Publications

• 2025: Can Lessons From Human Teams Be Applied to Multi-Agent Systems?The Role of Structure, Diversity, and Interaction Dynamics

  Under Review in ARR May 2025 Cycle
  Multi-Agent Systems (MAS) with Large Language Model (LLM)-powered agents are gaining attention, yet fewer studies explore their team dynamics. Inspired by human team science, we propose a multi-agent framework to examine core aspects of team science: structure (flat vs. hierarchical teams), diversity (via demographic personas), and interaction dynamics (through pre-/post-task interviews and GPT-4o-based conversation analysis). We evaluate team performance across four tasks: CommonsenseQA, StrategyQA, Social IQa, and Latent Implicit Hate, spanning commonsense and social reasoning. Our results show that flat teams tend to perform better than hierarchical ones, while diversity has a nuanced impact. Interviews suggest agents are overconfident about their team performance, yet post-task reflections reveal both appreciation for collaboration and challenges in integration. GPT-4o analysis highlights limited conversational coordination among agents.

• 2025: Federating Governance: Scaling Community Rules on Mastodon

  Under Review at CSCW 2026
  The rise of decentralized social media platforms like Mastodon and Bluesky is highlighting the challenge of scaling self-governance and moderation. As communities grow, they face new issues that demand increasingly complex governance structures. Because moderation in these spaces is typically volunteer-driven, there is limited formal guidance on how community rules and moderation practices should evolve with growth. This study investigates how formalized rules scale with Mastodon instances by analyzing community rules across servers of varying sizes. We categorize these rules to identify key governance priorities, finding that smaller communities focus on narrower sets of topics, while larger servers maintain a more balanced coverage of a broad range of topics. In line with previous research, we find that topics about problematic content, such as Harassment,''Hate Speech,’’ and ``Illegal Content’’ consistently dominate moderation rules regardless of server size, although their relative importance may decrease with server size. Our analysis of rule formalization reveals that as community size increases, rule scope expands, while linguistic diversity and readability decrease. Finally, we examine the role of internal and external factors on rule development. Our regression analysis shows that internal factors are strong predictors: community size predicts every aspect of rule formalization, and community maturity becomes a significant predictor alongside its size. By contrast, the number of cross-server interactions (federation) accounts only for the expansion of rule scope.

• 2024: Profiled with Purpose: LLMs’ Role in User Profiles for Personalized Conversations

  Semi-structured interviews are commonly used in social sciences that combine a protocol with the flexibility to follow topical trajectories as the conversation unfolds, allowing discovery, exploration, and meaning-making considering the lived experiences of the person being interviewed The project draws inspiration from the Japanese concept of ``ikigai’’ or meaning and purpose in life \cite{garcia2017ikigai} that can be derived from personal, interpersonal, and community sources, such as hobbies, family, and volunteering \cite{randall2022finding}. Photographs have been used as a common tool to initiate conversations and elicit memories that can connect to older adults’ relationships, family, and activities that bring them meaning in their lives. Expanding on our study to develop an ikigai-based activity. where older adults discuss personal photographs through conversations that bring out the special moments in their lives along with their interests, life stories and events, close relations, and even their ongoing experiences.

• 2024: Let’s talk about you: Development and evaluation of an autonomous robot to support Ikigai reflection in older adults

  The sources of a person’s \emph{ikigai}—their sense of meaning and purpose in life—often change as they age. Reflecting on past and new sources of ikigai may help people renew their sense of meaning as their life circumstances shift. Building on insights from an initial Wizard-of-Oz robot prototype \cite{randall2023realizing}, we describe the design of an autonomous robot that uses a semi-structured conversation format to help older adults reflect on what gives their life meaning and purpose. The robot uses both pre-determined (scripted) and Large Language Model (LLM) generated questions to personalize conversations with older adults around themes of social interaction, planning, accomplishments, goal setting, and the recent past. We evaluated the autonomous robot with 19 older adult participants in a lab setting and at two eldercare facilities. Analysis of the older adults’ conversations with the robot and their responses to an evaluative survey allowed us to identify several design considerations for an autonomous robot that can support ikigai reflection. Interweaving simple yet detailed pre-determined questions with LLM-generated follow-up questions yielded enjoyable, in-depth conversations with older adults. We also recognized the need for the robot to be able to offer relevant suggestions when participants cannot recall events and people they find meaningful. These findings aim to further refine the design of an interactive robot that can support users in their exploration of life’s purpose.

• 2024: Digital Epidemiology of Prescription Drug References on X (Formerly Twitter): Neural Network Topic Modeling and Sentiment Analysis

  This study sought to use next-generation techniques (beyond LDA) from natural language processing to reprocess X data and automatically cluster groups of tweets into topics to differentiate between street- and brand-name data sets. We also aimed to analyze the differences in emotional valence between the 2 data sets to study the relationship between engagement on social media and sentiment.