Hi! I'm Divya Ramesh.

As a human-centered AI and design researcher, I take a holistic approach to designing AI systems that enhance, rather than diminish the agency of underserved communities. To do this, I combine insights from human-computer interaction (HCI) and science and technology studies (STS), iteratively generating designs and socially-engaged critiques of human-AI interactions. I use a wide-range of computational and interpretivist social science methods in my work.

I expect to graduate with a PhD in Computer Science and Engineering and a certificate in Science, Technology, and Public Policy from the University of Michigan, Ann Arbor in 2025. I hold an MS in EE from the University of Southern California, and a BE in ECE from Ramaiah Institute of Technology, India. Between 2014-18, I was a deep learning and computer vision engineer at CloudSight Inc., where I architected the company's human-AI interaction pipeline for real-time image recognition and authored a US patent for it. I have also had short research stints at Microsoft Research and Google Research.

I am on the academic and industry research job market for 2024-25!


Designing Human-AI Interactions: Accountability & Agency of Underserved Communities

My research straddles dual worlds of human-computer interaction (HCI) and science and technology studies (STS), of design and critique, and of computational and interpretivist social science methods. I generate and evaluate human-AI interaction designs using experimental techniques including crowdsourcing, reinforcement learning from human feedback (RLHF), and wizard-of-oz studies. To ensure ecological validity, I ground evaluations in lived experiences of marginalized communities using interpretive qualitative studies including interviews, and content and discourse analyses.

During my PhD, I drew on theories of HCI and cognitive psychology to help people exercise their agency during human-AI interactions, such as by identifying complementary performance characteristics of AI algorithms that helped people appropriately rely on AI. Building on insights from STS, I generated socially-engaged design critiques, by understanding how and why AI interactions diminish people’s agency in everyday contexts. Drawing on thick descriptions of low-income individuals’ everyday AI encounters in India and the US, my work establishes that enhancing human agency during AI interactions is a challenge of both appropriate reliance and stakeholder power relations, with implications for accountability.

My on-going and future work reframes the goal of human-AI interactions as designing for accountability and agency simultaneously. By situating human-AI interaction designs and critiques in social, historical, and cultural contexts of use, and by leveraging a praxis-based approach, my research aims to build an accountable AI ecosystem that enhances the agency of underserved communities.

Featured Publications

Divya Ramesh, Vaishnav Kameswaran, Ding Wang, Nithya Sambasivan. "How Platform-User Power Relations Shape Algorithmic Accountability: A Case Study of Instant Loan Platforms and Financially Stressed Users in India." In Proceedings of 2022 ACM Conference on Fairness, Accountability and Transparency (FAccT 2022). Seoul, South Korea.

[Preprint] [Techcrunch Perceptron]

Divya Ramesh, Caitlin Henning, Nel Escher, Haiyi Zhu, Min Kyung Lee, Nikola Banovic. "Ludification as a Lens for Algorithmic Management: A Case Study of Gig-Workers’ Experiences of Ambiguity in Instacart Work." In Designing Interactive Systems Conference (DIS ’23), July 10–14, 2023, Pittsburgh, PA, USA.

[Preprint]

Madison Cutler, Erin Keith, Kleinman Molly, Navarrete Ky, Kaci Pellar, Divya Ramesh, Eric Welsby. "Community Partnerships Playbook: How to Create Equitable Partnerships between Technical and Community Experts." A publication of the University of Michigan Science, Technology, and Public Policy program; Detroit Disability Power; the Detroit Justice Center; and We the People Michigan, February 2024, Ann Arbor, MI, USA. [Playbook]

*Alphabetical ordering of authors

Divya Ramesh, Anthony Liu, Jean Song, Andres Echeverria, Nicholas Waytowich, Walter Lasecki."Yesterday's Reward is Today's Punishment: Contrast Effects in Human Feedback to Reinforcement Learning Agents." In Proceedings of the 19th International Conference on Autonomous Agents and Multiagent Systems (AAMAS 2020). Auckland, New Zealand. [Preprint] [Talk] [Slides]

Pragnesh Jay Modi Best Student Paper Award

This is a featured list. For some of my other work, please visit my Google Scholar page.

Research Impact

My research has appeared in premier human-computer interaction, AI, and AI ethics venues such as ACM CHI, FAccT, DIS, TOCHI and AAMAS, where I have received a Best Paper nomination, and a Pragnesh Jay Modi Best Student Paper award. My first-author work with Google Research has had significant social impact through appearance in media outlets, directly informing Google’s research direction on culturally-inclusive AI and shaping policy. This work ranks in the top 5% of all research outputs scored by Altmetric Explorer for its impact. I have received recognition through the Quad, Barbour, and CSST Fellowships. I have also mobilized my research findings in externally invited talks, workshops, and panels. Most notably, in June 2023, I was invited to discuss the role of AI and emerging technologies in India-US relationships with President Biden’s Chief Foreign Policy officer and National Security Council Chief of Staff, Curtis Ried. My work has received generous support from the Schmidt Futures Foundation, Rackham Graduate School and Google Research.