Michelle L. Ding
LightNews LightNews
banner
michelleding.bsky.social
Michelle L. Ding
@michelleding.bsky.social
99 followers 66 following 23 posts
organizer/researcher critically investigating how AI systems impact communities. cs phd @ brown cntr. she/her.

🌷 https://michelle-ding.github.io/
💭 https://michellelding.substack.com/
Posts Replies Media Videos
michelleding.bsky.social
Grey background with CNTR logo and black text that says: Testing LLMs in a sandbox isn’t responsible. Focusing on community uses and needs is. Michelle L. Ding, Jo Gasior-Kavishe, Victor Ojewale, and Suresh Venkatasubramanian Third Workshop on Socially Responsible Language Modelling Research (SoLaR) 2025 Full abstract available here bit.ly/solar-cntr Photos of authors in order of names above Brown Data Science Institute, Center for Technological Repsonsibility, Reimagination, and Redesign
October 11, 2025 at 4:12 PM
michelleding.bsky.social
More resources here: bit.ly/solar-cntr 🌱 Happy to chat with anyone interested in learning more!
Grey background with black text that says: Slide 1: What can you do as a researcher? Look into your local community - develop sustainable, long term partnerships with city-level or state-level coalitions, networks, news agencies, or non-profit organizations that do on the ground work to discover harms University researchers – look beyond your department into other disciplines in the humanities and social sciences to ground your research in interdisciplinary methods Industry researchers – look towards trust & safety teams, policy teams, human rights teams and establish collaborations where research reflects real uses. Or advocate to (re)establish these teams. Slide 2: Contribute to a crowdsourced, open source database: bit.ly/solar-cntr Case studies are a great way for us to share, learn, teach, and reflect on methods together. At SoLaR 2025, we hope to crowdsource a spreadsheet of community-driven LLM evaluation case studies that orient us towards contextualized, participatory, expertise driven evaluations. Please add any your own or related work to our spreadsheet :) And let’s stay connected! We also have a tab for contact info!
October 11, 2025 at 4:08 PM
michelleding.bsky.social
Brown background with CNTR logo with white text that says: Responsibility is not marginally improving LLMs to fit every use case. Responsibility is knowing when to use them and when not to.
October 11, 2025 at 4:04 PM
michelleding.bsky.social
Brown background with CNTR logo with white text that says: In LLM evaluations, we - researchers - are not the experts. We are the facilitators… …who enable communities to evaluate LLMs for themselves When we allow their expertise and lived experiences to guide our methodology… We will organically develop more meaningful scientific insights… And escape the vicious cycle of abstract benchmarks that only serve to misdirect and confuse us.
October 11, 2025 at 4:02 PM
michelleding.bsky.social
Thank you everyone who came to our talk! Here are some highlights:
Brown background and CNTR logo with white text that says: Community-driven evaluations yield good scientific insights. They strip away marketing, hype, and unnecessary abstractions And reveal the contexts where LLMs might genuinely add value and where they do not. This opens space for meaningful dialogue between stakeholders across academia, industry, government, and civil society And empowers the public to engage with and assess for themselves whether the tools they are being presented with can be used responsibly.
October 11, 2025 at 4:01 PM