Rasika Muralidharan
@rasikamurali.bsky.social
PhD Student @ Indiana University Bloomington | Cooperative behaviors on social networks and agentic teams
Huge thanks to my collaborators Arthur D. Soto-Vásquez, Ph.D., Maria Montenegro, Ph.D., and Danny Valdez, Ph.D. for their support.
📄 Read the full paper: lnkd.in/gCauNwFS
💬 Questions or thoughts? Reach out via DM or email: rasimura@iu.edu
📄 Read the full paper: lnkd.in/gCauNwFS
💬 Questions or thoughts? Reach out via DM or email: rasimura@iu.edu
October 20, 2025 at 4:33 PM
Huge thanks to my collaborators Arthur D. Soto-Vásquez, Ph.D., Maria Montenegro, Ph.D., and Danny Valdez, Ph.D. for their support.
📄 Read the full paper: lnkd.in/gCauNwFS
💬 Questions or thoughts? Reach out via DM or email: rasimura@iu.edu
📄 Read the full paper: lnkd.in/gCauNwFS
💬 Questions or thoughts? Reach out via DM or email: rasimura@iu.edu
👍 Engagement patterns differ across languages, revealing distinct communication ecosystems. More qualitative research is needed to understand how health info spreads in Spanish-speaking communities and how local institutions shape public health messages online. #BreastCancer #HealthComms
October 20, 2025 at 4:32 PM
👍 Engagement patterns differ across languages, revealing distinct communication ecosystems. More qualitative research is needed to understand how health info spreads in Spanish-speaking communities and how local institutions shape public health messages online. #BreastCancer #HealthComms
🤷♂️ Spanish-language posts mainly come from local governments, media, and municipalities. Many promote at-home self-exams—no longer recommended in the U.S. Most engagement comes from health sites, politicians, and local reps, showing distinct voices shaping Spanish-language breast cancer discourse.
October 20, 2025 at 4:31 PM
🤷♂️ Spanish-language posts mainly come from local governments, media, and municipalities. Many promote at-home self-exams—no longer recommended in the U.S. Most engagement comes from health sites, politicians, and local reps, showing distinct voices shaping Spanish-language breast cancer discourse.
🧩 English-language posts are mostly from major non-profits like Susan G. Komen and the National Breast Cancer Alliance. These groups lead awareness efforts, driving engagement through strong branding and messages focused on screening, survivorship, and fundraising in the U.S. #BreastCancer
October 20, 2025 at 4:30 PM
🧩 English-language posts are mostly from major non-profits like Susan G. Komen and the National Breast Cancer Alliance. These groups lead awareness efforts, driving engagement through strong branding and messages focused on screening, survivorship, and fundraising in the U.S. #BreastCancer
🧠 Findings:
💭English and Spanish language content centered on similar topics—mammography, breast cancer awareness events, Pink Ribbon Month, and personal narratives.
💭English and Spanish language content centered on similar topics—mammography, breast cancer awareness events, Pink Ribbon Month, and personal narratives.
October 20, 2025 at 4:26 PM
🧠 Findings:
💭English and Spanish language content centered on similar topics—mammography, breast cancer awareness events, Pink Ribbon Month, and personal narratives.
💭English and Spanish language content centered on similar topics—mammography, breast cancer awareness events, Pink Ribbon Month, and personal narratives.
⚙️ We employed topic modeling to identify clusters of topics discussed in Facebook posts related to breast cancer. Metadata analysis was then used to determine the primary sources of content and the types of posts that received the most engagement.
October 20, 2025 at 4:24 PM
⚙️ We employed topic modeling to identify clusters of topics discussed in Facebook posts related to breast cancer. Metadata analysis was then used to determine the primary sources of content and the types of posts that received the most engagement.
Big shoutout to my advisors: @haewoon.bsky.social @jisunan.bsky.social for their guidance and support on this project!
📄 Paper link: arxiv.org/pdf/2510.07488
📄 Paper link: arxiv.org/pdf/2510.07488
October 13, 2025 at 2:11 PM
Big shoutout to my advisors: @haewoon.bsky.social @jisunan.bsky.social for their guidance and support on this project!
📄 Paper link: arxiv.org/pdf/2510.07488
📄 Paper link: arxiv.org/pdf/2510.07488
🔮 As LLMs evolve from individual tools to collaborative agents, understanding their team dynamics becomes essential.
Applying team science offers a powerful lens for designing AI systems that collaborate effectively, leverage diversity, and adapt like human teams do.
Applying team science offers a powerful lens for designing AI systems that collaborate effectively, leverage diversity, and adapt like human teams do.
October 13, 2025 at 2:10 PM
🔮 As LLMs evolve from individual tools to collaborative agents, understanding their team dynamics becomes essential.
Applying team science offers a powerful lens for designing AI systems that collaborate effectively, leverage diversity, and adapt like human teams do.
Applying team science offers a powerful lens for designing AI systems that collaborate effectively, leverage diversity, and adapt like human teams do.
🧠 Key findings
🧩 Agents tend to overestimate team performance before the task; post-task reflections surface misalignment and integration difficulties.
🧑⚖️ Evaluations via GPT-4o (“LLM-as-a-judge”) agree that flat teams score better in comprehension, coherence, reasoning, and confidence.
🧩 Agents tend to overestimate team performance before the task; post-task reflections surface misalignment and integration difficulties.
🧑⚖️ Evaluations via GPT-4o (“LLM-as-a-judge”) agree that flat teams score better in comprehension, coherence, reasoning, and confidence.
October 13, 2025 at 2:05 PM
🧠 Key findings
🧩 Agents tend to overestimate team performance before the task; post-task reflections surface misalignment and integration difficulties.
🧑⚖️ Evaluations via GPT-4o (“LLM-as-a-judge”) agree that flat teams score better in comprehension, coherence, reasoning, and confidence.
🧩 Agents tend to overestimate team performance before the task; post-task reflections surface misalignment and integration difficulties.
🧑⚖️ Evaluations via GPT-4o (“LLM-as-a-judge”) agree that flat teams score better in comprehension, coherence, reasoning, and confidence.
🧠 Key findings
⚖️ Flat (decentralized) teams often outperform hierarchical ones on reasoning tasks.
⚔️ Diversity is a double-edged sword: it can enrich reasoning in some settings, but also introduce coordination friction especially under hierarchical structure.
⚖️ Flat (decentralized) teams often outperform hierarchical ones on reasoning tasks.
⚔️ Diversity is a double-edged sword: it can enrich reasoning in some settings, but also introduce coordination friction especially under hierarchical structure.
October 13, 2025 at 2:04 PM
🧠 Key findings
⚖️ Flat (decentralized) teams often outperform hierarchical ones on reasoning tasks.
⚔️ Diversity is a double-edged sword: it can enrich reasoning in some settings, but also introduce coordination friction especially under hierarchical structure.
⚖️ Flat (decentralized) teams often outperform hierarchical ones on reasoning tasks.
⚔️ Diversity is a double-edged sword: it can enrich reasoning in some settings, but also introduce coordination friction especially under hierarchical structure.
We designed flat and hierarchical teams of LLM agents, assigning them personas (e.g. demographics) to inject controlled diversity. These teams were tested on reasoning and social reasoning tasks. We combined quantitative performance evaluation + qualitative analysis to analyze interaction dynamics
October 13, 2025 at 2:03 PM
We designed flat and hierarchical teams of LLM agents, assigning them personas (e.g. demographics) to inject controlled diversity. These teams were tested on reasoning and social reasoning tasks. We combined quantitative performance evaluation + qualitative analysis to analyze interaction dynamics