Article

Towards context aware data fusion: Modeling and integration of situationally qualified human observations to manage uncertainty in a hard + soft fusion process

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

Abstract

This paper presents a framework for characterizing errors associated with different categories of human observation combined with a method for integrating these into a hard + soft data fusion system. Error characteristics of human observers (often referred to as soft data sensors) have typically been artificially generated and lack contextual considerations that in a real-world application can drastically change the accuracy and precision of these characteristics. The proposed framework and method relies on error values that change based upon known and unknown states of qualifying variables empirically shown to affect observation accuracy under different contexts. This approach allows fusion systems to perform uncertainty alignment on data coming from human observers. The preprocessed data yields a more complete and reliable situation assessment when it is processed by data association and stochastic graph matching algorithms. This paper also provides an approach and results of initial validation testing of the proposed methodology. The testing approach leverages error characterization models for several exemplar categories of observation in combination with simulated synthetic data. Results have shown significant performance improvements with respect to both data association and situation assessment fusion processes with an average F-measure improvement of 0.16 and 0.20 for data association and situation assessment respectively. These F-measure improvements are representative of fewer incorrect and missed associations and fewer graph matching results, which then must be considered by human analysts. These benefits are expected to translate into a reduction of the overall cognitive workload facing human analysts in situations where they are tasked with developing and maintaining situational awareness.

No full-text available

Request Full-text Paper PDF

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

... Indudablemente, el interés por la fusión de datos es cada vez mayor debido a la creciente incorporación de sensores en los dispositivos y sistemas de cómputo, el objetivo es tener información útil como apoyo en el proceso de la toma de decisiones sobre un determinado suceso, objeto o acción [5,6]. Recientemente se está empleando también la fusión de datos para integrar fuentes variadas para hacer detecciones y clasificaciones de actividades en hogares inteligentes [9], ambientes virtuales inmersivos [10], interfaces tangibles [11] y escritorios inteligentes [12], apoyo en la realización de actividades del usuario [4], fusión de datos de observaciones humanas [13], entre otros. ...
... Algunos métodos utilizados para minimizar la inconsistencia de los datos son [4,13,31]: teoría de razonamiento evidencial Dempster Shafer, lógica difusa, estimación Bayesiana, filtro de Kalman, entre otros. Para la fusión de datos estos métodos toman en cuenta los factores ambientales y la imprecisión de los sensores que afectan las mediciones, así como la ambigüedad de los datos y la dificultad para distinguirlos [32]. ...
... La selección del algoritmo depende de la naturaleza, el tipo y el nivel de abstracción de los datos, como: datos, atributos, símbolos, identidad, patrones o decisiones. Estos algoritmos deben ser capaces de distinguir la inconsistencia y combinar diferentes tipos de datos, a diferentes niveles, para obtener una descripción consistente [13,32]. Por lo que, para este trabajo los algoritmos definidos como parte del proceso de fusión son: ...
Article
Context–aware systems use data obtained from various sources to adapt and provide services of interest to the user according to their needs, location or interaction with the environment. However, the use of heterogeneous sources creates a large volume of data that may differ in format, transmission speed and may be affected by environmental noise. This generates some inconsistency in the data, which must be detected in time to avoid erroneous analysis. This is done using data fusion, which is the action of integrating diverse sources to be analyzed according to a determined context. This paper presents the conceptual design of a data fusion method of heterogeneous sources, obtained from contextual information, with the aim of maintaining the consistency of the data during the fusion process (extraction, preprocessing, fusion, and loading data).
... The context is the physical, emotional and social environment in which the user is immersed and gives sense and value to the actions or activities that take place around the user [2]. Through context analysis one may characterize the status of an entity, such as people, place or object, which is considered relevant for the interaction between user and system [3], [4]. ...
... Context-aware systems have used data fusion [4], [6], [9], [23] to integrate data from various sources. In this sense, data fusion is a key and critical aspect of systems with heterogeneous sources. ...
... In this sense, data fusion is a key and critical aspect of systems with heterogeneous sources. The objective is to integrate data efficiently from multiple sources to overcome the limitations encountered when using a single source [4]. In this sense, data fusion from heterogeneous sources represents a significant advantage in unifying diverse data, to obtain a global, robust and complete view of the event or situation that is being observed and to be analyzed. ...
... In this sense, a complex activity, prior to the context analysis, is data fusion from heterogeneous sources. This data fusion is the process for detecting, associating, correlating, estimating and combining data at various levels [51], which come from different sources, such as sensors and databases [18], signals and decisions [13] and even human observations [22] or expert experience [6]. ...
... This knowledge field has been employed in several areas, such as signal processing [49], information theory [22], estimation [18], inference and artificial intelligence [30]; taking greater advances in military applications through automatic target recognition, autonomous vehicle navigation, remote sensing and threat identification [19,31,32,36]. Other non-military applications are industrial processes monitoring [24], robotics [42,47], medical applications [27], among others. ...
... Recently, data fusion has also been used to integrate various sources in the Internet of Things [11], fault detection in industrial processes [17,24], fusion of quantitative and qualitative data [6,22] as well as making detection and classification of activities in intelligent homes and environments [5,38], immersive virtual environments [44], tangible interfaces [7,35] and smart desks [54]. ...
Article
Full-text available
Nowadays, context-aware systems use data obtained from various sources to adapt and provide services of interest to users according to their needs, location or interaction with the corresponding environment. However, the use of heterogeneous sources creates a huge amount of data that may differ in format, transmission speed and may be affected by environmental noise. This generates some inconsistency in data, which must be detected in time to avoid erroneous analysis. This is done using data fusion, which is the action for integrating diverse sources to be analyzed according to a given context. In this work, we propose a scheme of data fusion of heterogeneous sources, supported by a distributed architecture and Bayesian inference as fusion method. As a practical experiment, data were collected from three DHT22 sensors, whose measurements were relative humidity and temperature. The purpose of the experiment was to analyze the variation of these measurements over 24 hours, and fusion them to obtain integrated data. This proposed of data fusion represents an important field of action for the knowledge generation of interest in context-aware systems, for example for the analysis of the environment in order to take advantage of the use of energy and provide a comfortable working environment for the users.
... These two methods look at the uncertainty from different points of view. In literature one can find various terms for fuzzy data, such as possibilistic, soft and subjective [3], as opposed to random called probabilistic, hard and objective [21]. These terms are somewhat arbitrary and there are authors who used probability distributions to represent subjective information [43,44,45]. ...
... It is more representation of our confidence level in an uncertain phenomenon. If a need arises for fuzzy-random data fusion, [7], [21], each distribution is typically handled separately for a specific problem at hand. To our knowledge no rigorous mathematical methodologies exist for a practical uncertainty alignment between two types of data. ...
... The opposite applies as well, i.e. given two or more random events their interplay can be interpreted as a fuzzy event. We believe our approach can bring about new avenues in aligning fuzzy and random data, in particular in very important area of soft-hard (humanmachine) data fusion [21]. In our previous papers [40], [41], we presented the basics of our uncertainty alignment methodology. ...
Chapter
Full-text available
The paper advances our on going work in the area of uncertainty alignment and transformation between fuzzy (soft, human generated, possibilistic) and random (hard, machine generated, probabilistic) data. As reported in our previous papers, the Uncertainty Balance Principle was defined to express uncertain data vagueness as represented by a fuzzy data models, with a non uniqueness of related random data distributions. The underlying assumption is that both fuzzy and random data are described in terms of the same independent uncertain variables. The connection between fuzzy and random data is done via cumulative rather than probability density functions. In this paper we clarify and extend our previous work whereas an initial fuzzy distributions (membership functions) are supplied and the aim is to determine corresponding and related random distributions. The next step in this analysis will focus on Bayesian data mining to determine random distributions from a given large set (data base) of soft data modeled as fuzzy (triangular, trapezoidal or other convex) distributions. This work has been inspired by an ever increasing need to fuse human and machine data in order to perform decision making procedures. Areas of applications include Bank Risk Assessment in financial industry as well as Command and Control Integration in defense industry and any other applications where soft and hard data fusion is required.
... These two methods look at the uncertainty from different points of view. In literature one can find various terms for fuzzy data, such as possibilistic, soft and subjective [3], as opposed to random called probabilistic, hard and objective [21]. These terms are somewhat arbitrary and there are authors who used probability distributions to represent subjective information [43,44,45]. ...
... It is more representation of our confidence level in an uncertain phenomenon. If a need arises for fuzzy-random data fusion, [7], [21], each distribution is typically handled separately for a specific problem at hand. To our knowledge no rigorous mathematical methodologies exist for a practical uncertainty alignment between two types of data. ...
... The opposite applies as well, i.e. given two or more random events their interplay can be interpreted as a fuzzy event. We believe our approach can bring about new avenues in aligning fuzzy and random data, in particular in very important area of soft-hard (humanmachine) data fusion [21]. In our previous papers [40], [41], we presented the basics of our uncertainty alignment methodology. ...
Conference Paper
Full-text available
The paper advances our on going work in the area of uncertainty alignment and transformation between fuzzy (soft, human generated, possibilistic) and random (hard, machine generated, probabilistic) data. As reported in our previous papers, the Uncertainty Balance Principle was defined to express uncertain data vagueness as represented by a fuzzy data models, with a non uniqueness of related random data distributions. The underlying assumption is that both fuzzy and random data are described in terms of the same independent uncertain varia-bles. The connection between fuzzy and random data is done via cumulative rather than probability density functions. In this paper we clarify and extend our previous work whereas an initial fuzzy distributions (membership functions) are supplied and the aim is to determine corresponding and related random distributions. The next step in this analysis will focus on Bayesian data mining to determine random distributions from a given large set (data base) of soft data modeled as fuzzy (triangular, trapezoidal or other convex) distributions. This work has been inspired by an ever increasing need to fuse human and machine data in order to perform decision making procedures. Areas of applications include Bank Risk Assessment in financial industry as well as Command and Control Integration in defense industry and any other applications where soft and hard data fusion is required
... These two methods look at the uncertainty from different points of view. In literature one can find various terms for fuzzy data, such as possibilistic, soft and subjective [3], as opposed to random called probabilistic, hard and objective [21]. These terms are somewhat arbitrary and there are authors who used probability distributions to represent subjective information [42,43,44]. ...
... It is more representation of our confidence level in an uncertain phenomenon. If a need arises to combine fuzzy and random data, such as in soft/hard data fusion, [7], [21], each distribution is typically handled separately for a specific problem at hand, and to our knowledge no rigorous mathematical methodologies exist for a practical uncertainty alignment between two types of data. In a fundamental paper by Zadeh [2], a concept of possibilistic fuzzy distributions was introduced as opposed to random and probabilistic distributions. ...
... In the case of multimodal fuzzy data, this representation consists of a number of fixed and variable random events. We believe our approach can bring about new avenues in aligning fuzzy and random data, in particular in very important area of soft-hard (humanmachine) data fusion [21]. In our previous introductory paper [40] we presented the basics of our uncertainty alignment methodology. ...
Article
Full-text available
The objective of this paper is to present new and simple mathematical approach to deal with uncertainty transformation for fuzzy to random or random to fuzzy data. In particular we present a method to describe fuzzy (possibilistic) distribution in terms of a pair (or more) of related random (probabilistic) events, both fixed and variable. Our approach uses basic properties of both fuzzy and random distributions, and it assumes data is both possibilistic and probabilistic. We show that the data fuzziness can be viewed as a non uniqueness of related random events, and prove our Uncertainty Balance Principle. We also show how Zadeh’s fuzzy-random Consistency Principle can be given precise mathematical meaning. Various types of fuzzy distributions are examined and several numerical examples presented.
... CR modules convert the content provided by IE modules into a common format via, for instance: (1) uncertainty alignment [42], as the heterogeneous data sources may express it in inconsistent forms (e.g., hard sensor uncertainty is expressed in probabilistic/quantitative format whereas soft sensor uncertainty is stated in possibilistic/fuzzy terms due to the qualitative nature of linguistic expression); or (2) temporal alignment [43], which is required when the disparate data sources present temporal features in incongruent forms (e.g., exact timestamp vs. approximate time of day vs. linguistic phrasing encompassing multiple verbal tense expressions). ...
... PT was used to integrate imprecise qualitative data in a soft fusion framework in [42], and to model the uncertainty in soft data in an SA system in [55]. ...
... Innate differences between physical sensors and humans make it difficult or impossible to use these MOPs and MOEs when people are integrated into the fusion process; physical sensors can be calibrated for consistent performance, and although people can be trained, the differences in internal factors (such as cognitive abilities, biases, and stress) can cause unknown performance variation [42] [58] [59]. ...
Thesis
Full-text available
This thesis presents a framework capable of integrating hard (physics-based) and soft (people-generated) data for the purpose of achieving increased situational assessment (SA) and effective course of action (CoA) generation upon risk identification. The proposed methodology is realized through the extension of an existing Risk Management Framework (RMF). In this work, the RMF’s SA capabilities are augmented via the injection of soft data features into its risk modeling; the performance of these capabilities is evaluated via a newly-proposed risk-centric information fusion effectiveness metric. The framework’s CoA generation capabilities are also extended through the inclusion of people-generated data, capturing important subject matter expertise and providing mission-specific requirements. Furthermore, this work introduces a variety of CoA-related performance measures, used to assess the fitness of each individual potential CoA, as well as to quantify the overall chance of mission success improvement brought about by the inclusion of soft data. This conceptualization is validated via experimental analysis performed on a combination of real- world and synthetically-generated maritime scenarios. It is envisioned that the capabilities put forth herein will take part in a greater system, capable of ingesting and seamlessly integrating vast amounts of heterogeneous data, with the intent of providing accurate and timely situational updates, as well as assisting in operational decision making.
... Data fusion systems combine information originating from multiple, heterogeneous sources into an estimated state of the environment in order to contribute to better decisionmaking. Data may come from either "hard sources" (physical sensors) or "soft sources" (including both qualitative or quantitative estimations provided by human observers) (Jenkins, Gross, Bisantz, & Nagi, 2015;Kerker, Jenkins, Gross, Bisantz, & Nagi, 2014). Understanding how people use language to provide information about situations and events is critical in the design of grounded language systems that seek to incorporate such descriptions into situational estimates (Roy & Reiter, 2005). ...
... It is challenging to understand characteristics of human estimates (e.g. accuracy or precision) over a range of situational and environmental conditions (Jenkins et al., 2015). Soft data also requires unique processing methodologies due to the variety and ambiguity inherent to natural language (Gross et al., 2012). ...
... Soft data also requires unique processing methodologies due to the variety and ambiguity inherent to natural language (Gross et al., 2012). A necessary first step in utilizing qualitative data in the fusion process is to map it to a quantitative value, or distribution of values, in order for it to be processed by a computer (Jenkins et al., 2015). However, standard methods or definitions to systematically and accurately convert linguistic terms into quantitative data are lacking. ...
Article
Qualitative linguistic data provides unique, valuable information that can only come from human observers. Data fusion systems find it challenging to incorporate this “soft data” as they are primarily designed to analyze quantitative, hard-sensor data with consistent formats and qualified error characteristics. This research investigates how people produce linguistic descriptions of human physical attributes. Thirty participants were asked to describe seven actors’ ages, heights, and weights in two naturalistic video scenes, using both numeric estimates and linguistic descriptors. Results showed that not only were a large number of linguistic descriptors used, but they were also used inconsistently. Only 10% of the 189 unique terms produced were used by four or more participants. Especially for height and weight, we found that linguistic terms are poor devices for transmitting estimated values due to the large and overlapping ranges of numeric estimates associated with each term. Future work should attempt to better define the boundaries of inclusion for more frequently used terms and to create a controlled language lexicon to gauge whether or not that improves the precision of natural language terms.
... Unlike physical sensors, human observations are not precalibrated, meaning uncertainties must be considered prior to analysis. This point was illustrated in the research team's past work on uncertainty alignment of human observational data [12,13]. Uncertainty alignment uses context-aware models of human observations to correct for observational biases and variance, specifically by combining data from three main contextual categories that have been empirically shown to influence human estimation capabilities: [1] Observer characteristics (e.g., observers display tendencies to anchor by their own characteristics when estimating those of a target individual [14][15][16][17]); [2] Target characteristics (e.g., familiarity of the target to the observer [18], true attribute values of the target [19]); [3] Environmental characteristics (e.g., observations made in the dark vs. the light [20]). ...
... Next, the variance and distribution family (as indicated by the uncertainty alignment model) are applied to the bias adjusted observation, resulting in the true value containing distribution. The numerical results from these past efforts point to the tangible benefits of considering uncertainty aligned information to fusion tasks such as data association and situation assessment [13]. The uncertainty alignment models developed in the team's past study [13] relied upon contextual factors and observational biases and variances extracted from empirical literature. ...
... The numerical results from these past efforts point to the tangible benefits of considering uncertainty aligned information to fusion tasks such as data association and situation assessment [13]. The uncertainty alignment models developed in the team's past study [13] relied upon contextual factors and observational biases and variances extracted from empirical literature. This literature spanned multiple domains (e.g., perception, psychology, psychophysics, human factors), but typically focused only on narrow application areas and particular sets of contextual factors (e.g., age proofing for tobacco and alcohol sales [14]). ...
Conference Paper
Full-text available
Visual estimations of target attributes in a realworld environment are highly context-dependent when the estimations are provided by human observers. For example, the accuracy of an individual estimating the age, height, or weight of another person is dependent upon environmental (e.g., viewing distance), observer (e.g., age/height/weight), and target (e.g., clothing, gate) factors. Prior efforts have attempted to characterize the ability of humans to estimate attributes of other humans; however, these studies typically only present observers with static images in controlled settings. The present study instead characterizes observations of attributes made of a more dynamic, real world. Participants provide estimates of target individuals' ages, heights, and weights, along with other descriptive data, as they watched video recorded scenes of simulated, realistic security incidents. Results indicate the anchoring effect demonstrated in prior efforts may not be as prevalent under more ecologically-valid viewing conditions; however, individuals are still able to provide relatively accurate estimations of individuals' age, height, and weight, with minimal influence of the observers' own physical attributes.
... The candidate classifiers are traditionally machines (socalled hard sensors). More recently, researchers started to investigate the advantages of keeping humans (so-called soft sensors) in the loop [8]- [10]. Human-machine collaboration allows pattern recognition protocols to exploit the unique capabilities of both humans and machines. ...
... , 1, 2, 3, . . . , 8,9,10 . (3) A specific trial is executed by considering all of the cost gradients in (3). ...
Article
Full-text available
Despite growing interest in human-machine collaboration for enhanced decision-making, little work has been done on the optimal fusion of human and machine decisions for cost-sensitive biometric authentication. An elegant and robust protocol for achieving this objective is proposed. The merits of the protocol is illustrated by simulating a scenario where a workforce of human experts and a score-generating machine are available for the authentication of handwritten signatures on, for example, bank cheques. The authentication of each transaction is determined by its monetary value and the quality of the claimed author’s signature. A database with 765 signatures is considered, and an experiment that involves 24 human volunteers and two different machines is conducted. When a reasonable number of experts are kept in the loop, the average expected cost associated with the workforce-machine hybrid is invariably lower than that of the unaided workforce and that of the unaided machine.
... This is considered hard data fusion. The main idea is to associate different pieces of data based on the likelihood that they represent the same or related events, people, and objects [1]. On the other hand, humans produce soft subjective data most of the time, which can not be modeled by probabilistic means. ...
... We believe this new approach reduces HO errors, and increases effectiveness of HOME Team. The problem which human errors pose for soft data and hard/soft fusion process is well described in [1,2,18]. We believe that our approach and SDBase built-in-learning feature diminishes this issue, and produces better hybrid fusion process. ...
... This is considered hard data fusion. The main idea is to associate different pieces of data based on the likelihood that they represent the same or related events, people, and objects [1]. On the other hand, humans produce soft subjective data most of the time, which can not be modeled by probabilistic means. ...
... We believe this new approach reduces HO errors, and increases effectiveness of HOME Team. The problem which human errors pose for soft data and hard/soft fusion process is well described in [1,2,18]. We believe that our approach and SDBase built-in-learning feature diminishes this issue, and produces better hybrid fusion process. ...
Chapter
Full-text available
The overall goal in this concept paper is to present an innovative and rigorous mathematical methodology and an expert self learning data fusion and decision platform, which is scalable and effective for a variety of applications. This includes (i) Interface design that incorporates the understanding of how both machines and humans fuse soft and hard data and information, and (ii) Forming a shared perception and understanding of the environment between the human and the machine, which supports human decisions and reduces human soft and decision making errors. With this paper we continue our research on Uncertainty Balance Principle (UBP) which is at the core of our soft hard data fusion and decision making strategy. The proposed methodology can be employed in the context of Artificial Intelligence (AI) and Machine Learning (ML) applications, such as banking risk assessment, Block Chain peer to peer systems, ecological and climate modeling, social sciences, econometrics, as well as defense applications such as battle management.
... This is considered hard data fusion. The main idea is to associate different pieces of data based 2 on the likelihood that they represent the same or related events, people, and objects [1]. On the other hand, humans produce soft subjective data most of the time, which can not be modeled by probabilistic means. ...
... We believe this new approach reduces HO errors, and increases effectiveness of HOME Team. The problem which human errors pose for soft data and hard/soft fusion process is well described in [1,2,18]. We believe that our approach and SDBase built-in-learning feature diminishes this issue, and produces better hybrid fusion process. ...
... This makes perfect intuitive sense as well. We have the same situation for Case 2, except that the non unique probability is now P(A 1 ). Figure 2 to [5,10,20]. This corresponds to a non symmetric TFN with a larger spread of x. ...
... Π(x) 1 1 dΠ/dx d(∆P)/dx probabilistic, sensor based) data are to be fused with subjective (soft, fuzzy, possibilistic, human based) data [9], [10]. One way to interpret the results is as a precise mathematical description of fuzzy-to-random "consistency principle" first introduced by Zadeh in his classic paper [2], as a loose and intuitive notion. ...
Article
Full-text available
The objective of this paper is to present new and simple mathematical approach to deal with uncertainty alignment between fuzzy and random data. In particular we present a method to describe fuzzy (possibilistic) distribution in terms of a pair (or more) of related random (probabilistic) events, both fixed and variable. Our approach uses basic properties of both fuzzy and random distributions. We show that the data fuzziness can be viewed as a non uniqueness of related random events. We also show how fuzzy-random consistancy principle can be given precise mathemtaical meaning. Various types of fuzzy distributions are examined, special cases considered, and several numerical examples presented.
... In CNO, soft and hard data are mass generated. Hard data are generated by nonhuman entities, whereas soft data are generated by a human in natural language [22]. Additionally, the data can be classified according to their frequency of use as cold data, which are accessed sporadically, and hot data, which are accessed frequently. ...
Article
Full-text available
A Collaborative Networked Organization (CNO) is a set of entities that operate in heterogeneous contexts and aim to collaborate to take advantage of a business opportunity or solve a problem. Big data allows CNOs to be more competitive by improving their strategy, management and business processes. To support the development of big data ecosystems in CNOs, several frameworks have been reported in the literature. However, these frameworks limit their application to a specific CNO manifestation and cannot conduct intelligent processing of big data to support decision making at the CNO. This paper makes two main contributions: (1) the proposal of a metaframework to analyze existing and future frameworks for the development of big data ecosystems in CNOs and (2) to show the Collaborative Networked Organizations–big data (CNO-BD) framework, which includes guidelines, tools, techniques, conceptual solutions and good practices for the building of a big data ecosystem in different kinds of Collaborative Networked Organizations, overcoming the weaknesses of previous issues. The CNO-BD framework consists of seven dimensions: levels, approaches, data fusion, interoperability, data sources, big data assurance and programmable modules. The framework was validated through expert assessment and a case study.
... For example, Larry Stone worked with another Metron scientist, Jim Ferry, on search theory for fused physical and online location data (Popp and Yen, 2006). We note that data fusion methods and event representation models (e.g., using networks (Jenkins et al., 2015;Farasat et al., 2016)) are likely to be useful research directions in such efforts as well. Dr. Smith first pointed out that many efforts in security research seek to assess the vulnerability of a system (whether it is a network or not) to failure. ...
Article
Full-text available
Homeland security research has gone through a significant transformation since the events of September 11, 2001, and continues to evolve. This article identifies opportunities that the industrial engineering and operations research communities can seize. By drawing together insights from thought leaders in these communities, a path outlining research problems and discovery is provided that will serve to guide industrial engineering and operations research innovations and help move homeland security research forward over the next decade.
... In the military domain, threat assessment is an example of a complex problem of this kind, which may trigger a sequence of actions and decisions with high cost or consequence [6,7,8]. Threat assessment can be solved as a multi-intelligence (multi-INT) fusion problem [9,7], in accordance with the paradigm of Activity-Based Intelligence [10,11]. Indeed, relevant information may be sourced by different agencies, from different nations, in a variety of formats and shared in a more or less trustful way. ...
Article
Explanation abilities are required for data-driven models, where the high number of parameters may render its internal reasoning opaque to users. Despite the natural transparency brought by the graphical model structure of Bayesian networks, decisions trees or valuation networks, additional explanation abilities are still required due to both the complexity of the problem as well as the consequences of the decision to be taken. Threat assessment is an example of such a complex problem in which several sources with partially unknown behaviour provide information on distinct but related frames of discernment. In this paper, we propose a solution as an evidential network with explanation abilities to detect and investigate threat to maritime infrastructure. We propose a post-hoc explanation approach to an already transparent by design threat assessment model, combining feature relevance and natural language explanations with some visual support. To this end, we extend the sensitivity analysis method of generation of explanations for evidential reasoning to a multi-source model where sources can have several and disparate behaviours. Natural language explanations are generated on the basis of a series of sensitivity measures quantifying the impact of both direct reports and source models. We conclude on challenges to be addressed in future work.
... 。 地面是人类周围环境中最重要的一个表面。 大量研究表明, 在中等距离范围内(约 2 米以外 30 米以内), 人类是以地面为参照框架对置于其表面 上的物体的空间布局进行编码的 Feria et al., 2003;He & Ooi, 2000;Bian et al., 2005Bian et al., , 2006Thompson & Creem-Regehr, 2007;Bian & Andersen, 2010;Kavšek & Granrud, 2013;Loomis, 2014;Wu, Zhou, Shi, He, & Ooi, 2015;Nakashima & Kumada, 2017;Loyola, 2018 , 且得到了大 量实证研究的支持 (Ooi, Wu, & He, 2006;Jansen, Toet, & Werkhoven, 2011;Bian & Andersen, 2011; 周 佩 灵 , 黎 安 娟 , 2011; Wu, He, & Ooi, 2014;Erkelens, 2015;Jenkins et al., 2015;Norman et al., 2015;Proulx et al., 2016 ...
... The candidate classifiers are traditionally machines (socalled hard sensors). More recently, researchers started to investigate the advantages of keeping humans (so-called soft sensors) in the loop [6], [7]. Humans are proficient at integrating information and incorporating context, while machines are adept at making fast, consistent and objective decisions. ...
... The main purpose of this study is to investigate and test whether the accuracy of probabilistic credit risk assessment of corporates, evaluated with logistic regression, can be improved using soft and hard data modeling, followed by soft-hard data fusion, in particular using Uncertainty Balance Principle. In literature one can find various methods on how to integrate fuzzy and random data in meaningful ways [34][35][36]53], as well as about the area of "random fuzzy sets" and "fuzzy random variables", as well as various "fuzzy" applications [e.g., 1-2, 5, 11-20, 22-24, 33, 37, 39-45, 47-52]. However, this study does not deal with aforementioned. ...
Article
Full-text available
This study introduces Uncertainty Balance Principle (UBP) as a new concept/method for incorporating additional soft data into probabilistic credit risk assessment models. It shows that soft banking data, used for credit risk assessment, can be expressed and decomposed using UBP and thus enabling more uncertainty to be handled with a precise mathematical methodology. The results show that this approach has relevance to credit risk assessment models in the sense that it proved its usefulness for the purpose of soft-hard data fusion, it modified Probability of Default with soft data modeled using possibilistic (fuzzy) distributions and fused with hard probabilistic data via UBP and it obtained better classification prediction results on the overall sample. This was demonstrated on a simple example of one soft variable, two experts and a small sample and thus this is an approach/method that requires further research, enhancements and rigorous statistical testing for the application to a complete scoring and/or rating system
... The main purpose of this study is to investigate and test whether the accuracy of probabilistic credit risk assessment of corporates, evaluated with logistic regression, can be improved using soft and hard data modeling, followed by soft-hard data fusion, in particular using Uncertainty Balance Principle. In literature one can find various methods on how to integrate fuzzy and random data in meaningful ways [34][35][36]53], as well as about the area of "random fuzzy sets" and "fuzzy random variables", as well as various "fuzzy" applications [e.g., 1-2, 5, 11-20, 22-24, 33, 37, 39-45, 47-52]. However, this study does not deal with aforementioned. ...
Article
Full-text available
This study introduces Uncertainty Balance Principle (UBP) as a new concept/method for incorporating additional soft data into probabilistic credit risk assessment models. It shows that soft banking data, used for credit risk assessment, can be expressed and decomposed using UBP and thus enabling more uncertainty to be handled with a precise mathematical methodology. The results show that this approach has relevance to credit risk assessment models in the sense that it proved its usefulness for the purpose of soft-hard data fusion, it modified Probability of Default with soft data modeled using possibilistic (fuzzy) distributions and fused with hard probabilistic data via UBP and it obtained better classification prediction results on the overall sample. This was demonstrated on a simple example of one soft variable, two experts and a small sample and thus this is an approach/method that requires further research, enhancements and rigorous statistical testing for the application to a complete scoring and/or rating system
... They discuss the use of context in the classical lower-level tasks of sensor characterisation, signal fusion, data association, filtering, tracking and classification, and consider as well the higher-level tasks of knowledge representation, situation assessment, decision making, intent assessment and process refinement. In relation to natural language processing, context is also used in hard and soft fusion [7] to incorporate soft information (i.e. from human sources) with hard information (i.e. from sensors) using fuzzy membership functions to define the semantics of vague concepts. Indeed, with the growth of social media data processing, the notion of context is instrumental to natural language processing techniques, where words and sentences often need some contextual setting to be properly understood. ...
... Machines are more efficient at combining structured, hard (machine-derived) data. Humans are currently better than machines at combining soft (humanderived) data as it is usually unstructured [16] and uncertainty levels are unknown [17,18]. One human role is frequently to support automated computer reasoning techniques with visual and aural pattern recognition and semantic reasoning [19]. ...
... It makes more flexible to reproduce the desired sensing data. Besides that, the context-aware, semantic information matching [19], and fuzzy reasoning methods [20,21] are also utilized to adjust the output data as required. ...
Article
Full-text available
Sensor fusion techniques have made a significant contribution to the success of the recently emerging mobile applications era because a variety of mobile applications operate based on multi-sensing information from the surrounding environment, such as navigation systems, fitness trackers, interactive virtual reality games, etc. For these applications, the accuracy of sensing information plays an important role to improve the user experience (UX) quality, especially with gyroscopes and accelerometers. Therefore, in this paper, we proposed a novel mechanism to resolve the gyro drift problem, which negatively affects the accuracy of orientation computations in the indirect Kalman filter based sensor fusion. Our mechanism focuses on addressing the issues of external feedback loops and non-gyro error elements contained in the state vectors of an indirect Kalman filter. Moreover, the mechanism is implemented in the device-driver layer, providing lower process latency and transparency capabilities for the upper applications. These advances are relevant to millions of legacy applications since utilizing our mechanism does not require the existing applications to be re-programmed. The experimental results show that the root mean square errors (RMSE) before and after applying our mechanism are significantly reduced from 6.3 × 10(-1) to 5.3 × 10(-7), respectively.
... These automated processes are enabled in software, and thus vulnerable to the " garbage-in/garbage-out " constraint. IF functional Ontologies [64] Situation assessment Activity monitoring [65][66][67][68]Situation understanding [69][70][71]Natural language understanding and linguistics [72, 73] Decision-making [74][75][76][77][78][79]Level 3 Intent assessment [80][81][82]Level 4 Process refinement Context discovery [83] Context adaptation [79] Context learning [51, 84, 85] capabilities are developed to estimate some aspect of a real world that is of interest to some user. Those capabilities are bounded by a number of factors, such as the quality of the available data or information that can be used to form the estimates 2 as well as the complexity of the world being observed. ...
Chapter
Full-text available
This chapter attempts to cover two topics which themselves are complex and multidisciplinary: the concept of “Context” and the concept of “Information Fusion” , both of which have long histories of research publications. This chapter thus attempts to provide the reader concise introductions to these two topics by providing a review of an established framework for data and information fusion that derives from the well-known functional model of the fusion process called the Joint Directors of Laboratories or JDL model of fusion. The latter part of the chapter introduces two frameworks for how information fusion and contextual information can possibly be joined together that would allow for improved exploitation and inferencing in a variety of applications; these frameworks should be viewed as suggestions of notional processing concepts for these purposes. The chapter also provides numerous references for the reader to follow up and explore any of the ideas offered herein.
... For example, while describing height of a person, the observer might note it down as "about 6 feet" or "very tall." This ambiguity in the attribute values is captured and quantified using a process called uncertainty alignment [8], [9], in which the ambiguous attribute values are modeled using probabilistic or possibilistic distribution functions. The similarity score for such attributes is calculated as a function of the probability values obtained from the respective distributions. ...
Article
Full-text available
The information gathered by sources during counter insurgency (COIN) operations can be classified into two types. The data gathered by humans and recorded in the textual format such as field reports is referred to as soft data. Data gathered with the help of physical sensors such as video cameras, LIDAR, acoustic sensors, etc. is referred to as hard data. To process this information, various hard and soft processing techniques are used, which convert these data into relational graphs. Many times this hard and soft data contains duplicate references of the same entities, events and relationships, caused by multiple sources reporting on the same entity, event or relationship or due to successive reports on an entity, event or relationship. The role of data association is to identify these duplicate references across the different observations and merge them into fused (cumulative) evidence. This cumulative evidence will contain more information about the real world than offered by any single observation. We want the cumulative evidence to describe the real world as accurately as possible, so as to draw satisfactory conclusions on the state of the real world. This calls for development of an objective strategy for evaluating the performance of data association processes. In this paper, we describe in detail the testing and evaluation strategy developed for this task. This strategy was deployed for the evaluation of three different data association algorithms on the Sunni Criminal (SUN) Thread of the Synthetic Counterinsurgency (SYNCOIN) dataset. The SUN thread consists of 114 soft messages and 13 hard messages. This test and evaluation strategy can also help in selecting the best data association algorithm to deploy based on the different properties of the input dataset and processing time permissible.
... Examples of nonmilitary applications of data fusion include air traffic control [22], healthcare [23,24], speaker detection and tracking [25], mobile robot navigation [26], mobile robot localization [27], intelligent transportation systems [28], remote sensing [29,30], environment monitoring [31,32], and situational awareness [33]. For example, a Bayesian approach with pre-and postfiltering to handle data uncertainty and inconsistency in mobile robot local positioning is described in Ref. [27]. ...
... While the processing (fusion) of information from hard sensors is widely covered within the fusion literature for years [5], the fusion of information from soft sources received only recently quite more attention [6], [7], [8]. Combining both types of information has also attracted increasing interest in the domain of Hard and Soft information fusion within the last years: [9], [10], [11], [12], [13], [14], [15] propose new methods, algorithms, frameworks, architectures and uncertainty alignment approaches. ...
Conference Paper
Full-text available
Physical sensors (hard sources) and humans (soft sources) have complementary features in terms of perception, reasoning, memory. It is thus natural to combine their associated information for a wider coverage of the diversity of the available information and thus provide an enhanced situation awareness for the decision maker. While the fusion domain mainly considers (although not only) the processing and combination of information from hard sources, conciliating these two broad areas is gaining more and more interest in the domain of hard and soft fusion. In order to better understand the diversity and specificity of sources of information, we propose a functional model of a source of information, and a structured list of dimensions along which a source of information can be qualified. We illustrate some properties on a real data gathered from an experiment of light detection in a fog chamber involving both automatic and human detectors.
Article
With the development of ordnance technology, the survival and safety of individual combatants in high-tech warfare are under serious threat, and the Personal Protective Equipment (PPE), as an important guarantee to reduce casualties and maintain military combat effectiveness, is widely developed. This paper systematically reviewed various PPE based on individual combat through literature research and comprehensive discussion, and introduced in detail the latest application progress of PPE in terms of material and technology from three aspects: individual integrated protection system, traditional protection equipment, and intelligent protection equipment, respectively, and discussed in depth the functional improvement and optimization status brought by advanced technology for PPE, focusing on the achievements of individual equipment technology application. Finally, the problems and technical bottlenecks in the development of PPE were analyzed and summarized, and the development trend of PPE were pointed out. The results of the review will provide a forward-looking reference for the current development of individual PPE, and are important guidance for the design and technological innovation of advanced equipment based on the future technological battlefield.
Article
Full-text available
The diverse applications and widespread effects of cyberspace in most military and civilian applications have led to the rapid growth of data, information, knowledge, technology, methods, tools, and cyber systems. One of the important and strategic requirements in the cyber command and control cycle is the capability to extract, process, fusion and analyze data and information from various sources to achieve the desired situational awareness of the cyberspace and cyber operations environment. Considering the various aspects of this issue, the use of soft data, ie data and information from human resources, along with hard data, ie data and information of machine resources, can help in achieving more accurate and reliable recognition and decision. One of the main research topics in this area is designing a conceptual model and appropriate processing framework for analysis and inference based on data and information with the ability of linking various information from different sources with each other and modeling different types of uncertainty in hard and soft data and fusing them. This paper presents an ontology-based approach to the processing and fusion of hard and soft data in cyber command and control, in which various aspects of the issue has been addressed including ontology-based architecture and processing framework of information inference and fusion, the method of representation of uncertainty and reliability in hard and soft data and fusing these data, conversion of beliefs into probabilities to be able to make decisions on the hypotheses under consideration, designing an appropriate ontology model for cyber command and control purposes, and design and implementation of ontology-based logic of inference and fusion. The results of applying the proposed model in a typical scenario of cyber command and control show that it is operational in fusion of hard and soft cyber data. In addition to the ability of inference and fusion, one of the notable features of the proposed approach is its scalability and adaptability to new extensions in the tools and requirements of cyber command and control.
Conference Paper
This paper appeared in media: "AI research strengthens certainty in battlefield decision-making [https://www.army.mil/article/249169]" ---------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- The resurgence of AI in the recent decade dramatically changes the design of modern sensor data fusion systems, leading to new challenges, opportunities, and research directions. One of these challenges is the management of uncertainty. This paper develops a framework to reason about sources of uncertainty, develops representations of uncertainty, and investigates uncertainty mitigation strategies in modern intelligent data processing systems. Insights are developed into workflow composition that maximizes efficacy at accomplishing mission goals despite the sources of uncertainty, while leveraging a collaboration of humans, algorithms, and machine learning components.
Conference Paper
Mobile users can be recommended services or goods precisely according to their actual needs even in different contexts. Therefore, it is necessary to construct a framework integrating following functions: context identification, context reasoning, services or product recommendations and other tasks for the mobile terminal. In this paper, we firstly introduce mobile context awareness theory, and describe the composition of context-aware mobile systems. Secondly, we construct a framework of mobile context-aware recommendation system in line with the characteristics of mobile terminal devices and mobile context-aware data. Then, we build a nested key-value storage model and an up-to-date algorithm for mining mobile context-aware sequential pattern, in order to find both the user’s long-term behavior pattern and the new trend of his recent behavior, to predict user’s next behavior. Lastly, we discuss the difficulties and future development trend of mobile context-aware recommendation system.
Conference Paper
This contribution presents the application of Dempster-Shafer theory to the prediction of China’s stock market. To be specific, we predicted the most promising industry in the next month every trading day. This prediction can help investors to select stocks, but is rarely seen in previous literatures. Instead of predicting the fluctuation of the stock market from scratch all by ourselves, we fused ratings of 44 industries from China’s securities companies using Shafer’s evidence theory. Our preliminary experiment is a daily prediction since 2012-05-02 with ratings published 10 days before that day. Our predicted industries have an average rank of 19.85 in earnings, 11.8% better than random guessing (average rank is 22.5). The average rise rate of predicted industries in a month is 0.59%, 0.86% higher than overall (which is -0.274%), and nearly 0.7% higher than simple voting (which is -0.097%). Our predictions are posted on Weibo every day since 2014-04-28.
Article
Intelligence analysis is a domain characterized by a torrent of streaming data within which a very small portion contains useful knowledge or actionable intelligence. Intelligence analysts have to sift through the compiled data and weave through a complex web of convoluted connections in an attempt to illuminate information requirements (IR) and maintain situational awareness. Automated methodologies have eased the manual burden of this process to some extent. Data are naturally modeled in a graphical form representing the known people, places, events and the relationships between them. Graph matching algorithms in which an information requirement is formulated as a template graph or situation of interest to be found in the observed data graph have been successfully employed in intelligence analysis processes. Absent from these past contributions is the recognition that partial information requirements, such as indicators and warnings, are not mutually exclusive to a specific IR, and an understanding of the characteristics of the underlying data can lead to significant performance benefits. The knowledge of overlapping template sections forms the motivation for precedence tree guided search and AND/OR templates. Through the recognition of the overlapping sections, a single AND/OR template can be created to answer many information requirements. This paper presents a novel algorithm for the intelligent traversal of an AND/OR template, providing increased algorithmic efficiency over the execution of multiple sequential graph matching instances. This paper focuses on development of an algorithm for intelligent AND/OR template traversal with computational results illustrating the effectiveness of the developed methods. The results indicate a significant improvement in runtime (with a speedup over 5 in some cases) while maintaining a good solution quality (within 2% of multiple AND path graph matching executions) in AND/OR and precedence tree guided graph matching.
Article
This survey aims to provide a comprehensive status of recent and current research on context-based Information Fusion (IF) systems, tracing back the roots of the original thinking behind the development of the concept of “context”. It shows how its fortune in the distributed computing world eventually permeated in the world of IF, discussing the current strategies and techniques, and hinting possible future trends. IF processes can represent context at different levels (structural and physical constraints of the scenario, a priori known operational rules between entities and environment, dynamic relationships modelled to interpret the system output, etc.). In addition to the survey, several novel context exploitation dynamics and architectural aspects peculiar to the fusion domain are presented and discussed.
Article
Full-text available
The perception of distance to real and virtual objects using two methods of distance estimation (verbal estimation and manual replication) along a 110 foot hallway was tested. Results suggest that verbal estimates of distance may not accurately reflect perceived distances. Replication procedure significantly improves the estimation of the previously viewed object distance. Furthermore, the effects of distance judgment method were greater than were the effects of environment type. The magnitude of the distance judgment error was considerably larger for the estimation condition in the real environment than it was for the replication condition in the augmented environment. These results lend further support to the notion that verbal estimates of distance do not accurately represent perceived distance. Unless the task being performed specifically requires a numerical estimate of distance, it is recommended that methods similar to our distance replication method be used to accurately determine perceived distance.
Conference Paper
Full-text available
Soft information is information contained in natural language messages written by human informants or human intelligence gatherers. Tractor is a system that automatically processes natural language messages and represents the information extracted from them as propositional graphs. Propositional graphs have several benefits as a knowledge representation formalism for information fusion: n-ary relations may be represented as simply as binary relations; meta-information and pedigree may be represented in the same format as object-level information; they are amenable to graph matching techniques and fusion with information from other sources; they may be used by reasoning systems to draw inferences from explicitly conveyed information and relevant background information. The propositional graphs produced by Tractor are based on the FrameNet system of deep lexical semantics. A method of producing propositional graphs is proposed using dependency parse information and rules written in the SNePS knowledge representation and reasoning system.
Article
Full-text available
Intelligence analysis is a prototypical example of a situation that can be described by the phrase “coping with complexity,” given a variety of complicating factors including (but certainly not limited to) massive amounts of data, inherent uncertainty in the data, dynamism in the underlying world, and the risks associated with the conclusions drawn from the data. Given this complex nature of intelligence analysis, a wide variety of advanced visualization tools are continually being developed in hopes of designing support for the analysis process. In addition, a number of different cognitive analysis activities are ongoing in hopes of deriving requirements for future support tools. The goal of this panel is to present a sample from both areas in hopes of providing an integration of efforts and thus more effective support tools for intelligence analysts. We have four speakers presenting an analytic perspective, one from a tool development perspective, and one from a support function (the medium between analysis and design) perspective. This should provide an interesting set of complementary discussions into this topic area.
Article
Full-text available
Disputes over the size of mass protests have become routine. Organizers want to highlight widespread commitment to their cause and distrust conservative estimates by the media and police. More accurate counts probably will not end these disagreements, but will provide better evidence amidst the partisan squabbling.
Article
Full-text available
This paper describes an experiment to evaluate a procedure for measuring distance perception in immersive VEs. Forty-eight subjects viewed a VE with a Head Mounted Display (HMD), a Binocular Omni-Oriented Monitor (BOOM), or a computer monitor. Subjects estimated the distance to a figure of known height that was initially 40 ft away. As the figure moved forward, subjects indicated when the figure was perceived to be 30, 20, 10, 5, and 2.5 ft away. A separate group of 36 subjects performed the task in a real-world setting roughly comparable to the VE. VE distance estimation was highly variable across subjects. For distance perception involving a moving figure, in the VE conditions most subjects called out before the figure had closed to the specified distances. Distance estimation was least accurate with the monitor. In the real world, most subjects called out after the figure had closed to or passed the specified distances. Ways to improve the procedure are discussed.
Conference Paper
Full-text available
Historically, data fusion has focused on processing hard or physical sensor data while soft or human observed data has been neglected within fusion processes. This human observed data has much to offer towards obtaining comprehensive situational awareness, particularly in a domain such as intelligence analysis where subtle connections and interactions are difficult to observe with physical sensors. This paper describes the processing architecture designed and implemented for the fusion of hard and soft data in the multi-university research initiative on network-based hard and soft information fusion. The processing elements designed to successfully fuse and reason over the hard and soft data include the natural language processing elements to form propositional graphs from linguistic observations, conversion of the propositional graphs to attributed graphical form, alignment and tagging of the uncertainties extant in the human observations, conversion of hard data tracks to a graphical format, association of entities and relations in observational hard and soft data graphs and the matching of situations of interest to the cumulative data or evidential graph. To illustrate these processing elements within the integrated processing architecture a small synthetic data set entitled the bomber buster scenario is utilized, presenting examples of each processing element along the processing flow. The value of fusing hard and soft information is illustrated by demonstrating that individually, neither hard nor soft information could provide the situation estimate.
Article
Full-text available
The purpose of this project was to conduct applied research with exemplary technology to support post-graduate instruction in intelligence analysis. The first phase of research used Cognitive Task Analysis (CTA) to understand the nature of subject matter expertise for this domain, as well as leverage points for technology support. Results from the CTA and advice from intelligence analysis instructors at the Naval Postgraduate School lead us to focus on the development of a collaborative computer tool (CACHE) to support a method called the Analysis of Competing Hypotheses (ACH). We first evaluated a non-collaborative version of an ACH tool in an NPS intelligence classroom setting, followed by an evaluation of the collaborative tool, CACHE at NPS. These evaluations, along with similar studies conducted in coordination with NIST and MITRE, suggested that ACH and CACHE can support intelligence activities and mitigate confirmation bias. However, collaborative analysis has a number of trade-offs: it incurs overhead costs, and can mitigate or exacerbate confirmation bias, depending on the mixture of predisposing biases of collaborators.
Article
Full-text available
It is a rare season when the intelligence story in the news concerns intelligence analysis, not secret operations abroad. The United States is having such a season as it debates whether intelligence failed in the run-up to both September 11 and the second Iraq war, and so Rob Johnston's wonderful book is perfectly timed to provide the back-story to those headlines. The CIA's Center for the Study of Intelligence is to be commended for having the good sense to find Johnston and the courage to support his work, even though his conclusions are not what many in the world of intelligence analysis would like to hear. He reaches those conclusions through the careful procedures of an anthropologist -- conducting literally hundreds of interviews and observing and participating in dozens of work groups in intelligence analysis -- and so they cannot easily be dismissed as mere opinion, still less as the bitter mutterings of those who have lost out in the bureaucratic wars. His findings constitute not just a strong indictment of the way American intelligence performs analysis, but also, and happily, a guide for how to do better. Johnston finds no baseline standard analytic method. Instead, the most common practice is to conduct limited brainstorming on the basis of previous analysis, thus producing a bias toward confirming earlier views. The validating of data is questionable -- for instance, the Directorate of Operation's (DO) "cleaning" of spy reports doesn't permit testing of their validity -- reinforcing the tendency to look for data that confirms, not refutes, prevailing hypotheses. The process is risk averse, with considerable managerial conservatism. There is much more emphasis on avoiding error than on imagining surprises. The analytic process is driven by current intelligence, especially the CIA's crown jewel analytic product, the President's Daily Brief (PDB), which might be caricatured as "CNN plus secrets."
Article
Full-text available
Two experiments investigated the importance of spatial and surface cues in the age-processing of unfamiliar faces aged between one and 80 years. Three manipulations known to affect face recognition were used, individually and in various combinations: inversion, negation, and blurring. Faces were presented either in whole or in part. Age-estimation performance was largely unaffected by most of these manipulations; age-processing appears to be a highly robust process, due to the numerous cues available. Experiment 1 showed that, in contrast to face recognition, age-perception appears to be substantially unimpaired by inversion or negation. Experiment 2 suggests that age-estimates can be made on the basis of either surface information (the 2D disposition of the internal facial features, together with texture information) or shape information (head-shape plus feature configuration, as long as shape-from-shading information is present).
Article
Full-text available
This study addressed the question of how people remember the time of past events. Stimuli were 10 news events that had occurred from 6 months to 20 years before the study. In contrast to previous studies of memory for time, subjects were asked to provide estimates of the stimulus events on multiple time scales, including year, month, day of the month, day of the week, and hour. If judgments are based on direct information about the age of the memory, accuracy should decrease monotonically as one moves to finer scales. Alternatively, if subjects reconstruct the time from fragmentary information associated with the event, one would expect that estimates on finer time scales would often exceed grosser scales in accuracy. Results for accuracy, confidence, and number of recall cues supported the latter position. In addition, subjects reported a variety of types of recall cues, the most common being memory for personal experiences or events that were contiguous with the news event.
Conference Paper
Full-text available
Fusion of observational data acquired by human observers and couched in linguistic form is a modern-day challenge for the fusion community. This paper describes a basic research effort examining various strategies for associating and exploiting such data for intelligence analysis purposes. An overall approach is described that involves Latent Semantic Analysis, Inexact Graph Matching, formal ontology development, and Social Network Analyses. Not all the methods have yet been employed but the exploitation of the developed ontology and graphical techniques have been implemented in a working prototype and preliminary results have shown promise. Planned future research will complete the implementation of the methods described herein and add yet further enhancements.
Article
Full-text available
Our goal in this paper is to develop a practical framework for obtaining a uniform sample of users in an online social network (OSN) by crawling its social graph. Such a sample allows to estimate any user property and some topological properties as well. To this end, first, we consider and compare several candidate crawling techniques. Two approaches that can produce approximately uniform samples are the Metropolis-Hasting random walk (MHRW) and a re-weighted random walk (RWRW). Both have pros and cons, which we demonstrate through a comparison to each other as well as to the "ground truth." In contrast, using Breadth-First-Search (BFS) or an unadjusted Random Walk (RW) leads to substantially biased results. Second, and in addition to offline performance assessment, we introduce online formal convergence diagnostics to assess sample quality during the data collection process. We show how these diagnostics can be used to effectively determine when a random walk sample is of adequate size and quality. Third, as a case study, we apply the above methods to Facebook and we collect the first, to the best of our knowledge, representative sample of Facebook users. We make it publicly available and employ it to characterize several key properties of Facebook.
Conference Paper
Full-text available
This paper describes a methodology for incorporating human observations into a hard+soft information fusion process for counterinsurgency intelligence analysis. The goal of incorporating human observations into the information fusion process is important as it extends the ability of the fusion algorithms to associate and merge disparate pieces of information by allowing for information collected from soft data sources (e.g., human observations) to be included in the process along with information collected from hard data sources (e.g., radar sensors). This goal is accomplished through the employment of fuzzy membership functions used in similarity scoring, for data association and situation assessment. These membership functions are based on situationally qualified error characteristics. Error characteristics represent the key to this process by allowing for accurate uncertainty alignment based on the known and/or unknown state of context dependent variables that have been empirically determined to influence the accuracy of human estimation for a given category- in this case human age estimation.
Conference Paper
Full-text available
In conventional warfare as well as counter-insurgency (COIN) operations, the understanding of the situation is extremely vital to assure a sense of security. Intelligence in COIN is about people, and deployed units in the field are the best sources of intelligence. Past and present intelligence data is analyzed to look for changes in the insurgents' approach or tactics. To do this, graphical methods have proven to be effective. In recent work, have developed an inexact subgraph matching algorithm as a variation of the subgraph isomorphism approach for situation assessment. This paper enhances this procedure to represent inaccurate observations or data estimates, and inaccurate structural representations of a state of interest, thus accounting for the uncertainties. Various probabilistic and possibilistic uncertainty representations, transformations between representations and methods for establishing similarities between representations have been reviewed. This comprehensible approach will give pragmatic estimates providing rigor and sound understanding during situation assessment.
Article
Full-text available
This paper presents a theoretical model of situation awareness based on its role in dynamic human decision making in a variety of domains. Situation awareness is presented as a predominant concern in system operation, based on a descriptive view of decision making. The relationship between situation awareness and numerous individual and environmental factors is explored. Among these factors, attention and working memory are presented as critical factors limiting operators from acquiring and interpreting information from the environment to form situation awareness, and mental models and goal-directed behavior are hypothesized as important mechanisms for overcoming these limits. The impact of design features, workload, stress, system complexity, and automation on operator situation awareness is addressed, and a taxonomy of errors in situation awareness is introduced, based on the model presented. The model is used to generate design implications for enhancing operator situation awareness and future directions for situation awareness research.
Conference Paper
Full-text available
Situation awareness (SA) is an important component of pilot/system performance in all types of aircraft. It is the role of the human factors engineer to develop aircraft cockpits which will enhance SA. Research in the area of situation awareness is is vitally needed if system designers are to meet the challenge of providing cockpits which enhance SA. This paper presents a discussion of the SA construct, important considerations facing designers of aircraft systems, and current research in the area of SA measurement.
Book
The 2005 BISC International Special Event-BISCSE’05 " FORGING THE FRONTIERS" was held in the University of California, Berkeley, “WHERE FUZZY LOGIC BEGAN, from November 3 – 6, 2005. The successful applications of fuzzy logic and it’s rapid growth suggest that the impact of fuzzy logic will be felt increasingly in coming years. Fuzzy logic is likely to play an especially important role in science and engineering, but eventually its influence may extend much farther. In many ways, fuzzy logic represents a significant paradigm shift in the aims of computing - a shift which reflects the fact that the human mind, unlike present day computers, possesses a remarkable ability to store and process information which is pervasively imprecise, uncertain and lacking in categoricity. The chapters of the book are evolved from presentations made by selected participants at the meeting and organized in two books. The papers include reports from the different front of soft computing in various industries and address the problems of different fields of research in fuzzy logic, fuzzy set and soft computing. The book provides a collection of forty four (44) articles in two volumes.
Book
The Psychology of Intelligence Analysis has been required reading for intelligence officers studying the art and science of intelligence analysis for decades. Richards Heuer, Jr. discusses in the book how fundamental limitations in human mental processes can prompt people to jump to conclusions and employ other simplifying strategies that lead to predictably faulty judgments known as cognitive biases. These analytic mindsets cannot be avoided, but they can be overcome through the application of more structured and rigorous analytic techniques including the Analysis of Competing Hypotheses.
The primary goal of this effort was to understand the problems faced by military intelligence analysis personnel as well as how, and to what degree, the identification of these problems could guide the development of computational support systems. To develop this understanding, we performed a literature review, knowledge elicitation interviews and a cognitive task analysis (CTA) in the domain of Army Intelligence Analysis at the Brigade Combat Team. This effort consisted of identifying: (1) the major functions or cognitive tasks entailed in Army Intelligence Analysis; and (2) the complexities in the domain that pose challenges to performance of these cognitive tasks. Identifying the cognitive tasks and the challenges faced in performing those tasks provided a basis for determining opportunities for more effective support of human information processing and decision-making. In this paper, we document selected results of this analysis effort.
Article
Graph association is the problem of merging many graphs that collectively describe a set of possibly repetitive entities and relationships into a single graph that contains unique entities and relationships. As a form of data association, graph association can be used to identify when two sensors are observing the same object so information from both sensors can be combined and analyzed in a meaningful and consistent way. Graph association between two graphs is related to the problem of graph matching, and between multiple graphs it is related to the common labeling of a graph set (also known as multiple graph matching) problem. This article contribution is to formulate graph association as a binary linear program and introduce a heuristic for solving multiple graph association using a Lagrangian relaxation approach to address issues with between-graph transitivity requirements. The algorithms are tested on a representative dataset. The developed model formulation was found to accurately solve the graph association problem. Furthermore, the Lagrangian heuristic was found to solve the developed model within 3% of optimal on many problem instances, and found better solutions to large problems than is possible by directly using CPLEX. © 2013 Wiley Periodicals, Inc. Naval Research Logistics, 2013
Article
In intelligence analysis a situation of interest is commonly obscured by the more voluminous amount of unimportant data. This data can be broadly divided into two categories, hard or physical sensor data and soft or human observed data. Soft intelligence data is collected by humans through human interaction, or human intelligence (HUMINT). The value and difficulty in manual processing of these observations due to the volume of available data and cognitive limitations of intelligence analysts necessitate an information fusion approach toward their understanding. The data representation utilized in this work is an attributed graphical format. The uncertainties, size and complexity of the connections within this graph make accurate assessments difficult for the intelligence analyst. While this graphical form is easier to consider for an intelligence analyst than disconnected multi-source human and sensor reports, manual traversal for the purpose of obtaining situation awareness and accurately answering priority information requests (PIRs) is still infeasible. To overcome this difficulty an automated stochastic graph matching approach is developed. This approach consists of three main processes: uncertainty alignment, graph matching result initialization and graph matching result maintenance. Uncertainty alignment associates with raw incoming observations a bias adjusted uncertainty representation representing the true value containing spread of the observation. The graph matching initialization step provides template graph to data graph matches for a newly initialized situation of interest (template graph). Finally, the graph matching result maintenance algorithm continuously updates graph matching results as incoming observations augment the cumulative data graph. Throughout these processes the uncertainties present in the original observations and the template to data graph matches are preserved, ultimately providing an indication of the uncertainties present in the current situation assessment. In addition to providing the technical details of this approach, this paper also provides an extensive numerical testing section which indicates a significant performance improvement of the proposed algorithm over a leading commercial solver.
Article
As a response to Air Combat Command (ACC/A2RT), the 711HPW/RHAS reviewed published literature on critical thinking skills and training to enhance skills as they relate to improving performance of intelligence analysts. While there are many critical training curriculums available in the intelligence community, current literature shows a lack of empirical evidence correlating critical thinking and intelligence analysis. This report suggests some considerations for an effective critical thinking curriculum as it relates to intelligence analysis.
Conference Paper
Multisensor data fusion is a key enabling technology in which information from a number of sources is integrated to form a unified picture [1]. This concept has been applied to numerous fields and new applications are being explored constantly. Even though most multisensor data fusion applications have been developed relatively recently, the notion of data fusion has always been around. In fact, all of us employ multisensor data fusion principles in our daily lives. The human brain is an excellent example of an operational fusion system that performs extremely well. It integrates sensory information, namely sight, sound, smell, taste and touch data and makes inferences regarding the problem at hand. It has been a natural desire of researchers in different disciplines of science and engineering to emulate this information fusion ability of the human brain. The idea is that fusion of complementary information available from different sensors will yield more accurate results for information processing problems. Significant advances in this important field have been made but perfect emulation of the human brain remains an elusive goal.
Article
" . . . Ss who received the pretraining were superior to control Ss in both constant and variable error. Absolute estimation was improved even though Ss were not tested in the same field where they were trained, the targets were unfamiliar, and the distances varied. It was proposed that S learned a scale relating responses, in yards, to gradients of stimulation deriving from the ground surface." When S was asked to make relative judgments of distance pretraining did not lower DL's. (PsycINFO Database Record (c) 2006 APA, all rights reserved).
Article
Using targets which provided no cues, an experimental group judged the distance, in yards, to 18 targets, then made 90 corrected judgments, and finally repeated the first series. A control group made the two estimations without the intervening training. It was found that improvement in absolute judgment was one result of the training series along with other effects upon constant and variable errors. "The development of a conceptual scale of distance in a psychophysical relationship with stimulation provided by a receding stretch of ground was hypothesized."
Article
Subjects were asked to date 28 ‘recent’ public events that had occurred in the previous 7 years, and 28 ‘historical’ events that occurred between 1765 and 1947. For both kinds of event, the age of older events was underestimated and that of more recent events overestimated, a result agreeing with previous research. Whether the events were well or poorly known, as rated by a separate sample of subjects, affected the dating error of historical but not recent events. The results suggest that both recent and historical events are dated by a rather abstract, constructive process, rather than by cues relating to the age of the memory or the time of its formation.
Article
This study examined the possibility to improve accuracy of age estimates through training. Thirty-four participants were divided into an experimental and a control group. The sessions included a pre-test before training, six feedback or no-feedback training tests and a post-test after training. The experimental group performed the feedback tests and the control group the no-feedback tests. Training was found to improve age estimation accuracy, particularly estimations of old stimuli, and training with feedback seemed to be superior to training without feedback. No difference was found between the groups at pre-test, but at post-test the experimental group exhibited greater accuracy in age estimation. Moreover, the experimental group increased its accuracy between the pre- and post-tests. Copyright © 2006 John Wiley & Sons, Ltd.
Article
Accurate age estimation is important in a variety of settings, particularly those in which age is a condition for access to a product such as alcohol or tobacco. The current paper reviews data on the estimation of age, focusing on changes that occur to the ageing face, methodology and findings. The accumulated data suggest that age estimation of unfamiliar faces can be quite accurate, is sensitive to cues at the local and global level and may vary on the basis of group membership. Potential directions for future research are highlighted and findings are discussed with respect to their implications for policies regarding the solicitation of age identification. Copyright © 2008 John Wiley & Sons, Ltd.
Article
Although the ability of subjects to judge the duration of a temporal interval has been investigated using a variety of paradigms, most of the studies have used very short intervals ranging from a few hundred milliseconds to several seconds. Therefore, the purpose of the present study was to examine the psychophysical relationship between judged and actual durations of longer intervals ranging from months to years. In Experiment 1, college students were asked to estimate the ages of 12 major news events which had occurred during the previous 5 years. Analysis of median estimates of each item showed that subjects underestimated the ages of the oldest items and overestimated the ages of the most recent items. To determine whether this regression effect was due to a response bias or to the context effect produced by other items on the questionnaire, Experiments 2 and 3 systematically varied the range of allowable response alternatives and the age of the background items that formed the context in which the experimental items were judged. Although the range of allowable response alternatives had a significant effect upon the estimates, no evidence was obtained for an item context effect. The data are interpreted in terms of a memory-based hypothesis which suggests that the subjects were attempting to recall the time of occurrence of the items and then deriving elapsed time rather than directly judging duration.
Article
In the past four years, we have worked with several research programs that were developing intelligent software for use by intelligence analysts. Our involvement in these programs was to develop the metrics and methodologies for assessing the impact on users; in this case, on intelligence analysts. In particular, we focused on metrics to evaluate how much the intelligent systems contribute to the users’ tasks and what the cost is to the user in terms of workload and process deviations. In this paper, we describe the approach used. We started with two types of preliminary investigations – first, collecting and analyzing data from analysts working in an instrumented environment for a period of 2 years, and second, developing and conducting formative evaluations of research software. The long-term studies informed our ideas about the processes that analysts use and provided potential metrics in an environment without intelligent software tools. The formative evaluations helped us to define sets of application-specific metrics. Finally, we conducted assessments during and after technology insertions. We describe the metrics and methodologies used in each of these activities, along with the lessons learned.
Article
This research explores the problem of how people determine the time of public events, such as the attempted assassination of Ronald Reagan or the Three-Mile Island accident. According to what here is called the accessibility principle, the subjective dates of these events depend in part on the amount that can be recalled about them: The more known, the more recent the event will seem. Experiments 1 and 2 demonstrate this effect when subjects estimate explicit dates for important news stories of the 1970s and 1980s. The same effect appears in Experiment 3 for subjects who rate the recency of less known events drawn from a single week. Accessibility also contributes to the amount of time needed to compare the subjective date of an event (e.g., the Jonestown suicides) to an explicitly presented date (e.g., November 1979), as shown in Experiment 4. The accessibility principle for time estimation can be conceived as one of a related group of retrieval-based inferences that plays a part in judgments of frequency and probability and judgments about the falsity of a putative fact.
Article
Records of everyday autobiographical events were gathered from a small group of adults during a 4-month period. This was followed by five memory tests extending over years. Recognition memory, temporal ordering, and dating accuracy declined as the events tested became more remote. Recognition accuracy on original items was high over the entire study; whereas the false recognition of nonevent, foil items increased after a 1- to 3-month delay. Confidence ratings of recognition accuracy remained consistently high over all tests, even though recognition accuracy deteriorated. Additional analyses of foil items indicated that false recognitions of nonevents as one's own memories were related positively to the semantic similarity between foils and the original records from which they were constructed. Taken together, the data support the hypothesis that the same autobiographical schemata account for the correct recognition of actual events, the false recognition of certain nonevents as one's own memories, the correct rejection of other nonevents, and an overconfidence in the “facts” of one's life.
Article
The intent of this paper is to show enhancements in Levels 2 and 3 fusion capabilities through a new class of models and algorithms in graph matching. The problem today is not often lack of data, but instead, lack of information and data overload. Graph matching algorithms help us solve this problem by identifying meaningful patterns in voluminous amounts of data to provide information. In this paper we investigate a classical graph matching technique for subgraph isomorphism. A complete implementation of a heuristic approach (since the problem under consideration is NP-Hard) using an inexact isomorphism technique has been used. The heuristic approach is called Truncated Search Tree algorithm (TruST), where the state space of the problem is constrained using breadth and depth control parameters. The breadth and depth control parameters are then studied using design of experiment based inferential statistics. Finally, a software implementation of the procedure has been completed.
Article
The importance of a Web page is an inherently subjective matter, which depends on the readers interests, knowledge and attitudes. But there is still much that can be said objectively about the relative importance of Web pages. This paper describes PageRank, a mathod for rating Web pages objectively and mechanically, effectively measuring the human interest and attention devoted to them. We compare PageRank to an idealized random Web surfer. We show how to efficiently compute PageRank for large numbers of pages. And, we show how to apply PageRank to search and to user navigation.
Article
In this paper a fuzzy distance measure between two generalized fuzzy numbers is developed. The metric properties of this distance measure are also studied. The new distance measure is compared with the other fuzzy distance measures proposed by Voxman [W. Voxman, Some remarks on distances between fuzzy numbers, Fuzzy Sets and Systems 100 (1998) 353–365] and Chakraborty and Chakraborty [C. Chakraborty, D. Chakraborty, A theoretical development on fuzzy distance measure for fuzzy numbers, Mathematical and Computer Modelling 43 (2006) 254–261] and turned out to be more reasonable. A new similarity measure is also developed with the help of the fuzzy distance measure. Examples are given to compare this similarity measure with the other similarity measure previously proposed. A decision making scheme is proposed using this similarity measure and this scheme is found to be more acceptable than the existing methods due to the fact that it considers the degrees of confidence of the experts’ opinion.
Article
The ability to accurately estimate distance is an essential component of navigating large-scale spaces. Although the factors that influence distance estimation have been a topic of research in real-world environments for decades and are well known, research on distance estimation in virtual environments (VEs) has only just begun. Initial investigations of distance estimation in VEs suggest that observers are less accurate in estimating distance in VEs than in the real world (Lampton et al., 1995). Factors influencing distance estimates may be divided into those affecting perceived distance (visual cues only) and those affecting traversed distance to include visual, cognitive, and proprioceptive cues. To assess the contribution of the various distance cues in VEs, two experiments were conducted. The first required a static observer to estimate the distance to a cylinder placed at various points along a 130-foot hallway. This experiment examined the effects of floor texture, floor pattern, and object size on distance estimates in a VE. The second experiment required a moving observer to estimate route segment distances and total route distances along four routes, each totaling 1210 feet. This experiment assessed the effects of movement method, movement speed, compensatory cues, and wall texture density. Results indicate that observers underestimate distances both in VEs and in the real world, but the underestimates are more extreme in VEs. Texture did not reliably affect the distance estimates, providing no compensation for the gross underestimates of distance in VE. Traversing a distance improves the ability to estimate that distance, but more natural means of moving via a treadmill do not necessarily improve distance estimates over traditional methods of moving in VE (e.g., using a joystick). The addition of compensatory cues (tone every 10 feet traversed on alternate route segments) improves VE distance estimation to almost perfect performance.
Article
Although distance estimation has been extensively studied in the laboratory, our ability to judge large distances in the field is not well researched. We challenge the notion that large distances are uniformly underestimated. We presented different targets to observers at distances ranging from 25 to 500 m to obtain egocentric distance judgments in natural environments. Three experiments showed that observers tend to underestimate distances below 75 m in a large open field, whereas they overestimate farther distances. Both the eye height of the observer and the size of the target also influenced distance estimation. We conclude that the notion of a uniform vista space has to be reconceived.