Sarah A. Jessup
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Research Subjects
Sarah A. Jessup has a total of 53 co-authors. Here are the top co-authors by number of shared publications:
- Gene Michael Alarcon
Air Force Research Laboratory
16 shared publications - August Capiola
Air Force Research Laboratory
11 shared publications Gene M. Alarcon
6 shared publications- Sasha M. Willis
Air Force Research Laboratory, Wright-Paterson AFB, Dayton, OH, USA
5 shared publications - Rose Gamble
Tandy School of Computer Science University of Tulsa 74104 Tulsa Oklahoma USA
4 shared publications - Anthony M. Gibson
Consortium of Universities, Washington, DC 20036, USA
3 shared publications
Latest Publications
- 10 October 2026
- Published by Elsevier in Applied Ergonomics
- Vol. 140
Understanding preferences for explainable AI: A study on LIME variations
- 10 October 2026
- Published by Elsevier in Applied Ergonomics
- Vol. 140
Understanding preferences for explainable AI: A study on LIME variations
- Open Access
- 2 August 2026
- Published by Elsevier in SSRN Electronic Journal
Performance, Process, and Purpose in Human–Machine Interaction: Three Experimental Studies Investigating Machine Learning Transparency Techniques
Lee and See’s (2004) model of trust in automation identifies performance, process, and purpose as the primary antecedents of perceived machine trustworthiness. Although this framework has become foundational in human–machine interaction research, few studies have experimentally manipulated these trustworthiness dimensions with design features that directly correspond to their theoretical definitions. Across three experiments, we tested Lee and See’s model by manipulating objective ML transparency features in an image-classification task. Study 1 examined calibrated confidence estimates as a performance manipulation, Study 2 investigated explainable artificial intelligence (XAI) using Local Interpretable Model-Agnostic Explanations (LIME) as a process manipulation, and Study 3 manipulated transfer learning to examine perceptions of model purpose. Trustworthiness perceptions and behavioral performance were assessed throughout the interaction. Across studies, trustworthiness perceptions generally declined over time as participants observed model imperfections. Higher objective accuracy consistently improved all three trustworthiness perceptions and enhanced task performance. The transparency manipulations influenced their intended trustworthiness dimensions, with calibrated confidence estimates strengthening performance perceptions, XAI enhancing process perceptions, and transfer-learning information shaping purpose perceptions, but cross-dimensional effects were also observed. Together, the findings provide one of the first comprehensive experimental tests of Lee and See’s trustworthiness framework using contemporary ML design techniques. The results demonstrate that objective transparency mechanisms influence users’ subjective trustworthiness perceptions in distinct ways and highlight how performance, process, and purpose information can be intentionally designed to improve trust calibration and human–ML collaboration.
- 2 August 2026
- Published by Elsevier in SSRN Electronic Journal
Open AccessPerformance, Process, and Purpose in Human–Machine Interaction: Three Experimental Studies Investigating Machine Learning Transparency Techniques
Lee and See’s (2004) model of trust in automation identifies performance, process, and purpose as the primary antecedents of perceived machine trustworthiness. Although this framework has become foundational in human–machine interaction research, few studies have experimentally manipulated these trustworthiness dimensions with design features that directly correspond to their theoretical definitions. Across three experiments, we tested Lee and See’s model by manipulating objective ML transparency features in an image-classification task. Study 1 examined calibrated confidence estimates as a performance manipulation, Study 2 investigated explainable artificial intelligence (XAI) using Local Interpretable Model-Agnostic Explanations (LIME) as a process manipulation, and Study 3 manipulated transfer learning to examine perceptions of model purpose. Trustworthiness perceptions and behavioral performance were assessed throughout the interaction. Across studies, trustworthiness perceptions generally declined over time as participants observed model imperfections. Higher objective accuracy consistently improved all three trustworthiness perceptions and enhanced task performance. The transparency manipulations influenced their intended trustworthiness dimensions, with calibrated confidence estimates strengthening performance perceptions, XAI enhancing process perceptions, and transfer-learning information shaping purpose perceptions, but cross-dimensional effects were also observed. Together, the findings provide one of the first comprehensive experimental tests of Lee and See’s trustworthiness framework using contemporary ML design techniques. The results demonstrate that objective transparency mechanisms influence users’ subjective trustworthiness perceptions in distinct ways and highlight how performance, process, and purpose information can be intentionally designed to improve trust calibration and human–ML collaboration.
- Open Access
- 13 July 2026
- Published by Elsevier in SSRN Electronic Journal
Calibrated Confidence and the Trust Process: Empirical Tests of the Lee and See Model in AI Decision-Making
Artificial intelligence and machine learning (AI/ML) models are becoming increasingly prevalent in decision-making. Despite their increase application in many areas, little research has been conducted on human-machine interaction with ML. New models that are developed in computer science literature advocate they increase transparency, trust, and task performance for users, but little to no research has empirically tested these postulates. We tested aspects of the Lee and See (2004) theoretical model in relation to machine trust, exploring differences in the trust process between calibrated confidence models and uncalibrated confidence models across three studies in a dual-task environment. Results indicate calibrated confidence intervals increased task performance on the non-automated task (i.e., the task without ML), but not the automated task. Calibrated confidence models also resulted in higher trustworthiness perceptions and reliance intentions over time. Propensity to trust machines was related to initial trustworthiness perceptions but not actual behaviors in all three studies. Trustworthiness perceptions accounted for significant variance in relevant criteria, but reliance intentions did not. Results demonstrated calibrated confidence models increased the trust process, but we found mixed support for Lee and See’s theoretical model.
- 13 July 2026
- Published by Elsevier in SSRN Electronic Journal
Open AccessCalibrated Confidence and the Trust Process: Empirical Tests of the Lee and See Model in AI Decision-Making
Artificial intelligence and machine learning (AI/ML) models are becoming increasingly prevalent in decision-making. Despite their increase application in many areas, little research has been conducted on human-machine interaction with ML. New models that are developed in computer science literature advocate they increase transparency, trust, and task performance for users, but little to no research has empirically tested these postulates. We tested aspects of the Lee and See (2004) theoretical model in relation to machine trust, exploring differences in the trust process between calibrated confidence models and uncalibrated confidence models across three studies in a dual-task environment. Results indicate calibrated confidence intervals increased task performance on the non-automated task (i.e., the task without ML), but not the automated task. Calibrated confidence models also resulted in higher trustworthiness perceptions and reliance intentions over time. Propensity to trust machines was related to initial trustworthiness perceptions but not actual behaviors in all three studies. Trustworthiness perceptions accounted for significant variance in relevant criteria, but reliance intentions did not. Results demonstrated calibrated confidence models increased the trust process, but we found mixed support for Lee and See’s theoretical model.
- 26 January 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Human-Machine Systems
- Vol. 56 (2) , 301-310
The Trust Process Applied to Machine Learning Algorithms: The Influence of Calibrated Confidence Estimates
Little research has been conducted on user interactions with machine learning models and users' trust. The current study explores differences in psychological perceptions of trust between two machine learning models. Across three studies, participants (N = 130, 134, 132) were recruited online and completed an image-binning monitoring task with either an uncalibrated classification model (UCM) or a calibrated classification model (CCM) that provided confidence estimates about their decisions. The UCM was highly confident, regardless of images from out-of-distribution classes, whereas the CCM was better calibrated between accuracy and confidence of out-of-distribution images. Results revealed participants performed better on the task in the CCM condition across all studies. In addition, trustworthiness perception results differed depending on whether the study included out-of-distribution images. Overall, calibrated confidence estimates displayed by CCMs can be an effective means of increasing transparency and enhancing our understanding of trust in machines, but more research is needed on trust perceptions in image classification tasks.
- 26 January 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Human-Machine Systems
- Vol. 56 (2) , 301-310
The Trust Process Applied to Machine Learning Algorithms: The Influence of Calibrated Confidence Estimates
Little research has been conducted on user interactions with machine learning models and users' trust. The current study explores differences in psychological perceptions of trust between two machine learning models. Across three studies, participants (N = 130, 134, 132) were recruited online and completed an image-binning monitoring task with either an uncalibrated classification model (UCM) or a calibrated classification model (CCM) that provided confidence estimates about their decisions. The UCM was highly confident, regardless of images from out-of-distribution classes, whereas the CCM was better calibrated between accuracy and confidence of out-of-distribution images. Results revealed participants performed better on the task in the CCM condition across all studies. In addition, trustworthiness perception results differed depending on whether the study included out-of-distribution images. Overall, calibrated confidence estimates displayed by CCMs can be an effective means of increasing transparency and enhancing our understanding of trust in machines, but more research is needed on trust perceptions in image classification tasks.
- 19 November 2025
- Published by Elsevier in Applied Ergonomics
- Vol. 132, 104657
“Are you sure about that?” The effects of calibrated classification model task accuracy and confidence on trustworthiness, trust, and performance
- 19 November 2025
- Published by Elsevier in Applied Ergonomics
- Vol. 132, 104657
“Are you sure about that?” The effects of calibrated classification model task accuracy and confidence on trustworthiness, trust, and performance
- 4 August 2025
- Published by Springer Nature in International Journal of Social Robotics
- Vol. 17 (8) , 1437-1452
The Influence of Robot Anthropomorphism and Trust Violation Types on Trustworthiness Perceptions and Trust Behaviors
- 4 August 2025
- Published by Springer Nature in International Journal of Social Robotics
- Vol. 17 (8) , 1437-1452
The Influence of Robot Anthropomorphism and Trust Violation Types on Trustworthiness Perceptions and Trust Behaviors
- Open Access
- 1 January 2025
- Published by HICSS Conference Office
Exploring User Evaluations of Machine Learning Models: A Qualitative Study on the Impact of Confidence Intervals
Research on artificial intelligence and machine learning models has burgeoned in the last decade. However, research has seldom utilized qualitative methods for assessing user-based experiences and system evaluations of AI/ML models. This study aims to provide an example of how thematic text analysis can be used to provide greater insight into user experiences with these systems and examine how varying levels of model transparency affects evaluations. Participants (N = 130) completed an image binning monitoring task with either an uncalibrated classification model (UCM), which displayed high confidence regardless of classification accuracy or a calibrated classification model (CCM), which had greater calibration between accuracy and confidence. Results revealed detailed information on user evaluations for both models including various performance perceptions, impressions, and strategy behaviors. Furthermore, we identified key differences in user evaluations between these models and our confidence manipulation, such as greater trust and confidence display use. Qualitative analysis has been shown to be an effective approach for detailed investigation of user experiences and model evaluation.
- 1 January 2025
- Published by HICSS Conference Office
Open AccessExploring User Evaluations of Machine Learning Models: A Qualitative Study on the Impact of Confidence Intervals
Research on artificial intelligence and machine learning models has burgeoned in the last decade. However, research has seldom utilized qualitative methods for assessing user-based experiences and system evaluations of AI/ML models. This study aims to provide an example of how thematic text analysis can be used to provide greater insight into user experiences with these systems and examine how varying levels of model transparency affects evaluations. Participants (N = 130) completed an image binning monitoring task with either an uncalibrated classification model (UCM), which displayed high confidence regardless of classification accuracy or a calibrated classification model (CCM), which had greater calibration between accuracy and confidence. Results revealed detailed information on user evaluations for both models including various performance perceptions, impressions, and strategy behaviors. Furthermore, we identified key differences in user evaluations between these models and our confidence manipulation, such as greater trust and confidence display use. Qualitative analysis has been shown to be an effective approach for detailed investigation of user experiences and model evaluation.
- Open Access
- 1 January 2025
- Published by HICSS Conference Office
What do I (and You) Have to Lose? Investigating Effects of Risk Asymmetry on the Trust Process in Distributed Dyads
Popular theories of trust note the importance of risk in the trust process. However, real-world risk is not homogenous: one party may stand to lose more (or less) based on the actions of another, and the degree of this asymmetry may change. The present research focused on comparative vulnerability and risk discrepancy between parties in a novel task. The study investigated their moderating effects on the trust process through the lens of exchange theory. Participants were recruited online to participate in a novel task designed to emulate real-world constraints present in computer-mediated ad hoc dyads. The results showed that in several cases, relative vulnerabilities and the extent of risk discrepancy moderated the effect of partner behaviors on trust-relevant criteria. This work is a step toward quantifying the effects of changing comparative vulnerability and risk discrepancy on the trust process and has implications for distributed dyads.
- 1 January 2025
- Published by HICSS Conference Office
Open AccessWhat do I (and You) Have to Lose? Investigating Effects of Risk Asymmetry on the Trust Process in Distributed Dyads
Popular theories of trust note the importance of risk in the trust process. However, real-world risk is not homogenous: one party may stand to lose more (or less) based on the actions of another, and the degree of this asymmetry may change. The present research focused on comparative vulnerability and risk discrepancy between parties in a novel task. The study investigated their moderating effects on the trust process through the lens of exchange theory. Participants were recruited online to participate in a novel task designed to emulate real-world constraints present in computer-mediated ad hoc dyads. The results showed that in several cases, relative vulnerabilities and the extent of risk discrepancy moderated the effect of partner behaviors on trust-relevant criteria. This work is a step toward quantifying the effects of changing comparative vulnerability and risk discrepancy on the trust process and has implications for distributed dyads.
- 25 October 2024
- Published by Elsevier in Personality and Individual Differences
- Vol. 233
Jazz hands and jitters: Exploring valence and arousal dimensions with multidimensional scaling techniques
- 25 October 2024
- Published by Elsevier in Personality and Individual Differences
- Vol. 233
Jazz hands and jitters: Exploring valence and arousal dimensions with multidimensional scaling techniques