Cognitive Vulnerability and Human Judgment
Understanding how framing, anchoring, primacy effects, functional fixedness, and other cognitive shortcuts influence human–AI interactions.
Trustworthy AI Technologies and Experiences (TATE)
Trustworthy AI Technologies and Experiences (TATE) is a research initiative within the HCIR Lab at the University of Wisconsin–Milwaukee. TATE investigates how people interact with increasingly capable AI systems, how cognitive limitations and algorithmic behavior may jointly shape human judgment, and how AI technologies can be designed and evaluated to support informed, equitable, and resilient decision-making.
Our research integrates information retrieval, machine learning, human-computer interaction, behavioral science, and participatory methods. We seek to develop reproducible evaluation frameworks, AI auditing methods, and practical interventions that address cognitive vulnerability, unreliable AI assistance, and harmful patterns of personalization.
Research Themes
AI Auditing, Reliability, and Bias Mitigation
Developing benchmarks, controlled experiments, and auditing methods for identifying unreliable or biased AI judgments and interactions.
Agentic AI, Personalization, and Human Agency
Examining how AI agents learn from interaction histories, retain user preferences, and potentially reinforce narrow choices or behavioral patterns over time.
AI Literacy, Learning, and Community Engagement
Evaluating AI-supported education, library services, and accessible technologies that help individuals and communities develop informed and responsible AI practices.
Research Methods and Infrastructure
TATE combines computational evaluation and behavioral research through experimental AI systems, benchmark datasets, controlled and longitudinal studies, and human-centered auditing. Open research infrastructure, including the ECHO platform, supports reproducibility and interdisciplinary collaboration.
Education and Community Engagement
Proposed and developing activities
TATE also seeks to connect research in computing and AI with socially meaningful applications. Potential activities include interdisciplinary student research, postdoctoral research opportunities, community-engaged technology capstones, public-facing workshops on responsible AI, and student research showcases. We aim to develop collaborations across SOIS, UWM's College of Community Engagement and Professions, and external community and industry partners.
These activities are in development and are not currently funded programs. No postdoctoral positions are being advertised at this time.
Selected Publications
- Liu, J. (2026). Bounded Minds, Generative Machines: Envisioning Conversational AI that Works with Human Heuristics and Reduces Bias Risk. Communications of the ACM.
- De Lima, V., Liu, J., & Yang, G. H. (2026). Stability Regimes for Framing-Sensitive Fine-Tuning in Language Models. NeurIPS 2026 (accepted).
- Oh, P. S., Luo, Y., & Liu, J. (2026). Stuck on the First Answer? Auditing Primacy Bias in Multi-Turn LLM Preference Revision via Causal Masking. ACM/IEEE JCDL 2026.
- He, J., & Liu, J. (2026). Not All Transparency Is Equal: Source Presentation Effects on Attention, Interaction, and Persuasion in Conversational Search. ACM CHIIR 2026.
- Chen, N., Fang, H., Liu, J., Wei, W., Sakai, T., & Wu, X.-M. (2026). Mitigating the Threshold Priming Effect in Large Language Model-Based Relevance Judgments via Personality Simulation. ACM WSDM 2026.
- Liu, J., & He, J. (2026). The Decoy Dilemma in Online Medical Information Evaluation: A Comparative Study of Credibility Assessments by LLM and Human Judges. ACM Transactions on Interactive Intelligent Systems.