Ahana Biswas | Human–AI Interaction, Complex Systems & Computational Social Science

About Me

I am a PhD Candidate in Information Science at the University of Pittsburgh, PICSO Lab, advised by Prof. Yu-Ru Lin. My research lies at the intersection of human–AI interaction, responsible AI, and computational social science. I study how people decide when to trust, verify, and rely on AI systems, and how these individual interactions scale through feedback, social learning, and network structure into broader collective outcomes.

My dissertation examines verification in real-world LLM information seeking: what verification support AI systems provide, how users respond to epistemic cues such as sources and uncertainty, and how people decide when checking an AI response is worth the effort. More broadly, I study feedback between people and computational systems, including how algorithmic environments shape attention, engagement, and subsequent behavior.

Methodologically, I combine large-scale behavioral data, causal inference, NLP and LLM-based measurement, network analysis, computational modeling, mixed-effects and longitudinal analysis, and qualitative research. My research has been published in venues including ICWSM, Computational Social Science (CSS), and the Harvard Kennedy School Misinformation Review, and presented at conferences and workshops across human–AI interaction, computational social science, network science, and complex systems.

If you are interested in my work or would like to connect, feel free to reach out!

On the 2026–27 job market for postdoctoral and industry research positions in human–AI interaction, complex systems, and computational social science!

Research Areas

Human–AI Interaction: Trust, Verification, and Reliance

I study how people evaluate and act on information produced by AI systems: when they trust an AI response, when they verify it, what signals trigger checking, and when they ultimately rely on AI-generated information. My dissertation examines these processes in naturally occurring multi-turn LLM interactions and through qualitative studies of users' verification and reliance decisions.

Verification and User Scrutiny in Real-World LLM Conversations: I study the verification environments created by deployed LLM systems and examine how sources, attribution, uncertainty, reasoning, and other response cues relate to users' subsequent accepting, checking, and challenging behavior. Paper
Human Verification and Reliance Decisions: I study how users interpret verification affordances, decide when checking is worth the effort, and adjust their reliance on AI across different tasks and levels of risk.

Human–AI Feedback and Complex Adaptive Systems

I study human–AI interaction not only as an individual decision problem, but also as a population-level process. I use computational and network models to examine how repeated reliance, social learning, visible behavior, verification, and network structure interact to produce collective outcomes.

Modeling AI Overreliance as a Complex Adaptive System (CSS 2026): A computational model examining how individual learning, social learning, and visible reliance can generate population-level overreliance, and how verification can alter these dynamics.
Regimes of Influence in Trust-Uncertainty Gated Networks: Network models examining how trust, distrust, uncertainty, and network structure jointly determine patterns of interpersonal influence and collective outcomes.

Algorithmic Feedback and Behavior in Socio-Technical Systems

My broader work examines how algorithmic visibility, audience feedback, and social interaction shape subsequent behavior in large-scale socio-technical systems. Using causal inference, network analysis, and longitudinal behavioral data, I study how local interactions between people and computational environments accumulate into larger patterns of attention, engagement, and communication.

From Attention to Dialogue (ICWSM 2026): A longitudinal study of how audience engagement feeds back into subsequent communication behavior.
Political Elites in the Attention Economy (ICWSM 2025): A network- and causally informed analysis of how visibility signals and audience responses shape communication incentives.
Toxic Politics & TikTok Engagement (HKS Misinformation Review, 2025): A multimodal analysis of relationships among content characteristics, platform visibility, and audience engagement.

Research Experience

Graduate Student Researcher — PICSO Lab, University of Pittsburgh
2023–Present
Human–AI interaction, trust and reliance, computational social science, causal inference, network analysis, and large-scale behavioral data.

Applied Scientist Intern — Amazon, Supply Chain Optimization Technologies
Summer 2026
Built a multi-agent LLM surrogate for a complex operational simulator and developed self-verification mechanisms for reliable counterfactual decision support. The project received the People's Choice Award at Amazon's Global AI Solutions Expo and won multiple internal hackathons.

Research Intern — Phenom People Inc.
Summer 2025
Developed causal and counterfactual explanation methods for improving the interpretability of AI-driven candidate–job matching.

Updates

[Nov 2026] New preprint: Signals, Evidence, and Context: User Scrutiny in Real-World LLM Conversations is now available on arXiv.
[Oct 2026] Passed my dissertation proposal defense!
[Oct 2026] Presented Modeling AI Overreliance as a Complex Adaptive System at Computational Social Science (CSS) 2026 in Santa Fe, NM.
[Sep 2026] Wrapped up my Applied Scientist internship at Amazon, where I developed a multi-agent LLM surrogate with self-verification for rapid counterfactual analysis. The project received the People’s Choice Award at Amazon’s Global AI Solutions Expo and won multiple internal hackathons.
[Sep 2026] Teaching Fellow for INFSCI 0201: Intermediate Programming with Python at the University of Pittsburgh this fall.
[Aug 2026] New paper: Modeling AI Overreliance as a Complex Adaptive System, accepted at Computational Social Science (CSS) 2026.
[Aug 2026] Received a Student Fellowship to attend Computational Social Science (CSS) 2026.
[Jun 2026] Participated in the Santa Fe Institute (SFI) Graduate Workshop in Computational Social Science.
[Apr 2026] 3 submissions accepted at IC2S2 2026: 2 parallel talks + 1 poster!
[Mar 2026] Our paper From Attention to Dialogue: Does Audience Engagement Reinforce Constructive Cross-Party Communication? was accepted at ICWSM 2026.
[Mar 2026] Abstract accepted for a Contributed Talk at NetSci 2026.
[Dec 2025] Our work on toxic politics on TikTok was mentioned in Forbes.

[Aug 2025] Presented a poster at PolNet-PACSS, Boston!
[Jul 2025] Our paper Toxic Politics and TikTok Engagement in the 2024 U.S. Election was accepted in the HKS Misinformation Review.
[Jul 2025] Started a summer research internship at Phenom, working on interpretable AI explanations for matching models.
[Jun 2025] Two oral presentations at ICWSM 2025 (Copenhagen) and Sunbelt 2025 (Paris)!
[Apr 2025] Our findings on elite influence were featured in The Conversation.
[Apr 2025] Passed my Comprehensive Exam!
[Apr 2025] Presented a talk on elite visibility dynamics at the IEEE Technical Symposium in Western Pennsylvania.
[Jul 2024] Paper accepted to ICWSM 2025: Political Elites in the Attention Economy: Visibility Over Civility and Credibility?
[Jul 2024] Attended SICSS and IC2S2 at the University of Pennsylvania.
[Mar 2024] Won People's Choice Award in the SCI 3-Minute Thesis Competition.
[Mar 2024] Passed my Preliminary Exam with Distinction.

Publications

Working Papers

Signals, Evidence, and Context: User Scrutiny in Real-World LLM Conversations. Biswas, A., & Lin, Y.-R. Under review at ACM FAccT, 2026. Paper

Regimes of Influence in Trust-Uncertainty Gated Networks. Masoumi, R., Biswas, A., & Lin, Y.-R. Under review at Nature Communications, 2026. Paper

Conference & Journal Publications

Modeling AI Overreliance as a Complex Adaptive System. Biswas, A. Computational Social Science (CSS), 2026. Paper

From Attention to Dialogue: Does Audience Engagement Reinforce Constructive Cross-Party Communication? Biswas, A., & Lin, Y.-R. International AAAI Conference on Web and Social Media (ICWSM), 2026. Paper

Toxic Politics and TikTok Engagement in the 2024 U.S. Election. Biswas, A., Javadian Sabet, A., & Lin, Y.-R. Harvard Kennedy School Misinformation Review, 2025. Paper Dataset Code

• Coverage in Wikipedia , Forbes , Pitt News

Political Elites in the Attention Economy: Visibility Over Civility and Credibility? Biswas, A., Lin, Y.-R., Tai, Y.-C., & Desmarais, B. A. International AAAI Conference on Web and Social Media (ICWSM), 2025. Paper

• Coverage in The Conversation

Political Narratives During the Russian Invasion of Ukraine. Biswas, A., Niven, T., & Lin, Y.-R. Social, Cultural, and Behavioral Modeling (SBP-BRiMS), 2023. Paper

Development of a COVID-19 Related Anti-Asian Tweet Data Set: Quantitative Study. Mokhberi, M., Biswas, A., Masud, Z., et al. JMIR Formative Research, 2023. Paper

Teaching

Teaching Fellow — INFSCI 0201: Intermediate Programming with Python
University of Pittsburgh · Fall 2026
Instructor for an intermediate Python course covering object-oriented and functional programming, software-engineering practices, and program design. I design and deliver lectures, live coding demonstrations, programming exercises, and assessments.

Graduate Student Mentor — NOUR'25
University of Pittsburgh · Summer 2025
Mentored undergraduate researchers and developed research, academic, networking, and professional-development programming.