Year: 2023
Type: dissertation
Abstract: Credibility assessment systems are a special class of information systems that collect and analyze psychophysiological and behavioral signals to remedy humans’ inability to judge lies and truth. On one hand, these technologies share commonalities with other information systems. The development and evaluation of IT artifacts (including credibility assessment systems) demonstrate rigor and relevance to the real world by drawing from a multidisciplinary knowledge base and addressing various organizational and societal needs, such as insider threat detection and criminal justice. On the other hand, credibility assessment researchers face their unique challenges. For example, for the application of insider threat detection, existing methods are ineffective for early warnings. To address this limitation, the first essay proposes to use insiders’ behavioral features (e.g., facial expressions, eye gaze, and head pose) triggered by goal-congruence stimuli to detect malicious insiders (MIs). Goal-congruence stimuli notifies individuals of information in agreement with the benevolent or malicious goal (e.g., whether an organizational goal has or has not been achieved) and enables early detection by inducing behavioral anomalies from MIs before insider attacks. The method is evaluated through lab experiments that closely simulate government and corporate espionage. Another challenge in credibility assessment research is the lack of data analytic methods that capture human behavioral coordination, although research has shown that behavioral coordination contains rich information about message senders’ mental states and message receivers’ perception towards them. In the second essay, I view multimodal human behaviors as a complex and coordinated system and propose a novel approach to explicitly modeling behavioral coordination with weighted behavioral co-expression networks (CoNets). I posit patterns of behavioral coordination are revealed by changes in different CoNet properties (e.g., degree) and illustrate CoNets’ utility with deception detection and interview performance evaluation as two example tasks. A third challenge is the widespread public concern over the ethical and proper use of credibility assessment systems. To address this concern and guide public policies, the third essay builds a computational framework to study the impact of credibility assessment systems on the fairness of criminal investigations.
Authors: Wang, Xinran, Wang, Xinran
Language: English
Subfield: Social Psychology
Field: Psychology
Domain: Social Sciences
Open Access status: green
OpenAlex ID: W6990412331