Pub Date : 2022-08-06DOI: 10.1177/15553434221116254
Kuan-Ting Chen, H. Chen, Ann Bisantz, Su Shen, Ercan Sahin
There will be circumstances where partial or conditionally automated vehicles fail to drive safely and require human interventions. Within the human factors community, the taxonomies surrounding control transitions have primarily focused on characterizing the stages and sequences of the transition between the automated driving system (ADS) and the human driver. Recognizing the variance in operational design domains (ODDs) across vehicles equipped with ADS and how variable the takeover situations may be, we describe a simple taxonomy of takeover situations to aid the identification and discussions of takeover scenarios in future takeover studies. By considering the ODD structure and the human information processing stages, we constructed a fault tree analysis (FTA) aimed to identify potential failure sources that would prevent successful control transitions. The FTA was applied in analyzing two real-world accidents involving ADS failures, illustrating how this approach can help identify areas for improvements in the system, interface, or training design to support drivers in level 2 and level 3 automated driving.
{"title":"Where Failures May Occur in Automated Driving: A Fault Tree Analysis Approach","authors":"Kuan-Ting Chen, H. Chen, Ann Bisantz, Su Shen, Ercan Sahin","doi":"10.1177/15553434221116254","DOIUrl":"https://doi.org/10.1177/15553434221116254","url":null,"abstract":"There will be circumstances where partial or conditionally automated vehicles fail to drive safely and require human interventions. Within the human factors community, the taxonomies surrounding control transitions have primarily focused on characterizing the stages and sequences of the transition between the automated driving system (ADS) and the human driver. Recognizing the variance in operational design domains (ODDs) across vehicles equipped with ADS and how variable the takeover situations may be, we describe a simple taxonomy of takeover situations to aid the identification and discussions of takeover scenarios in future takeover studies. By considering the ODD structure and the human information processing stages, we constructed a fault tree analysis (FTA) aimed to identify potential failure sources that would prevent successful control transitions. The FTA was applied in analyzing two real-world accidents involving ADS failures, illustrating how this approach can help identify areas for improvements in the system, interface, or training design to support drivers in level 2 and level 3 automated driving.","PeriodicalId":46342,"journal":{"name":"Journal of Cognitive Engineering and Decision Making","volume":"17 1","pages":"147 - 165"},"PeriodicalIF":2.0,"publicationDate":"2022-08-06","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"42865155","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2022-08-02DOI: 10.1177/15553434221113964
Claire Textor, Rui Zhang, Jeremy Lopez, Beau G. Schelble, Nathan J. Mcneese, Guo Freeman, R. Pak, Chad C. Tossell, E. D. de Visser
Advancements and implementations of autonomous systems coincide with an increased concern for the ethical implications resulting from their use. This is increasingly relevant as autonomy fulfills teammate roles in contexts that demand ethical considerations. As AI teammates (ATs) enter these roles, research is needed to explore how an AT’s ethics influences human trust. This current research presents two studies which explore how an AT’s ethical or unethical behavior impacts trust in that teammate. In Study 1, participants responded to scenarios of an AT recommending actions which violated or abided by a set of ethical principles. The results suggest that ethicality perceptions and trust are influenced by ethical violations, but only ethicality depends on the type of ethical violation. Participants in Study 2 completed a focus group interview after performing a team task with a simulated AT that committed ethical violations and attempted to repair trust (apology or denial). The focus group responses suggest that ethical violations worsened perceptions of the AT and decreased trust, but it could still be trusted to perform tasks. The AT’s apologies and denials did not repair damaged trust. The studies’ findings suggest a nuanced relationship between trust and ethics and a need for further investigation into trust repair strategies following ethical violations.
{"title":"Exploring the Relationship Between Ethics and Trust in Human–Artificial Intelligence Teaming: A Mixed Methods Approach","authors":"Claire Textor, Rui Zhang, Jeremy Lopez, Beau G. Schelble, Nathan J. Mcneese, Guo Freeman, R. Pak, Chad C. Tossell, E. D. de Visser","doi":"10.1177/15553434221113964","DOIUrl":"https://doi.org/10.1177/15553434221113964","url":null,"abstract":"Advancements and implementations of autonomous systems coincide with an increased concern for the ethical implications resulting from their use. This is increasingly relevant as autonomy fulfills teammate roles in contexts that demand ethical considerations. As AI teammates (ATs) enter these roles, research is needed to explore how an AT’s ethics influences human trust. This current research presents two studies which explore how an AT’s ethical or unethical behavior impacts trust in that teammate. In Study 1, participants responded to scenarios of an AT recommending actions which violated or abided by a set of ethical principles. The results suggest that ethicality perceptions and trust are influenced by ethical violations, but only ethicality depends on the type of ethical violation. Participants in Study 2 completed a focus group interview after performing a team task with a simulated AT that committed ethical violations and attempted to repair trust (apology or denial). The focus group responses suggest that ethical violations worsened perceptions of the AT and decreased trust, but it could still be trusted to perform tasks. The AT’s apologies and denials did not repair damaged trust. The studies’ findings suggest a nuanced relationship between trust and ethics and a need for further investigation into trust repair strategies following ethical violations.","PeriodicalId":46342,"journal":{"name":"Journal of Cognitive Engineering and Decision Making","volume":"16 1","pages":"252 - 281"},"PeriodicalIF":2.0,"publicationDate":"2022-08-02","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"46141456","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2022-06-12DOI: 10.1177/15553434221108002
G. Tokadlı, M. Dorneich
What makes an autonomous system a teammate? The paper presents an evaluation of factors that can encourage a human perceive an autonomous system as a teammate rather than a tool. Increased perception of teammate-likeness more closely matches the human’s expectations of a teammate’s behavior, benefiting coordination and cooperation. Previous work with commercial pilots suggested that autonomous systems should provide visible cues of actions situated in the work environment. These results motivated the present study to investigate the impact of feedback modality on the teammate-likeness of an autonomous system under low (sequential events) and high (concurrent events) task loads. A Cognitive Assistant (CA) was developed as an autonomous teammate to support a (simulated) Mars mission. With centralized feedback, an autonomous teammate provided verbal and written information on a dedicated display. With distributed feedback, the autonomous teammate provided visible cues of actions in the environment in addition to centralized feedback. Perception of teammate-likeness increased with distributed feedback due to increased awareness of the CA’s actions, especially under low task load. In high task load, teamwork performance was higher with distributed feedback when compared to centralized feedback, where in low task load there was no difference in teamwork performance between feedback modalities.
{"title":"Autonomy as a Teammate: Evaluation of Teammate-Likeness","authors":"G. Tokadlı, M. Dorneich","doi":"10.1177/15553434221108002","DOIUrl":"https://doi.org/10.1177/15553434221108002","url":null,"abstract":"What makes an autonomous system a teammate? The paper presents an evaluation of factors that can encourage a human perceive an autonomous system as a teammate rather than a tool. Increased perception of teammate-likeness more closely matches the human’s expectations of a teammate’s behavior, benefiting coordination and cooperation. Previous work with commercial pilots suggested that autonomous systems should provide visible cues of actions situated in the work environment. These results motivated the present study to investigate the impact of feedback modality on the teammate-likeness of an autonomous system under low (sequential events) and high (concurrent events) task loads. A Cognitive Assistant (CA) was developed as an autonomous teammate to support a (simulated) Mars mission. With centralized feedback, an autonomous teammate provided verbal and written information on a dedicated display. With distributed feedback, the autonomous teammate provided visible cues of actions in the environment in addition to centralized feedback. Perception of teammate-likeness increased with distributed feedback due to increased awareness of the CA’s actions, especially under low task load. In high task load, teamwork performance was higher with distributed feedback when compared to centralized feedback, where in low task load there was no difference in teamwork performance between feedback modalities.","PeriodicalId":46342,"journal":{"name":"Journal of Cognitive Engineering and Decision Making","volume":"16 1","pages":"282 - 300"},"PeriodicalIF":2.0,"publicationDate":"2022-06-12","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"43677892","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2022-06-03DOI: 10.1177/15553434221104615
Ksenia Appelganc, Tobias Rieger, Eileen Roesler, D. Manzey
Tasks classically performed by human–human teams in today’s workplaces are increasingly given to human–technology teams instead. The role of technology is not only played by classic decision support systems (DSSs) but more and more by artificial intelligence (AI). Reliability is a key factor influencing trust in technology. Therefore, we investigated the reliability participants require in order to perceive the support agents (human, AI, and DSS) as “highly reliable.” We then examined how trust differed between these highly reliable agents. Whilst there is a range of research identifying trust as an important determinant in human–DSS interaction, the question is whether these findings are also applicable to the interaction between humans and AI. To study these issues, we conducted an experiment (N = 300) with two different tasks, usually performed by dyadic teams (loan assignment and x-ray screening), from two different perspectives (i.e., working together or being evaluated by the agent). In contrast to our hypotheses, the required reliability if working together was equal regardless of the agent. Nevertheless, participants trusted the human more than an AI or DSS. They also required that AI be more reliable than a human when used to evaluate themselves, thus illustrating the importance of changing perspective.
{"title":"How Much Reliability Is Enough? A Context-Specific View on Human Interaction With (Artificial) Agents From Different Perspectives","authors":"Ksenia Appelganc, Tobias Rieger, Eileen Roesler, D. Manzey","doi":"10.1177/15553434221104615","DOIUrl":"https://doi.org/10.1177/15553434221104615","url":null,"abstract":"Tasks classically performed by human–human teams in today’s workplaces are increasingly given to human–technology teams instead. The role of technology is not only played by classic decision support systems (DSSs) but more and more by artificial intelligence (AI). Reliability is a key factor influencing trust in technology. Therefore, we investigated the reliability participants require in order to perceive the support agents (human, AI, and DSS) as “highly reliable.” We then examined how trust differed between these highly reliable agents. Whilst there is a range of research identifying trust as an important determinant in human–DSS interaction, the question is whether these findings are also applicable to the interaction between humans and AI. To study these issues, we conducted an experiment (N = 300) with two different tasks, usually performed by dyadic teams (loan assignment and x-ray screening), from two different perspectives (i.e., working together or being evaluated by the agent). In contrast to our hypotheses, the required reliability if working together was equal regardless of the agent. Nevertheless, participants trusted the human more than an AI or DSS. They also required that AI be more reliable than a human when used to evaluate themselves, thus illustrating the importance of changing perspective.","PeriodicalId":46342,"journal":{"name":"Journal of Cognitive Engineering and Decision Making","volume":"16 1","pages":"207 - 221"},"PeriodicalIF":2.0,"publicationDate":"2022-06-03","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"42031827","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2022-05-25DOI: 10.1177/15553434221103718
Stephen L. Dorton, Samantha B. Harper
Artificial Intelligence (AI) is often viewed as the means by which the intelligence community will cope with increasing amounts of data. There are challenges in adoption, however, as outputs of such systems may be difficult to trust, for a variety of factors. We conducted a naturalistic study using the Critical Incident Technique (CIT) to identify which factors were present in incidents where trust in an AI technology used in intelligence work (i.e., the collection, processing, analysis, and dissemination of intelligence) was gained or lost. We found that explainability and performance of the AI were the most prominent factors in responses; however, several other factors affected the development of trust. Further, most incidents involved two or more trust factors, demonstrating that trust is a multifaceted phenomenon. We also conducted a broader thematic analysis to identify other trends in the data. We found that trust in AI is often affected by the interaction of other people with the AI (i.e., people who develop it or use its outputs), and that involving end users in the development of the AI also affects trust. We provide an overview of key findings, practical implications for design, and possible future areas for research.
{"title":"A Naturalistic Investigation of Trust, AI, and Intelligence Work","authors":"Stephen L. Dorton, Samantha B. Harper","doi":"10.1177/15553434221103718","DOIUrl":"https://doi.org/10.1177/15553434221103718","url":null,"abstract":"Artificial Intelligence (AI) is often viewed as the means by which the intelligence community will cope with increasing amounts of data. There are challenges in adoption, however, as outputs of such systems may be difficult to trust, for a variety of factors. We conducted a naturalistic study using the Critical Incident Technique (CIT) to identify which factors were present in incidents where trust in an AI technology used in intelligence work (i.e., the collection, processing, analysis, and dissemination of intelligence) was gained or lost. We found that explainability and performance of the AI were the most prominent factors in responses; however, several other factors affected the development of trust. Further, most incidents involved two or more trust factors, demonstrating that trust is a multifaceted phenomenon. We also conducted a broader thematic analysis to identify other trends in the data. We found that trust in AI is often affected by the interaction of other people with the AI (i.e., people who develop it or use its outputs), and that involving end users in the development of the AI also affects trust. We provide an overview of key findings, practical implications for design, and possible future areas for research.","PeriodicalId":46342,"journal":{"name":"Journal of Cognitive Engineering and Decision Making","volume":"16 1","pages":"222 - 236"},"PeriodicalIF":2.0,"publicationDate":"2022-05-25","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"46457637","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2022-05-24DOI: 10.1177/15553434221097788
Ulf Andersson, M. Andersson Hagiwara, B. Wireklint Sundström, Henrik Andersson, Hanna Maurin Söderholm
In emergency medical services (EMS), the clinical reasoning (CR) of registered nurses (RNs) working in ambulance care plays an important role in providing care and treatment that is timely, accurate, appropriate and safe. However, limited existing knowledge about how CR is formed and influenced by the EMS mission hinders the development of service provision and decision support tools for RNs that would further enhance patient safety. To explore the nature of CR and influencing factors in this context, an inductive case study examined 34 observed patient–RN encounters in an EMS setting focusing on ambulance care. The results reveal a fragmented CR approach involving several parallel decision-making processes grounded in and led by patients’ narratives. The findings indicate that RNs are not always aware of their own CR and associated influences until they actively reflect on the process, and additional research is needed to clarify this complex phenomenon.
{"title":"Clinical Reasoning among Registered Nurses in Emergency Medical Services: A Case Study","authors":"Ulf Andersson, M. Andersson Hagiwara, B. Wireklint Sundström, Henrik Andersson, Hanna Maurin Söderholm","doi":"10.1177/15553434221097788","DOIUrl":"https://doi.org/10.1177/15553434221097788","url":null,"abstract":"In emergency medical services (EMS), the clinical reasoning (CR) of registered nurses (RNs) working in ambulance care plays an important role in providing care and treatment that is timely, accurate, appropriate and safe. However, limited existing knowledge about how CR is formed and influenced by the EMS mission hinders the development of service provision and decision support tools for RNs that would further enhance patient safety. To explore the nature of CR and influencing factors in this context, an inductive case study examined 34 observed patient–RN encounters in an EMS setting focusing on ambulance care. The results reveal a fragmented CR approach involving several parallel decision-making processes grounded in and led by patients’ narratives. The findings indicate that RNs are not always aware of their own CR and associated influences until they actively reflect on the process, and additional research is needed to clarify this complex phenomenon.","PeriodicalId":46342,"journal":{"name":"Journal of Cognitive Engineering and Decision Making","volume":"16 1","pages":"123 - 156"},"PeriodicalIF":2.0,"publicationDate":"2022-05-24","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"48106227","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2022-05-22DOI: 10.1177/15553434221104600
Dakota D. Scott, Lisa Vangsness, Joel Suss
The few perceptual–cognitive expertise and deception studies in the domain of law enforcement have yet to examine perceptual–cognitive expertise differences of police trainees and police officers. The current study uses methods from the perceptual–cognitive expertise and deception models. Participants watched temporally occluded videos of actors honestly drawing a weapon and deceptively drawing a non-weapon from a concealed location on their body. Participants determined if the actor was holding a weapon or a non-weapon. Using signal-detection metrics—sensitivity and response bias—we did not find evidence of perceptual–cognitive expertise; performance measures did not differ significantly between police trainees and experienced officers. However, consistent with the hypotheses, we did find that both police trainees and police officers became more sensitive in identifying the object as occlusion points progressed. Additionally, we found that across police trainees and police officers, their response bias became more liberal (i.e., more likely to identify the object as a weapon) as occlusion points progressed. This information has potential impacts for law enforcement practices and additional research.
{"title":"Perceptual–Cognitive Expertise in Law Enforcement: An Object-Identification Task","authors":"Dakota D. Scott, Lisa Vangsness, Joel Suss","doi":"10.1177/15553434221104600","DOIUrl":"https://doi.org/10.1177/15553434221104600","url":null,"abstract":"The few perceptual–cognitive expertise and deception studies in the domain of law enforcement have yet to examine perceptual–cognitive expertise differences of police trainees and police officers. The current study uses methods from the perceptual–cognitive expertise and deception models. Participants watched temporally occluded videos of actors honestly drawing a weapon and deceptively drawing a non-weapon from a concealed location on their body. Participants determined if the actor was holding a weapon or a non-weapon. Using signal-detection metrics—sensitivity and response bias—we did not find evidence of perceptual–cognitive expertise; performance measures did not differ significantly between police trainees and experienced officers. However, consistent with the hypotheses, we did find that both police trainees and police officers became more sensitive in identifying the object as occlusion points progressed. Additionally, we found that across police trainees and police officers, their response bias became more liberal (i.e., more likely to identify the object as a weapon) as occlusion points progressed. This information has potential impacts for law enforcement practices and additional research.","PeriodicalId":46342,"journal":{"name":"Journal of Cognitive Engineering and Decision Making","volume":"16 1","pages":"157 - 176"},"PeriodicalIF":2.0,"publicationDate":"2022-05-22","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"47000222","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2022-05-06DOI: 10.1177/15553434221096261
N. Tenhundfeld, Mustafa Demir, E. D. de Visser
Trust in automation is a foundational principle in Human Factors Engineering. An understanding of trust can help predict and alter much of human-machine interaction (HMI). However, despite the utility of assessing trust in automation in applied settings, there are inherent and unique challenges in trust assessment for those who seek to do so outside of the confines of the sterile lab environment. Because of these challenges, new approaches for trust in automation assessment need to be developed to best suit the unique demands of trust assessment in the real world. This paper lays out six requirements for these future measures: they should (1) be short, unobtrusive, and interaction-based, (2) be context-specific and adaptable, (3) be dynamic, (4) account for autonomy versus automation dependency, (5) account for task dependency, and (6) account for levels of risk. For the benefits of trust assessment to be realized in the “real world,” future research needs to leverage the existing body of literature on trust in automation while looking toward the needs of the practitioner.
{"title":"Assessment of Trust in Automation in the “Real World”: Requirements for New Trust in Automation Measurement Techniques for Use by Practitioners","authors":"N. Tenhundfeld, Mustafa Demir, E. D. de Visser","doi":"10.1177/15553434221096261","DOIUrl":"https://doi.org/10.1177/15553434221096261","url":null,"abstract":"Trust in automation is a foundational principle in Human Factors Engineering. An understanding of trust can help predict and alter much of human-machine interaction (HMI). However, despite the utility of assessing trust in automation in applied settings, there are inherent and unique challenges in trust assessment for those who seek to do so outside of the confines of the sterile lab environment. Because of these challenges, new approaches for trust in automation assessment need to be developed to best suit the unique demands of trust assessment in the real world. This paper lays out six requirements for these future measures: they should (1) be short, unobtrusive, and interaction-based, (2) be context-specific and adaptable, (3) be dynamic, (4) account for autonomy versus automation dependency, (5) account for task dependency, and (6) account for levels of risk. For the benefits of trust assessment to be realized in the “real world,” future research needs to leverage the existing body of literature on trust in automation while looking toward the needs of the practitioner.","PeriodicalId":46342,"journal":{"name":"Journal of Cognitive Engineering and Decision Making","volume":"16 1","pages":"101 - 118"},"PeriodicalIF":2.0,"publicationDate":"2022-05-06","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"42330259","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2022-04-25DOI: 10.1177/15553434221092930
Lillian M. Rigoli, Gaurav Patil, Patrick Nalepka, Rachel W. Kallen, S. Hosking, Christopher J. Best, Michael J. Richardson
Effective team performance often requires that individuals engage in team training exercises. However, organizing team-training scenarios presents economic and logistical challenges and can be prone to trainer bias and fatigue. Accordingly, a growing body of research is investigating the effectiveness of employing artificial agents (AAs) as synthetic teammates in team training simulations, and, relatedly, how to best develop AAs capable of robust, human-like behavioral interaction. Motivated by these challenges, the current study examined whether task dynamical models of expert human herding behavior could be embedded in the control architecture of AAs to train novice actors to perform a complex multiagent herding task. Training outcomes were compared to human-expert trainers, novice baseline performance, and AAs developed using deep reinforcement learning (DRL). Participants’ subjective preferences for the AAs developed using DRL or dynamical models of human performance were also investigated. The results revealed that AAs controlled by dynamical models of human expert performance could train novice actors at levels equivalent to expert human trainers and were also preferred over AAs developed using DRL. The implications for the development of AAs for robust human-AA interaction and training are discussed, including the potential benefits of employing hybrid Dynamical-DRL techniques for AA development.
{"title":"A Comparison of Dynamical Perceptual-Motor Primitives and Deep Reinforcement Learning for Human-Artificial Agent Training Systems","authors":"Lillian M. Rigoli, Gaurav Patil, Patrick Nalepka, Rachel W. Kallen, S. Hosking, Christopher J. Best, Michael J. Richardson","doi":"10.1177/15553434221092930","DOIUrl":"https://doi.org/10.1177/15553434221092930","url":null,"abstract":"Effective team performance often requires that individuals engage in team training exercises. However, organizing team-training scenarios presents economic and logistical challenges and can be prone to trainer bias and fatigue. Accordingly, a growing body of research is investigating the effectiveness of employing artificial agents (AAs) as synthetic teammates in team training simulations, and, relatedly, how to best develop AAs capable of robust, human-like behavioral interaction. Motivated by these challenges, the current study examined whether task dynamical models of expert human herding behavior could be embedded in the control architecture of AAs to train novice actors to perform a complex multiagent herding task. Training outcomes were compared to human-expert trainers, novice baseline performance, and AAs developed using deep reinforcement learning (DRL). Participants’ subjective preferences for the AAs developed using DRL or dynamical models of human performance were also investigated. The results revealed that AAs controlled by dynamical models of human expert performance could train novice actors at levels equivalent to expert human trainers and were also preferred over AAs developed using DRL. The implications for the development of AAs for robust human-AA interaction and training are discussed, including the potential benefits of employing hybrid Dynamical-DRL techniques for AA development.","PeriodicalId":46342,"journal":{"name":"Journal of Cognitive Engineering and Decision Making","volume":"16 1","pages":"79 - 100"},"PeriodicalIF":2.0,"publicationDate":"2022-04-25","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"44193336","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2022-04-21DOI: 10.1177/15553434221085114
Lamia Alam, Shane T. Mueller
AI systems are increasingly being developed to provide the first point of contact for patients. These systems are typically focused on question-answering and integrating chat systems with diagnostic algorithms, but are likely to suffer from many of the same deficiencies in explanation that have plagued medical diagnostic systems since the 1970s (Shortliffe, 1979). To provide better guidance about how such systems should approach explanations, we report on an interview study in which we identified explanations that physicians used in the context of re-diagnosis or a change in diagnosis. Seven current and former physicians with a variety of specialties and experience were recruited to take part in the interviews. Several high-level observations were made by reviewing the interview notes. Nine broad categories of explanation emerged from the thematic analysis of the explanation contents. We also present these in a diagnosis meta-timeline that encapsulates many of the commonalities we saw across diagnoses during the interviews. Based on the results, we provided some design recommendations to consider for developing diagnostic AI systems. Altogether, this study suggests explanation strategies, approaches, and methods that might be used by medical diagnostic AI systems to improve user trust and satisfaction with these systems.
{"title":"Examining Physicians’ Explanatory Reasoning in Re-Diagnosis Scenarios for Improving AI Diagnostic Systems","authors":"Lamia Alam, Shane T. Mueller","doi":"10.1177/15553434221085114","DOIUrl":"https://doi.org/10.1177/15553434221085114","url":null,"abstract":"AI systems are increasingly being developed to provide the first point of contact for patients. These systems are typically focused on question-answering and integrating chat systems with diagnostic algorithms, but are likely to suffer from many of the same deficiencies in explanation that have plagued medical diagnostic systems since the 1970s (Shortliffe, 1979). To provide better guidance about how such systems should approach explanations, we report on an interview study in which we identified explanations that physicians used in the context of re-diagnosis or a change in diagnosis. Seven current and former physicians with a variety of specialties and experience were recruited to take part in the interviews. Several high-level observations were made by reviewing the interview notes. Nine broad categories of explanation emerged from the thematic analysis of the explanation contents. We also present these in a diagnosis meta-timeline that encapsulates many of the commonalities we saw across diagnoses during the interviews. Based on the results, we provided some design recommendations to consider for developing diagnostic AI systems. Altogether, this study suggests explanation strategies, approaches, and methods that might be used by medical diagnostic AI systems to improve user trust and satisfaction with these systems.","PeriodicalId":46342,"journal":{"name":"Journal of Cognitive Engineering and Decision Making","volume":"16 1","pages":"63 - 78"},"PeriodicalIF":2.0,"publicationDate":"2022-04-21","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"49040783","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
京公网安备 11010802042870号
京ICP备2023020795号-1
扫码关注我们
求助内容：
应助结果提醒方式：