency and subsequently acceptance are scarce. In response, our research is concerned with the development of a theoretical model that explains end-user acceptance of intelligent systems. We utilize the unified theory of acceptance and use in information technology as well as explanation theory and related theories on initial trust and user trust in information systems. The proposed model is tested in an industrial maintenance workplace scenario using maintenance experts as participants to represent the user group. Results show that acceptance is performance-driven at first sight. However, transparency plays an important indirect role in regulating trust and the perception of performance.
... Asking the participant to self-report their own level of trust is extremely common within this field of research (Hancock et al., 2011). Many existing questionnaires to measure the perceived trustworthiness of another agent exist (e.g., Singh et al., 1993;Madsen and Gregor, 2000;Adams et al., 2003;Cahour and Forzy, 2009;Merritt, 2011). Several of these questionnaires are discussed and reviewed by Hoffman et al. (2018), where a final questionnaire is concluded, adapting many items from (Merritt, 2011). ...
Introduction: Collaboration in teams composed of both humans and automation has an interdependent nature, which demands calibrated trust among all the team members. For building suitable autonomous teammates, we need to study how trust and trustworthiness function in such teams. In particular, automation occasionally fails to do its job, which leads to a decrease in a human’s trust. Research has found interesting effects of such a reduction of trust on the human’s trustworthiness, i.e., human characteristics that make them more or less reliable. This paper investigates how automation failure in a human-automation collaborative scenario affects the human’s trust in the automation, as well as a human’s trustworthiness towards the automation.
Methods: We present a 2 × 2 mixed design experiment in which the participants perform a simulated task in a 2D grid-world, collaborating with an automation in a “moving-out” scenario. During the experiment, we measure the participants’ trustworthiness, trust, and liking regarding the automation, both subjectively and objectively.
Results: Our results show that automation failure negatively affects the human’s trustworthiness, as well as their trust in and liking of the automation.
Discussion: Learning the effects of automation failure in trust and trustworthiness can contribute to a better understanding of the nature and dynamics of trust in these teams and improving human-automation teamwork.
... Details of the trust estimation module are described in Section 3. In human-computer studies, the user's trust in a system is usually measured using subjective measurements based on self-reported questionnaires. For example, Madsen and Gregor [24] developed a questionnaire based on a hierarchical model of trust where subjects can agree or disagree with statements about the system's trustworthiness. Here, the authors differentiate between affect-based and cognition-based trust. ...
... Here, cognition-based trust refers to making a choice on whom to trust based on "good reasons" constituting evidence of trustworthiness according to prior knowledge and information [35]. Similarly, human-computer trust also includes both cognitive and affective components, although affective components are stronger indicators [36]. Accordingly, people's cognition of a trustee plays an important role in their trust in a person, organization, or machine. ...
Before Automated Driving Systems (ADS) with full driving automation (SAE Level 5) are placed into practical use, the issue of calibrating drivers' initial trust in Level 5 ADS to an appropriate degree to avoid inappropriate disuse or improper use should be resolved. This study aimed to identify the factors that affected drivers' initial trust in Level 5 ADS. We conducted two online surveys. Of these, one explored the effects of automobile brands and drivers' trust in automobile brands on drivers' initial trust in Level 5 ADS using a Structural Equation Model (SEM). The other identified drivers' cognitive structures regarding automobile brands using the Free Word Association Test (FWAT) and summarized the characteristics that resulted in higher initial trust among drivers in Level 5 ADS. The results showed that drivers' trust in automobile brands positively impacted their initial trust in Level 5 ADS, which showed invariance across gender and age. In addition, the degree of drivers' initial trust in Level 5 ADS was significantly different across different automobile brands. Furthermore, for automobile brands with higher trust in automobile brands and Level 5 ADS, drivers' cognitive structures were richer and varied, which included particular characteristics. These findings suggest the necessity of considering the influence of automobile brands on calibrating drivers' initial trust in driving automation.
... Prior information and beliefs have a role in forming the initial state of trust; however, trust and confidence may evolve over time as the system is explored and experienced. Common variables used to assess and study trust include user knowledge, familiarity, technical competence, confidence, emotions, beliefs, faith, and personal attachments [385,386]. These variables may be quantified by explicitly questioning users about their experiences with a system during and after usage. ...
Artificial intelligence (AI) is currently being utilized in a wide range of sophisticated applications, but the outcomes of many AI models are challenging to comprehend and trust due to their black-box nature. Usually, it is essential to understand the reasoning behind an AI model’s decision-making. Thus, the need for eXplainable AI (XAI) methods for improving trust in AI models has arisen. XAI has become a popular research subject within the AI field in recent years. Existing survey papers have tackled the concepts of XAI, its general terms, and post-hoc explainability methods but there have not been any reviews that have looked at the assessment methods, available tools, XAI datasets, and other related aspects. Therefore, in this comprehensive study, we provide readers with an overview of the current research and trends in this rapidly emerging area with a case study example. The study starts by explaining the background of XAI, common definitions, and summarizing recently proposed techniques in XAI for supervised machine learning. The review divides XAI techniques into four axes using a hierarchical categorization system: (i) data explainability, (ii) model explainability, (iii) post-hoc explainability, and (iv) assessment of explanations. We also introduce available evaluation metrics as well as open-source packages and datasets with future research directions. Then, the significance of explainability in terms of legal demands, user viewpoints, and application orientation is outlined, termed as XAI concerns. This paper advocates for tailoring explanation content to specific user types. An examination of XAI techniques and evaluation was conducted by looking at 410 critical articles, published between January 2016 and October 2022, in reputed journals and using a wide range of research databases as a source of information. The article is aimed at XAI researchers who are interested in making their AI models more trustworthy, as well as towards researchers from other disciplines who are looking for effective XAI methods to complete tasks with confidence while communicating meaning from data.
... Two dimensions of trust (cognitive and behavioral trust) were measured in the study [40]. Cognitive trust was measured by ve items adapted from [58]: "The website performs reliably, " "I can rely on the website to function properly, " "The website uses appropriate methods to reach decisions, " "The website has sound knowledge about how to recommend tness plans, " and "The website correctly uses the information I enter" (U = 0.92, " = 4.90, (⇡ = 1.32). Behavioral trust was accessed by three items adapted from [71]: "I will use this website again, " "I will use this website frequently, " and "I will tell my friends about this website" (U = 0.93, " = 3.59, (⇡ = 1.90). ...
Recommender systems (RS) have become increasingly vital for guiding health actions. While traditional systems filter content based on either demographics, personal history of activities, or preferences of other users, newer systems use social media information to personalize recommendations, based either on the users’ own activities and/or those of their friends on social media platforms. However, we do not know if these approaches differ in their persuasiveness. To find out, we conducted a user study of a fitness plan recommender system (N = 341), wherein participants were randomly assigned to one of six personalization approaches, with half of them given a choice to switch to a different approach. Data revealed that social media-based personalization threatens users’ identity and increases privacy concerns. Users prefer personalized health recommendations based on their own preferences. Choice enhances trust by providing users with a greater sense of agency and lowering their privacy concerns. These findings provide design implications for RS, especially in the preventive health domain.
... We collected the following variables to test explanatory channels: the perceived benefit to the individual user relative to the perceived benefit to the general public from the smart assistant, perceived individual preference for a sustainable environment and health system, profit-orientation, and trustworthiness and technical skills of each operating organization as perceived by the participants. Because psychological and attitudinal characteristics can explain cooperative behavior that includes temporal conflicts and technology usage, we consider the following control variables: humanassistant trust (Madsen and Gregor, 2000), interpersonal trust (Eurobarometer, 2014), future time orientation (specifically time perspective and anticipation of future consequences) (Gjesme, 1979;Steinberg et al., 2009), self-reported health (Idler and Angel, 1990), self-reported environmentally friendly behavior (Idler and Angel, 1990), and risk attitude (Weber et al., 2002). Given their particular relevance in explaining cooperative behavior under uncertainty, we collected social value orientation (Murphy et al., 2011) and the anticipated behavior of others using monetary incentivized tasks, to encourage honest and realistic responses (see Online Appendix A, part A7). ...
When using digital devices and services, individuals provide their personal data to organizations in exchange for gains in various domains of life. Organizations use these data to run technologies such as smart assistants, augmented reality, and robotics. Most often, these organizations seek to make a profit. Individuals can, however, also provide personal data to public databases that enable nonprofit organizations to promote social welfare if sufficient data are contributed. Regulators have therefore called for efficient ways to help the public collectively benefit from its own data. By implementing an online experiment among 1696 US citizens, we find that individuals would donate their data even when at risk of getting leaked. The willingness to provide personal data depends on the perceived risk level of a data leak but not on a realistic impact of the data on social welfare. Individuals are less willing to donate their data to the private industry than to academia or the government. Finally, individuals are not sensitive to whether the data are processed by a human-supervised or a self-learning smart assistant.
... Although Kabra et al. (2017) didn't get any significant association between user trust and behavioral intention. Madsen and Gregor (2000) described trust as the degree to which a customer is assured and eager to act based on suggestions, events, and judgments. Trust of the buyer reduces the perceived risk, decreasing the buyer's intention for transactions in business to the consumer market (Pavlou and Gefen, 2004). ...
The digital age has changed the way businesses are run today. Technology is not just a priviledge but also a necessity. The recent pandemic has given important lessons to business to be proactive and advanced in technology. Customers occupy the centrestage in any business and giving them solutions promptly for their queries can leave a positive impression and lead to long term customer enagagment. For this, a trained team of employees are required who can give their services incessantly. However, the rising employee retention costs have impacted the profit margins of organisations and more human intervention becomes a hurdle in standardization of processes. Therefore, organisations are roping in artificial intelligence to be more efficient and cost effective. Chatbots are artificial intelligence softwares that have enabled organisations to give answers to customer queries online. The study intends to examine the significant factors in determining customers’ intentions to use chatbots. This paper aims to understand the role of user experience, performance expectancy, effort expectancy, and trust in customer chatbot use intentions from the Indian point of view. A structured questionnaire was utilized to gather data for testing the proposed model, which was conceptualized based on extant literature on technology acceptance and consumer behavior. A survey response of 354 respondents was taken. In order to test the constructs, the collected data was analyzed through AMOS 21. The research findings depicted the positive impact of user experience, trust performance expectancy, and effort expectancy on customer intention to use chatbots, which influences actual usage. This paper empirically demonstrates the relationship among various variables affecting customer intentions to use chatbots. Since the paper uses data collected from a sample not randomly selected, it may regulate the generalization of the results. This study intends to add to the current research gap in the existing literature about customer intention to use chatbots, mainly in the Indian context. The research examined how positive user experience, performance, effort expectancy, and trust affect customer intentions to take support from chatbots.
... The EU self-assessment mechanisms for Trustworthy Artificial Intelligence created by the AIHLEG expert group is another example [78] of broadening the view. The same regards the Human-Computer trust (HTC) psychometric scales proposed by Madsen and Gregor [79], SHAPE Automation trust Index (SATI) [80]. We need new HCI mechanisms to measure potentially faulty TAI design practices, which can lead to risky, unsafe life or threatening social ramifications [16,47]. ...
The Internet revolution in 1990, followed by the data-driven and information revolution, has transformed the world as we know it. Nowadays, what seam to be 10 to 20 years ago, a science fiction idea (i.e., machines dominating the world) is seen as possible. This revolution also brought a need for new regulatory practices where user trust and artificial Intelligence (AI) discourse has a central role. This work aims to clarify some misconceptions about user trust in AI discourse and fight the tendency to design vulnerable interactions that lead to further breaches of trust, both real and perceived. Findings illustrate the lack of clarity in understanding user trust and its effects on computer science, especially in measuring user trust characteristics. It argues for clarifying those notions to avoid possible trust gaps and misinterpretations in AI adoption and appropriation.
... Some attributes such as empathy [26] and trust can be considered both a robot attribute (e.g. how trustworthy [88] is a robot) or more as a relationship attribute relevant to the overall perception and measurement of human-robot trust [57,71,84,114]. Games are usually multiplayer experiences featuring groups and teams. As such, this review includes several instances that have included robots in teams and groups. ...
During the past two decades, robots have been increasingly deployed in games. Researchers use games to better understand human-robot interaction and, in turn, the inclusion of social robots during gameplay creates new opportunities for novel game experiences. The contributions from social robotics and games communities cover a large spectrum of research questions using a wide variety of scenarios. In this article, we present the first comprehensive survey of the deployment of robots in games. We organise our findings according to four dimensions: (1) the societal impact of robots in games, (2) games as a research platform, (3) social interactions in games, and (4) game scenarios and materials. We discuss some significant research achievements and potential research avenues for the gaming and social robotics communities. This article describes the state of the art of the research on robots in games in the hope that it will assist researchers to contextualise their work in the field, to adhere to best practices and to identify future areas of research and multidisciplinary collaboration.
... Following Lu et al. (2019), the ethical objects were defined by trust and rejection behavior [11]. The perceived trust of the agent was measured using the Human-Computer Trust Scale on a five-point Likert scale [29]. This questionnaire had 25 items. ...
With the increasing adaptability and complexity of advisory artificial intelligence (AI)-based agents, the topics of explainable AI and human-centered AI are moving close together. Variations in the explanation itself have been widely studied, with some contradictory results. These could be due to users’ individual differences, which have rarely been systematically studied regarding their inhibiting or enabling effect on the fulfillment of explanation objectives (such as trust, understanding, or workload). This paper aims to shed light on the significance of human dimensions (gender, age, trust disposition, need for cognition, affinity for technology, self-efficacy, attitudes, and mind attribution) as well as their interplay with different explanation modes (no, simple, or complex explanation). Participants played the game Deal or No Deal while interacting with an AI-based agent. The agent gave advice to the participants on whether they should accept or reject the deals offered to them. As expected, giving an explanation had a positive influence on the explanation objectives. However, the users’ individual characteristics particularly reinforced the fulfillment of the objectives. The strongest predictor of objective fulfillment was the degree of attribution of human characteristics. The more human characteristics were attributed, the more trust was placed in the agent, advice was more likely to be accepted and understood, and important needs were satisfied during the interaction. Thus, the current work contributes to a better understanding of the design of explanations of an AI-based agent system that takes into account individual characteristics and meets the demand for both explainable and human-centered agent systems.
... Other questionnaires aim to assess users' trust after one or multiple interactions have taken place. Numerous scales exist, often with different systems in mind such as the Human-Robot trust scale [19], the Human-Computer Trust scale [20], and Trust in Industrial Human-Robot Collaboration scale [21]. The next sections will go more in-depth into the two scales used in this study, along with the rationale for why each scale was chosen. ...
Trust between humans and robots is a complex, multifaceted phenomenon and measuring it subjectively and reliably is challenging. It is also context dependent and so choosing the right tool for a specific study can prove difficult. This paper aims to evaluate various trust measures and compare them in terms of sensitivity to changes in trust. This is done by comparing two validated trust questionnaires (TAS and MDMT) and one single item assessment in a COVID-19 triage scenario. We found that trust measures are equivalent in terms of sensitivity to changes in trust. Furthermore, the study showed that trust could be measured similarly through a single item assessment in comparison with other lengthier scales, in scenarios with distinct breaks in trust. This finding would be of use for experiments where lengthy questionnaires are not appropriate, such as those in the wild.
... Explanations and their relationship with understanding have been studied for several years in human-AI interaction research, focusing mainly on figuring out what are the important characteristics of an ideal explanation [27]. ...
The lack of transparency of powerful Machine Learning systems paired with their growth in popularity over the last decade led to the emergence of the eXplainable Artificial Intelligence (XAI) field. Instead of focusing solely on obtaining highly performing models, researchers also develop explanation techniques that help better understand the system’s reasoning for a particular output. An explainable system can be designed, developed, and evaluated from different perspectives, which enables researchers from different disciplines to work together on this topic. However, the multidisciplinary nature of XAI systems creates new challenges for condensing and structuring adequate methodologies to design and evaluate such systems. This paper presents a survey of Human-centred and Computer-centred methods to evaluate XAI systems. We propose a new taxonomy to categorize XAI evaluation methods more clearly and intuitively. This categorization gathers knowledge from different disciplines and organizes the evaluation methods according to a set of categories that represent key properties of XAI systems. Possible ways to use the proposed taxonomy in the design and evaluation of XAI systems are also discussed, alongside with some concluding remarks and future directions of research.
... PT is another latent construct that that can only be measured indirectly, as is common in research assessing people's perceptions of system trustworthiness (e.g., by asking people to report on their perceived trustworthiness of a system [1,58,73,90,116], or by observing people's interactions with systems [135,149]). The facets that contribute to overall PT are similar to those that contribute to overall AT (see Figure 3). ...
Designing trustworthy systems and enabling external parties to accurately assess the trustworthiness of these systems are crucial objectives. Only if trustors assess system trustworthiness accurately, they can base their trust on adequate expectations about the system and reasonably rely on or reject its outputs. However, the process by which trustors assess a system’s actual trustworthiness to arrive at their perceived trustworthiness remains underexplored. In this paper, we conceptually distinguish between trust propensity, trustworthiness, trust, and trusting behavior. Drawing on psychological models of assessing other people’s characteristics, we present the two-level Trustworthiness Assessment Model (TrAM). At the micro level, we propose that trustors assess system trustworthiness based on cues associated with the system. The accuracy of this assessment depends on cue relevance and availability on the system’s side, and on cue detection and utilization on the human’s side. At the macro level, we propose that individual micro-level trustworthiness assessments propagate across different trustors – one stakeholder’s trustworthiness assessment of a system affects other stakeholder’s trustworthiness assessments of the same system. The TrAM advances existing models of trust and sheds light on factors influencing the (accuracy of) trustworthiness assessments. It offers implications for system design, stakeholder training, and regulation related to trustworthiness assessments.
In Explainable Artificial Intelligence (XAI) research, various local model-agnostic methods have been proposed to explain individual predictions to users in order to increase the transparency of the underlying Artificial Intelligence (AI) systems. However, the user perspective has received less attention in XAI research, leading to a (1) lack of involvement of users in the design process of local model-agnostic explanations representations and (2) a limited understanding of how users visually attend them. Against this backdrop, we refined representations of local explanations from four well-established model-agnostic XAI methods in an iterative design process with users. Moreover, we evaluated the refined explanation representations in a laboratory experiment using eye-tracking technology as well as self-reports and interviews. Our results show that users do not necessarily prefer simple explanations and that their individual characteristics, such as gender and previous experience with AI systems, strongly influence their preferences. In addition, users find that some explanations are only useful in certain scenarios making the selection of an appropriate explanation highly dependent on context. With our work, we contribute to ongoing research to improve transparency in AI.
Human-machine collaboration has shown great potential in sequential risky decision-making (SRDM). Human decision-makers made their decisions depending on the condition and their machine teammates. This paper aimed to explore attentional behaviors and decision-making processes under two human-machine collaboration contexts. To this end, 25 participants were asked to complete a modified Balloon Analog Risk Task with a highly accurate machine under different human-machine teams (HMTs). We collected and analyzed task performance, decision-making choices, eye-tracking data and subjective data. We employed the decision tree algorithm to describe decision-making processes and tested the performance through resubstitution validation and train-test validation. We found that both HMTs achieved comparable performance. Participants in the human-dominated team paid more attention to the machine-recommended value while participants in the human-machine joint team paid more attention to the inflation information of the machine. There were significant associations between choice ratios of inflation alternatives and decision choices for most subjects in both HMTs. In the human-machine joint team, we found a correlation between task profits and the fixation count on machine recommended value (r = 0.40, p = 0.05), and a correlation between the number of total explosions and the fixation count on whether the machine recommending to pump or not (r = –0.36, p = 0.07). Decision tree algorithm could cover and describe at least 67% of the decision-making choices and performed differently when subjects took different strategies. Our results revealed that eye-tracking and decision tree can be potential tools to describe and understand human SRDM behaviors.KeywordsHuman-machine collaborationBalloon Analog Risk Task (BART)Human-machine team (HMT)Eye-trackingDecision tree
Simulation has been widely used as a training tool in multiple industries. Due to the immersive and guided real-life experiences, learners can gain technical and non-technical skills in simulation-based training. However, despite the benefits of simulation-based training, research has shown that knowledge transfer may not be a guarantee. One possible explanation is how learning outcomes are assessed in current training programs using simulations. Therefore, in this report, we conducted a bibliometric analysis to understand the current research landscape related to simulation-based training programs and the learning assessments used in programs both in and outside the manufacturing industry. Based on our findings, more research is needed in simulation-based training in manufacturing and the learning assessment related to these training programs. Future research should focus on expanding the data collection to multiple databases and taking a more qualitative look into the current literature by conducting a systematized literature review.Keywordsbibliometric analysissimulation-based trainingmanufacturinglearning assessment
Fatigue is a loss in cognitive or physical performance due to physiological factors such as insufficient sleep, long work hours, stress, and physical exertion. It adversely affects the human body and can slow reaction times, reduce attention, and limit short-term memory. Hence, there is a need to monitor a person’s state to avoid extreme fatigue conditions that can result in physiological complications. However, tools to understand and assess fatigue are minimal. This paper primarily focuses on building an experimental setup that induces cognitive fatigue (CF) and physical fatigue (PF) through multiple cognitive and physical tasks while simultaneously recording physiological data. First, we built a prototype sensor suit embedded with numerous physiological sensors for easy use during data collection. Second, participants’ self-reported visual analog scores (VAS) are reported after each task to confirm fatigue induction. Finally, an evaluation system is built that utilizes machine learning (ML) models to detect states of CF and PF from sensor data, thus providing an objective measure. Our methods beat state-of-the-art approaches, where Random Forest performs the best in detecting PF with an accuracy of 80.5% while correctly predicting the true PF condition 88% of the time. On the other hand, the long short-term memory (LSTM) recurrent neural network produces the best results in detecting CF in the subjects (with 84.1% accuracy, 0.9 recall).Keywordsfatiguecognitive fatiguephysical fatiguemulti-modal sensorsmachine learning
Introduction
Whilst a theoretical basis for implementation research is seen as advantageous, there is little clarity over if and how the application of theories, models or frameworks (TMF) impact implementation outcomes. Clinical artificial intelligence (AI) continues to receive multi-stakeholder interest and investment, yet a significant implementation gap remains. This bibliometric study aims to measure and characterize TMF application in qualitative clinical AI research to identify opportunities to improve research practice and its impact on clinical AI implementation.
Methods
Qualitative research of stakeholder perspectives on clinical AI published between January 2014 and October 2022 was systematically identified. Eligible studies were characterized by their publication type, clinical and geographical context, type of clinical AI studied, data collection method, participants and application of any TMF. Each TMF applied by eligible studies, its justification and mode of application was characterized.
Results
Of 202 eligible studies, 70 (34.7%) applied a TMF. There was an 8-fold increase in the number of publications between 2014 and 2022 but no significant increase in the proportion applying TMFs. Of the 50 TMFs applied, 40 (80%) were only applied once, with the Technology Acceptance Model applied most frequently ( n = 9). Seven TMFs were novel contributions embedded within an eligible study. A minority of studies justified TMF application ( n = 51,58.6%) and it was uncommon to discuss an alternative TMF or the limitations of the one selected ( n = 11,12.6%). The most common way in which a TMF was applied in eligible studies was data analysis ( n = 44,50.6%). Implementation guidelines or tools were explicitly referenced by 2 reports (1.0%).
Conclusion
TMFs have not been commonly applied in qualitative research of clinical AI. When TMFs have been applied there has been (i) little consensus on TMF selection (ii) limited description of selection rationale and (iii) lack of clarity over how TMFs inform research. We consider this to represent an opportunity to improve implementation science's translation to clinical AI research and clinical AI into practice by promoting the rigor and frequency of TMF application. We recommend that the finite resources of the implementation science community are diverted toward increasing accessibility and engagement with theory informed practices. The considered application of theories, models and frameworks (TMF) are thought to contribute to the impact of implementation science on the translation of innovations into real-world care. The frequency and nature of TMF use are yet to be described within digital health innovations, including the prominent field of clinical AI. A well-known implementation gap, coined as the “AI chasm” continues to limit the impact of clinical AI on real-world care. From this bibliometric study of the frequency and quality of TMF use within qualitative clinical AI research, we found that TMFs are usually not applied, their selection is highly varied between studies and there is not often a convincing rationale for their selection. Promoting the rigor and frequency of TMF use appears to present an opportunity to improve the translation of clinical AI into practice.
We propose a method to specify and evaluate the trustworthiness of AI-based systems using scenarios and design tactics. Using our trustworthiness scenarios and design tactics, we can analyze the architectural design of AI-enabled systems to ensure trustworthiness has been properly achieved. Trustworthiness scenarios allow for the specification of trustworthiness, and design tactics are used to achieve the desired level of trustworthiness in the system. We illustrate the validity of our proposal through the design of the software architecture of a pollination robot. We find that our method opens discussions on ways to achieve trustworthiness and leads to the discovery of any weaknesses in the design concerning the trustworthiness of the AI system. Furthermore, our method allows for designing an AI system with trustworthiness in mind and therefore leads to greater analysis and identification of the sub-attributes that affect the trustworthiness of an AI system.KeywordsTrustworthinessTrustworthy AIUtility treeTrustSoftware architectureQuality attribute scenariosArchitectural tacticsArchitecture analysisATAM
When interacting with artificial intelligence (AI) in the medical domain, users frequently face automated information processing, which can remain opaque to them. For example, users with diabetes may interact daily with automated insulin delivery (AID). However, effective AID therapy requires traceability of automated decisions for diverse users. Grounded in research on human-automation interaction, we study Subjective Information Processing Awareness (SIPA) as a key construct to research users’ experience of explainable AI. The objective of the present research was to examine how users experience differing levels of traceability of an AI algorithm. We developed a basic AID simulation to create realistic scenarios for an experiment with N = 80, where we examined the effect of three levels of information disclosure on SIPA and performance. Attributes serving as the basis for insulin needs calculation were shown to users, who predicted the AID system’s calculation after over 60 observations. Results showed a difference in SIPA after repeated observations, associated with a general decline of SIPA ratings over time. Supporting scale validity, SIPA was strongly correlated with trust and satisfaction with explanations. The present research indicates that the effect of different levels of information disclosure may need several repetitions before it manifests. Additionally, high levels of information disclosure may lead to a miscalibration between SIPA and performance in predicting the system’s results. The results indicate that for a responsible design of XAI, system designers could utilize prediction tasks in order to calibrate experienced traceability.
Software evolution is a time-consuming and costly process due to its complex architecture. Software designers need to produce software that is effective as well as durable. Durability and effectiveness of software are the foremost priorities and challenges for developers. This book comprises real-life case studies of durability issues and their solutions that bring to light loopholes and show how to fix them, to enhance durability. Existing literature on software durability tells us that the first step is to recognise the problem. It gives information about durability, risk, estimation, knowledge, and governance based on five main characteristics: dependability, trustworthiness, usability, security, and human trust. The book serves as a complete package to get acquainted with assurance and risk management from a software durability perspective. It enhances our understanding of the concept of durability, its multi-dimensional approach, threats and their types, risk, mitigation techniques, and suggestive measures. The book reviews the emerging trends in the software development process in the context of durability concepts such as automated code reviews, coding standards, and software durability standards and their testing, cost management solutions, low-code or no-code solutions, and durability assurance.
In the future, Air Traffic Controllers (ATCOs) will be exposed to various new situations that will make their work much more complex and demanding. Automation of the system may enable support to ATCOs and facilitate decision-making and execution of different types of actions and tasks. Also, the introduction of automation can increase workload, information uncertainty, and potential negative outcome on performance at the task level. Trust in automation plays the main role in forming the connection between the ATCO and the automation. According to that, this paper aims to explain the concept of trust in general and in automation with a brief review of the methods and measurements of trust in the relevant literature.
Technological advances have increased the automation of Uncrewed Aerial Vehicles, allowing human operators to manage multiple vehicles at a high-level without the need to understand low-level system behaviors.Previous laboratory studies have explored the relationship between reliability, trust, use of automation, andthe effects of number of vehicles under supervision on subjective workload. Due to limitations resulting fromthe COVID-19 pandemic, in-person laboratory studies are not always possible. Therefore, this work aimed to investigate if remote data collection alternatives, such as Amazon’s Mechanical Turk, can provide comparativeresults as those obtained in laboratorysettings. A study was conducted in the context of small droneoperations. As expected, higher reliability led to higher trust ratings and the inclusion of more vehicles ledto higher workload. In contrast, reliability unexpectedly had no significant effect on intention to use theautomation. Though these results were encouraging, several limitations were identified
Traditional crowdsourcing has mostly been viewed as requester-worker interaction where requesters publish tasks to solicit input from human crowdworkers. While most of this research area is catered towards the interest of requesters, we view this workflow as a teacher-learner interaction scenario where one or more human-teachers solve Human Intelligence Tasks to train machine learners. In this work, we explore how teachable machine learners can impact their human-teachers, and whether they form a trustable relation that can be relied upon for task delegation in the context of crowdsourcing. Specifically, we focus our work on teachable agents that learn to classify news articles while also guiding the teaching process through conversational interventions. In a two-part study, where several crowd workers individually teach the agent, we investigate whether this learning by teaching approach benefits human-machine collaboration, and whether it leads to trustworthy AI agents that crowd workers would delegate tasks to. Results demonstrate the benefits of the learning by teaching approach, in terms of perceived usefulness for crowdworkers, and the dynamics of trust built through the teacher-learner interaction.
Proper calibration of human reliance on AI is fundamental to achieving complementary performance in AI-assisted human decision-making. Most previous works focused on assessing user reliance, and more broadly trust, retrospectively, through user perceptions and task-based measures. In this work, we explore the relationship between eye gaze and reliance under varying task difficulties and AI performance levels in a spatial reasoning task. Our results show a strong positive correlation between percent gaze duration on the AI suggestion and user AI task agreement, as well as user perceived reliance. Moreover, user agency is preserved particularly when the task is easy and when AI performance is low or inconsistent. Our results also reveal nuanced differences between reliance and trust. We discuss the potential of using eye gaze to gauge human reliance on AI in real-time, enabling adaptive AI assistance for optimal human-AI team performance.
Machine learning advances, particularly deep learning, have enabled us to design models that excel at increasingly complicated tasks. Because of the growing size and complexity of these models, it’s becoming more difficult to grasp how they arrive at their forecasts and when they go incorrect or even worse. Now, think of a situation in which we humans could open these black-box learning models and translate the content into a human-understandable format. This is known as Explainable Artificial Intelligence and there has been a lot of research in this field over the last few years mainly focusing on how to explain different types of models. The advancement of this research, raised a very important query: “Why does a model need to be explained?” So, the most accurate answer to this question is “TRUST”. TRUST that the models are making the correct decisions over the correct assumptions. TRUST that we can tell what happened when a model fails. TRUST that we can do on a model implemented on a large scale that the predictions are made in line with expectations. It’s hard to trust a system that’s not transparent about its internal processes. This paper discusses the evaluation measures and application areas of XAI. Some XAI-related concepts were also mentioned.KeywordsExplainable artificial intelligenceInterpretable machine learningMachine learningXAIDeep learning
With the rise of AI, algorithms have become better at learning underlying patterns from the training data including ingrained social biases based on gender, race, etc. Deployment of such algorithms to domains such as hiring, healthcare, law enforcement, etc. has raised serious concerns about fairness, accountability, trust and interpretability in machine learning algorithms. To alleviate this problem, we propose D-BIAS, a visual interactive tool that embodies human-in-the-loop AI approach for auditing and mitigating social biases from tabular datasets. It uses a graphical causal model to represent causal relationships among different features in the dataset and as a medium to inject domain knowledge. A user can detect the presence of bias against a group, say females, or a subgroup, say black females, by identifying unfair causal relationships in the causal network and using an array of fairness metrics. Thereafter, the user can mitigate bias by refining the causal model and acting on the unfair causal edges. For each interaction, say weakening/deleting a biased causal edge, the system uses a novel method to simulate a new (debiased) dataset based on the current causal model while ensuring a minimal change from the original dataset. Users can visually assess the impact of their interactions on different fairness metrics, utility metrics, data distortion, and the underlying data distribution. Once satisfied, they can download the debiased dataset and use it for any downstream application for fairer predictions. We evaluate D-BIAS by conducting experiments on 3 datasets and also a formal user study. We found that D-BIAS helps reduce bias significantly compared to the baseline debiasing approach across different fairness metrics while incurring little data distortion and a small loss in utility. Moreover, our human-in-the-loop based approach significantly outperforms an automated approach on trust, interpretability and accountability.
Smart voice assistants such as Amazon Alexa and Google Home are becoming increasingly pervasive in our everyday environments. Despite their benefits, their miniaturized and embedded cameras and microphones raise important privacy concerns related to surveillance and eavesdropping. Recent work on the privacy concerns of people in the vicinity of these devices has highlighted the need for ‘tangible privacy’, where
control and feedback mechanisms can provide a more assured sense of whether the camera or microphone is ‘on’ or ‘off’. However, current designs of these devices lack adequate mechanisms to provide such assurances. To address this gap in the design of smart voice assistants, especially in the case of disabling microphones, we evaluate several designs that incorporate (or not) tangible control and feedback mechanisms. By comparing people’s perceptions of risk, trust, reliability, usability, and control for these designs in a between-subjects online experiment (N=261), we find that devices with tangible built-in physical controls are perceived as more trustworthy and usable than those with non-tangible mechanisms. Our findings present an approach for tangible, assured privacy especially in the context of embedded microphones.
This study addressed the nature and functioning of relationships of interpersonal trust among managers and professionals in organizations, the factors influencing trust's development, and the implications of trust for behavior and performance. Theoretical foundations were drawn from the sociological literature on trust and the social-psychological literature on trust in close relationships. An initial test of the proposed theoretical framework was conducted in a field setting with 194 managers and professionals.
Discusses the limitations of the rational-structural and goal/expectation approaches to the problem of public goods (PGs), presents a new approach—the structural goal/expectation approach—intended to overcome these limitations, and tested 4 predictions derived from the new approach in a study of 48 4-person groups of undergraduates. According to this new approach, members who have realized the undesirable consequence of free riding and the importance of mutual cooperation will cooperate to establish a sanctioning system that assures other members' cooperation instead of trying to induce other members into mutual cooperation directly through cooperative actions. One important condition for their voluntary cooperation in the establishment of a sanctioning system is their realization that voluntarily based cooperation is impossible. In the study, each member of the group was given resource money that they could keep for themselves or contribute to the provision of a PG. The increase in the personal benefit due to one's contribution was reduced to zero, and Ss were not allowed to see each other in person. Some groups were given opportunities to develop a negative sanctioning system that punished the least cooperative group member. The level of punishment depended on the total amount of contribution made by the group members to the sanctioning system, which was separate from the contribution to the original PG. Results support the approach's predictions. (31 ref) (PsycINFO Database Record (c) 2012 APA, all rights reserved)
in this chapter we will examine the development and impact of trust in the context of close relationships
we will begin with a definition of trust and a discussion of its roots in individuals' interpersonal histories
we will go on to explore the development of trust in intimate relationships, emphasizing how its foundations are colored by the seminal experiences that mark different stages of interdependence
we will then consider the various states of trust that can evolve and their consequences for people's emotions and perceptions in established relationships (PsycINFO Database Record (c) 2012 APA, all rights reserved)
Two measures of nonverbal sensitivity to facial cues, sensitivity to unknown others, and sensitivity to an intimate other, along with a measure of general sending accuracy, were obtained from 48 married couples (20–31 yrs old). Rotter's Internal–External Locus of Control Scale and the Rotter Trust Scale were administered. It was predicted that (a) internal Ss would demonstrate better nonverbal encoding and decoding skills, (b) high-trust Ss would be better able to decode the nonverbal cues of other, and (c) the combination of internality and high trust would be associated with the highest level of decoding abilities and that the combination of external and low trust would correlate with the lowest level of decoding abilities. Results show no relation between either control or trust expectancies and sending accuracy. Trust expectancies covaried with nonverbal receiving abilities for both men and women, with high trust being associated with increased sensitivity. Control expectancies covaried with general nonverbal receiving abilities differently for both men and women, with internal women scoring higher and internal low-trust men scoring lowest on these measures. (25 ref) (PsycINFO Database Record (c) 2012 APA, all rights reserved)
Little attention has been given to the fit of specific types of computerized decision aids with various decision problem situations. Information regarding this potential contingency relationship is needed both for theoretical development and for guiding practical applications. This paper reports a laboratory study of the relationships between use of an ad hoc, personal, computerized decision aid, problem structure, and various dependent variables. Results indicate computerized decision aid users had positive attitudes toward the aid and, compared to a group of non-users, (1) considered fewer alternatives, (2) took more time making decisions, and (3) used more analytical tools. Identifying alternatives for decision makers seemed to seriously limit search behavior and alter choice behavior. Judges' assessments indicated that decision quality of the computerized decision aid users was generally worse than that of nonusers. There were few significant interaction effects between problem structure and use of the computerized aid.
Trust is widely acknowledged as an essential ingredient in patient-physician relationships. Given a dearth of situation-specific measures designed to quantify patients' trust in their physicians, we set out to develop an instrument to assess a patient's interpersonal trust in his physician. Findings from two studies are reported describing the development and validation of the Trust in Physician scale. Study 1 of 160 participants provided preliminary support for the reliability (Cronbach alpha = .90) and construct validity of the 11-item scale. Study 2, a replication study of 106 participants, supplied further evidence of the reliability and validity of the scale. Cronbach alpha was .85. Trust was significantly related to patients' desires for control in their clinical interactions and subsequent satisfaction with care. Research and clinical applications of the Trust in Physician scale are discussed.
The effectiveness and safety of large scale technological systems of all kinds is more and more dependent on their command and control. The effectiveness of command and control, in turn, is closely related to perceived trust in them, by operators, managers, and society as a whole. This paper examines concepts of trust, both rational and irrational, and both as cause and effect, and suggests some possibilities for quantitative modeling.
Although trust is an underdeveloped concept in sociology, promising theoretical formulations are available in the recent work
of Luhmann and Barber. This sociological version complements the psychological and attitudinal conceptualizations of experimental
and survey researchers. Trust is seen to include both emotional and cognitive dimensions and to function as a deep assumption
underwriting social order. Contemporary examples such as lying, family exchange, monetary attitudes, and litigation illustrate
the centrality of trust as a sociological reality.
Interpersonal trust is an aspect of close relationships which has been virtually ignored in social scientific research despite its importance as perceived by intimate partners and several family theorists. This article describes the development, validation, and correlates of the Dyadic Trust Scale, a tool designed for such research. It is unidimensional, reliable, relatively free from response biases, and purposely designed to be consistent with conceptualizations of trust from various perspectives. Dyadic trust proved to be associated with love and with intimacy of self-disclosure, especially for longer married partners. It varied by level of commitment, being lowest for ex-partners and highest for those engaged and living together, for newlyweds, and for those married over 20 years. Partners reciprocated trust more than either love or depth of self-disclosure. Future research could fruitfully relate dyadic trust to such issues as personal growth in relationships, resolving interpersonal conflict, and developing close relationships subsequent to separation or divorce.
"Construct validation was introduced in order to specify types of research required in developing tests for which the conventional views on validation are inappropriate. Personality tests, and some tests of ability, are interpreted in terms of attributes for which there is no adequate criterion. This paper indicates what sorts of evidence can substantiate such an interpretation, and how such evidence is to be interpreted." 60 references. (PsycINFO Database Record (c) 2006 APA, all rights reserved).
The aim of the present paper is to study users' perception of computers and human beings as advice givers in problem-solving situations. It will be asked if people's self-confidence and their perception of the advice vary depending on the origin of advice.Two studies showed somewhat different results. In the first study, people were given advice either by a (putative) computer or by a human being. Their self-confidence did not vary with the origin of the advice, but with the correctness of their own answer as well as of the advice. The perception of this advice did not differ for the two situations. Their general trust in computers was, however, much less than their trust in human beings. In the second study, the subjects had to attribute advice to a computer or a human being, without being told from whom the advice emanated. For Swedish subjects, the ratings showed consistently higher attributions to human beings regarding knowledge and explanation value of advice and higher attributions to computers regarding trust and understanding. For Indian subjects, humans always received the higher attributions.It was concluded that people's perception of computers seems to be related both to existing attitudes and to their experience of the advice given. Knowledge in the domain seems to be an important factor influencing the perception of the computer as trustworthy.
Interpersonal trust has long been known to influence cooperation. This study tested the hypothesis that one's degree of trust in others will influence the extent to which one reacts to the presence of fear (or the possibility of receiving no payoff for cooperative actions) in a payoff matrix. The hypothesis was formally tested with public goods games and resource dilemma games, with fear manipulated. Results support the hypothesis: when fear was present, high trusters cooperated more frequently than low trusters; when absent, high and low trusters cooperated at the same rate. The findings held across both games. However, the effects of fear within each game were not straightforward: removing fear from the resource dilemma increased low trusters' cooperation rates, but removing fear from the public goods game decreased high trusters' cooperation rates. Results imply that discussion of the role of trust in cooperation must consider whether the particular dilemma contains an element of fear.
This paper presents a model of trust and its interaction with information flow, influence, and control, and reports on an experiment based on the model to test several hypotheses about problem-solving effectiveness. The subjects were managers and the independent variable was the individual manager's initial level of trust. Groups of business executives were given identical factual information about a difficult manufacturing-marketing policy problem; half the groups were briefed to expect trusting behavior, the other half to expect untrusting behavior. There were highly significant differences in effectiveness between the high-trust groups and the low-trust groups in the clarification of goals, the reality of information exchanged, the scope of search for solutions, and the commitment of managers to implement solutions. The findings indicate that shared trust or lack of trust apparently are a significant determinant of managerial problem-solving effectiveness.
Three existing models of interpersonal trust are tested: (1) attitudinal, (2) situational, and (3) combined. Forty-four subjects were divided into five test groups and a control group. Interpersonal trust scores collected before and after a trust-building t group were analyzed by two-way ANO VA and F ratio. There was a significant increase in trust, measured by t test, in all test groups (p < .05), but no significant change in the control group. The ANO VA findings confirmed the existence of both attitudinal and situationalfactors present in interpersonal trust scores (p < .001) in all test groups. The F ratio demonstrated the situationalfactor as more important in explaining variations in interpersonal trust scores both before and after training (p < .01). Implications of these findings are discussed for traditional attitude theory and more recent contingency (situational) theory.
Calls for new directions in MIS research bring with them a call for renewed methodological rigor. This article offers an operating paradigm for renewal along dimensions previously unstressed. The basic contention is that confirmatory empirical findings will be strengthened when instrument validation precedes both internal and statistical conclusion validity and that, in many situations, MIS researchers need to validate their research instruments. This contention is supported by a survey of instrumentation as reported in sample IS journals over the last several years. A demonstration exercise of instrument validation follows as an illustration of some of the basic principles of validation. The validated instrument was designed to gather data on the impact of computer security administration on the incidence of computer abuse in the U.S.A.
The work we present deals with the trust of man in a teleoperation system. Trust is important because it is linked to stress which modifies human reliability. We are trying to quantify trust. In this paper, we'll present the theory of trust in relationships, and its extension for a man-machine system. Then, we explain the links between trust and human reliability. Then, we introduce our experimental process and the first results concerning selfconfidence.
Numerous researchers have proposed that trust is essential for understanding interpersonal and group behavior, managerial effectiveness, economic exchange and social or political stability, yet according to a majority of these scholars, this concept has never been precisely defined. This article reviews definitions from various approaches within organizational theory, examines the consistencies and differences, and proposes that trust is based upon an underlying assumption of an implicit moral duty. This moral duty-an anomaly in much of organizational theory-has made a precise definition problematic. Trust also is examined from philosophical ethics, and a synthesis of the organizational and philosophical definitions that emphasizes an explicit sense of moral duty and is based upon accepted ethical principles of analysis is proposed. This new definition has the potential to combine research from the two fields of study in important areas of inquiry.
In an investigation of the socialisation of trust, 72 parents and their 50 elementary schoolchildren completed trust belief scales and played a Prisoner's Dilemma (PD) game with each other and with a stranger. In addition, parents described incidents designed to reveal their promise fulfilment to their children. The study yielded positive correlations between: (1) mothers' promise fulfilment to their children and their children's trust beliefs in mothers, fathers, and teachers; and (2) mothers' trust beliefs and their children's trust beliefs in teachers. Additionally, fathers' promised co-operation in the PD game was correlated with their children's promised co-operation in the PD game in interactions with fathers, mothers, and strangers. The findings of this study suggest that mothers shape their children's trust beliefs whereas fathers shape their children's trusting behaviour in a play context.
The present paper reviews the research literature on trust in bargaining and mediation. Several models of trust within the bargaining process are also described. It is concluded that trust means different things, depending upon the relationship under investigation. Trust among negotiators can refer to a personality trail (how trusting a negotiator is of others) or to a temporary state. Within the state perspective, trust often refers to one of three orientations: (1) cooperative motivational orientation (MO), (2) patterns of predictable behavior, (3) a problem-solving orientation. Trust between a negotiator and constituents usually refers to a cooperative MO (i.e., shared loyalty) between these two groups. The addition of a mediator can impact both the opposing negotiators' relationship and each negotiator-constituent relationship; the mediator also has direct and indirect relationships with the parties and their constituents. Future directions for research on trust are identified.
Scholars in various disciplines have considered the causes, nature, and effects of trust. Prior approaches to studying trust are considered, including characteristics of the trustor, the trustee, and the role of risk. A definition of trust and a model of its antecedents and outcomes are presented, which integrate research from multiple disciplines and differentiate trust from similar constructs. Several research propositions based on the model are presented.
The purpose of this study was to develop a scale to assess individual differences in children's ascription to the trust-value basis of friendship. A sample of 130 children (70 girls and 60 boys) from 5th and 6th grades were administered the Chumship Checklist, and the newly constructed Trust-Value Friendship (T-V F) Scale, which measures children's desire for trust-value in forming friendships (friendship preferences) and children's demand for trust-value in existing friends (actual friendships). A subsample of children (103) was administered the T-V F Scale 1 month later. Factor analyses of the T-V F Scale yielded three factors for both friendship preferences and actual friendships scales: (a) trust confirming; (b) trust violating; and (c) school trust. Subscales of the T-V F Scale that were constructed to correspond to the factors demonstrated acceptable internal consistency and test–retest reliability. In support of the validity of the subscales, school trust for friendship preferences and trust confirming for actual friendships were correlated with the Chumship Checklist. The social antecedents and consequences of individual differences in children's ascription to the trust-value basis of friendship are discussed.
Examined the importance of 5 underlying interpersonal trust components (competence, consistency, integrity, loyalty, and openness) as they affect trust among supervisors, subordinates, and peers. Ss were 66 supervisory and managerial level employees who were required to (1) have at least 1 individual reporting to them and (2) report to 1 individual. Questionnaires were used to measure dyadic trust. While the importance of conditions of trust differed within each dyad, the order of importance was the same for all 3 dyads: integrity>competence>loyalty>consistency>openness. Among dyads, no differences were found for the importance of integrity, competence, or loyalty, while consistency and openness were more important for trust in peers than for trust in supervisors or subordinates (PsycINFO Database Record (c) 2012 APA, all rights reserved)
Gave the Imber Children's Trust Scale ICTS), a 40-item situational trust scale which measures father, mother, peer, and teacher trust, to 4th graders (N = 95) in 4 classrooms. The teachers of the 4 classes completed Imber's Teacher Trust Rating Scale (TTRS), a 5-point rating scale devised to measure trustworthiness, trust, security, and dependability of the students. In addition, data were obtained on intelligence scores from prior school testing. Results confirm the hypotheses that classroom achievement would be significantly correlated with both the TTRS and the Teacher subscale of the ICTS. Neither measure was significantly related to intelligence scores. (PsycINFO Database Record (c) 2012 APA, all rights reserved)
Discusses the positive and potential negative consequences of being high or low in interpersonal trust in current social life, particularly in interacting with ordinary people. A summary and analysis of previous investigations led to the following conclusions: People who trust more are less likely to lie and are possibly less likely to cheat or steal. They are more likely to give others a second chance and to respect the rights of others. The high truster is less likely to be unhappy, conflicted, or maladjusted, and is liked more and sought out as a friend more often, by both low-trusting and high-trusting others. When gullibility is defined as naiveté or foolishness and trust is defined as believing others in the absence of clear-cut reasons to disbelieve, then it can be shown over a series of studies that high trusters are not more gullible than low trusters. (17 ref) (PsycINFO Database Record (c) 2012 APA, all rights reserved)
Draws a distinction between validity (logic of inferences) and validation (the process of research using a specific design) to evaluate 11 criterion-related validation designs that differ in the timing of measurement of test behavior and job performance and in the selection procedure (random, existing tests, and experimental tests). Comparisons are based on T. D. Cook and D. T. Campbell's (1949) 4 interrelated criteria: statistical conclusion, internal, construct, and external validity. The analysis marks the selection procedure as a major design property that influences the validity of validation studies. Important differences between concurrent and predictive validation are identified for several aspects of validity. The fact that no specific design is considered most appropriate for all purposes suggests the use of validation research programs, based on a series of studies with different designs, to enhance the validity of validation research. (30 ref) (PsycINFO Database Record (c) 2012 APA, all rights reserved)
Participants in a laboratory experiment (
N = 79) role-played managers mediating a dispute between 2 peers. Building on previous research (e.g., P. J. Carnevale and D. E. Conlon; see record
1988-34828-001) and theory (e.g., D. G. Pruitt, 1981), a 2 × 3 factorial design varied time pressure on the mediators (high vs low time pressure) and trust exhibited between 2 preprogrammed disputants (high trust vs low trust vs a no-message control group). Participants could choose from messages exhibiting Carnevale's (1986) Strategic Choice Model of Conflict Mediation (inaction, pressing, compensating, or integrating), as well as rapport-building messages from K. Kressel's (1972) "reflexive" strategy. Results suggested that high time pressure increased the mediators' use of pressing messages and decreased the use of inaction messages. Participants also sent more reflexive messages when trust was low. Results are discussed in terms of mediation and conflict management theory. (PsycINFO Database Record (c) 2012 APA, all rights reserved)
Developed scales to assess one individual's trust in another in meaningful interpersonal relationships. For males, the scale included factors of reliableness, emotional trust, and general trust. For females, similar but not identical reliableness and emotional trust factors emerged. The scales demonstrated adequate reliability and were discriminable from the related constructs of liking and love. In Exp I, 435 undergraduates' responses on the Reliableness subscale varied appropriately as a function of the reliable or nonreliable behavior of the target person. In Exp II, 84 undergraduates' responses on the Emotional Trust subscale varied appropriately when the target person either betrayed or did not betray a confidence. In both experiments, the appropriate subscale was more sensitive to experimental manipulations than were the other trust subscales, attesting to the discriminant validity of the trust factors. (27 ref) (PsycINFO Database Record (c) 2012 APA, all rights reserved)
Providing explanations for recommended actions is deemed one of the most important capabilities of expert systems (ES). There is little empirical evidence, however, that explanation facilities indeed influence user confidence in, and acceptance of, ES-based decisions and recommendations. This paper investigates the impact of ES explanations on changes in user beliefs toward ES-generated conclusions. Grounded on a theoretical model of argument, three alternative types of ES explanations - trace, justification, and strategy - were provided in a simulated diagnostic expert system performing auditing tasks. Twenty practicing auditors evaluated the outputs of the system in a laboratory setting. The results indicate that explanation facilities can make ES-generated advice more acceptable to users and that justification is the most effective type of explanation to bring about changes in user attitudes toward the system. These findings are expected to be generalizable to application domains that exhibit similar characteristics to those of auditing: domains in which decision making tends to be judgmental and yet highly consequential, and the correctness or validity of such decisions cannot be readily verified.
A multistage, multivariate, recursive path model of the patient-physician treatment relationship and several plausible alternative path models were tested by means of a sample survey (N = 203) in an endeavor to explain the total effects of the quantity and quality of interactions, and the dimensions of interpersonal trust in the determination of client anxiety and pain behavior during the treatment process.While five of the eight hypotheses received empirical support from correlation analysis, the results of path analysis suggested that the relationships between interpersonal trust in the physician and the level of treatment anxiety, and perceived positive health gains from treatment and the level of treatment anxiety were augmented by the direct and indirect effects of past successful treatment.Several theoretical and practical implications of the results are discussed, including the application of the findings towards a theory of client-specialist relationships. In addition, a number of weaknesses in the internal validity of the study are addressed.
There is a danger inherent in labeling systems “expert.” Such identification implies some levels of “intelligence” or “understanding” within the confines of the system. It is important to know the limitations of any system, including realistic expectations of the real or implied power of an expert system. The “blindness” or boundaries inherent in expert system development extends to users who may misplace trust in false technology.This study investigates the use of an incorrect advice-giving expert system. Expert and novice engineers used the faulty system to solve a well test interpretation task. Measures of decision confidence, system success, state-anxiety and task difficulty were taken. Subjects expressed confidence in their “wrong” answer to the problem, displaying true dependence on a false technology. These results suggest implications for developers and/or users in areas of certification, evaluation, risk assessment, validation, and verification of systems conveying a level of “expertise.”
This field investigation studied the use of an expert system technology to gain some additional insight into specifiic behavioral implications for information system designers. Twenty-eight engineers in an oil and gas exploration and production company participated in this study by solving a well pressure buildup analysis problem. Half of the subjects utilized a well test interpretation expert system to assist them while the other subjects solved the problem manually. The groups were balanced across age, cognitive style and trait anxiety. Independent variables consisted of the expert system treatment, dogmatism and experience with performing the task. Impact measures consisted of decision confidence, decision quality, decision time, state anxiety and a system success indicator for those subjects utilizing the expert system.Although decision confidence was higher in the group utilizing the expert system, there was no corresponding increase in decision quality. Also, experts utilizing the expert system experienced an increase in state anxiety, and rated the expert system significantly worse than the novices did. This may imply that expert system technology may be more useful or appropriate to novices than experts.
A problem in the design of decision aids is how to design them so that decision makers will trust them and therefore use them appropriately. This problem is approached in this paper by taking models of trust between humans as a starting point, and extending these to the human-machine relationship. A definition and model of human-machine trust are proposed, and the dynamics of trust between humans and machines are examined. Based upon this analysis, recommendations are made for calibrating users' trust in decision aids.
This article advocates, as alternative for either the "classical" technology- or user-centred approach, to focus on the joint human-computer task performance in system design. Human involvement can be improved by designing system functions which complement human knowledge and capacities. Based on general needs for cognitive support, an aiding function is proposed which—in the process of task execution—takes the initiative to present context-specific, procedural task knowledge. Design of such aiding comprises two aspects: design of software and design of a human-computer system. Modern model-based software engineering methods provide strong support for the design of software systems, but little support for modelling the human-computer interaction. Current model-based methods are extended to address human-computer interaction issues. The resulting method comprises the design of easy-to-use-and-learn interfaces providing, if needed, aiding. In a case study, the method is applied to design a conventional plain interface and an aiding interface for the statistical program HOMALS. In an experiment, users with minor HOMALS expertise prove to perform their tasks better and to learn more with the aiding interface.
We show that individuals use inappropriate social rules in assessing machine behavior. Explanations of ignorance and individuals' views of machines as proxies for humans are shown to be inadequate; instead, individuals' responses to technology are shown to be inconsistent with their espoused beliefs. In two laboratory studies, computer-literate college students used computers for tutoring and testing. The first study (n = 22) demonstrates that subjects using a computer that praised itself believed that it was more helpful, contributed more to the subject's test score, and was more responsive than did subjects using a computer that criticized itself, although the tutoring and testing sessions were identical. In the second study (n = 44), the praise or criticism came from either the computer that did the tutoring or a different computer. Subjects responded as if they attributed a "self" and self-focused attributions (termed "ethopoeia") to the computers. Specifically, subjects responses followed the rules "other-praise is more valid and friendlier than self-praise", "self-criticism is friendlier than other-criticism", and "criticizers are smarter than praisers" to evaluate the computers, although the subjects claimed to believe that these rules should not be applied to computers.
This paper reports on the development of an instrument designed to measure the various perceptions that an individual may have of adopting an information technology (IT) innovation. This instrument is intended to be a tool for the study of the initial adoption and eventual diffusion of IT innovations within organizations. While the adoption of information technologies by individuals and organizations has been an area of substantial research interest since the early days of computerization, research efforts to date have led to mixed and inconclusive outcomes. The lack of a theoretical foundation for such research and inadequate definition and measurement of constructs have been identified as major causes for such outcomes. In a recent study examining the diffusion of new end-user IT, we decided to focus on measuring the potential adopters' perceptions of the technology. Measuring such perceptions has been termed a "classic issue" in the innovation diffusion literature, and a key to integrating the various findings of diffusion research. The perceptions of adopting were initially based on the five characteristics of innovations derived by Rogers (1983) from the diffusion of innovations literature, plus two developed specifically within this study. Of the existing scales for measuring these characteristics, very few had the requisite levels of validity and reliability. For this study, both newly created and existing items were placed in a common pool and subjected to four rounds of sorting by judges to establish which items should be in the various scales. The objective was to verify the convergent and discriminant validity of the scales by examining how the items were sorted into various construct categories. Analysis of inter- judge agreement about item placement identified both bad items as well as weaknesses in some of the constructs' original definitions. These were subsequently redefined. Scales for the resulting constructs were subjected to three separate field tests. Following the final test, the scales all demonstrated acceptable levels of reliability. Their validity was further checked using factor analysis, as well as conducting discriminant analysis comparing responses between adopters and nonadopters of the innovation. The result is a parsimonious, 38-item instrument comprising eight scales which provides a useful tool for the study of the initial adoption and diffusion of innovations. A short, 25-item, version of the instrument is also suggested.
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 time 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.