Please join us on Friday, February 27, 2026, from 12:30 to 1:30 p.m. for the in-person ISP Forum in SENSQ 5317. We will be featuring Andrew Aquilina and Li Zhang.   Lunch and refreshments will be provided, starting at noon.

Andrew Aquilina Presentation Information

 Whose Standard of Distress? Community Judgments and LLM Alignment on Well-Being Posts

Abstract: Mental health support tools increasingly use large language models (LLMs) to detect signals of psychological distress. For these tools to be accurate, equitable, and acceptable to users, the standards of distress they apply and their underlying rationales should be grounded in the perspectives of the people whose words are being analyzed. When standards instead reflect outsider assumptions, models can miss or misread cues and thereby amplify existing inequities. In this study, we investigate how perceptions of psychological distress differ across online communities, and how LLMs can be aligned to these community-specific norms. Using Reddit data from six identity-based subreddits, we develop a community-aware dataset and codebook for labelling distress and support-seeking behaviors. Both in-group and out-group raters annotate each post, allowing comparison of whose judgements LLMs reproduce. We then evaluate whether models better align with in-group consensus and whether prompting with community context improves sensitivity without stereotyping. Our findings highlight significant variation in how distress is expressed and recognized, underscoring how a single "universal" model of distress may not translate equally across groups. The work contributes a pluralistic evaluation framework and dataset for mental-health-related AI, emphasizing how equitable systems must reflect community-grounded standards in sensitive domains such as mental-health.

Bio: Andrew Aquilina is a second-year PhD student in Information Science at the University of Pittsburgh, working under the guidance of Prof. Yu-Ru Lin within the Computational Social Dynamics Lab. His research is centered on studying how human-centered language technologies can accommodate diverse individual preferences in a safe and transparent way, especially within sensitive contexts. He previously earned a master's in AI from Stockholm University and a bachelor's in AI from the University of Malta.

Li Zhang Presentation Information

Thinking Longer, Not Always Smarter: Evaluating LLM Capabilities in Hierarchical Legal Reasoning

Abstract: Case-based reasoning is a cornerstone of U.S. legal practice, requiring professionals to argue about a current case by drawing analogies to and distinguishing from past precedents. While Large Language Models (LLMs) have shown remarkable capabilities, their proficiency in this complex, nuanced form of reasoning needs further investigation. We propose a formal framework that decomposes the process of identifying significant distinctions between cases into three-stage reasoning tasks. Our framework models cases using factual predicates called factors, organizes them into a legal knowledge hierarchy, and defines verifiable rules for identifying distinctions, analyzing their argumentative support, and evaluating their significance. Through comprehensive evaluation of modern reasoning LLMs, we reveal a paradox: while models achieve high accuracy on surface-level reasoning (Task 1), performance degrades on hierarchical reasoning (Task 2: 64.82%-92.09%) and collapses on integrated analysis (Task 3: 11.46%-33.99%). Most strikingly, we find that models consistently expend more computational resources on incorrect responses than correct ones, suggesting that "thinking longer" does not always mean "thinking smarter." Our work provides a methodology for fine-grained analysis of LLM reasoning capabilities in complex domains and reveals fundamental limitations that must be addressed for robust and trustworthy legal AI.

Bio: Li Zhang is a second-year Ph.D. student in Intelligent Systems at the University of Pittsburgh, where he works with Professor Kevin Ashley. His research sits at the intersection of Artificial Intelligence and Law, focusing on evaluating and enhancing the reasoning capabilities of Large Language Models (LLMs) for legal tasks. He is particularly interested in trustworthy legal AI. Li’s work appears in top AI & Law venues such as CS&Law, ICAIL, and JURIX, where his paper on LLMs and overruled precedents recently won the Best Student Paper award (2025). Before his doctoral studies, he obtained a Juris Master from Tsinghua University and practiced as a Legal Counselor. In this talk, based on his upcoming paper at CS&Law 2026, Li will present "Thinking Longer, Not Always Smarter: Evaluating LLM Capabilities in Hierarchical Legal Reasoning." He will discuss a new framework for auditing how LLMs handle complex legal distinctions and share findings on the paradox where increased "reasoning" time in models does not always lead to smarter legal conclusions.