Up next
New events are in the works.
We're planning the next season of BLISS panels and workshops. Join the mailing list and you'll hear first.
Join the mailing listBoston LLMs Initiative for Social Sciences
Language models,
taken seriously as
research instruments.
BLISS is a student-led community of Boston-area researchers who learn, apply, and critically evaluate large language models — and question them like any other method.
Events
Past event · Fall panel · November 14, 2025
Computational Social Sciences in the Age of Generative AI
Our inaugural panel brought four leading researchers to one table to ask what generative AI is doing to the practice of social science — and what social science should do about it.
- David LazerDistinguished Professor, Political Science & Computer Sciences
- Tina Eliassi-RadJoseph E. Aoun Professor, Computer Sciences
- Nick BeauchampAssociate Professor, Political Science & Network Science
- Beatrice MagistroAssistant Professor, AI Governance
Moderated by Emre Tapan & Burak Ozturan · Renaissance Park 909, Northeastern University
Write-up coming soonThe program so far
-
Fall 2025
Workshops & panel
Two afternoon workshops — LLMs by Hand, working through model mechanics with paper and pencil, and LLMs in Action, hands-on with Hugging Face and APIs — alongside our inaugural panel, above.
-
Spring 2026
Biweekly reading group
BLISS became a standing conversation: a biweekly reading group discussing new work on LLMs and social science.
-
Summer 2026
Quiet season
On break — and planning what comes next. The mailing list hears first.
The Book
A book being written in the open — on reading the machines we study.
The Anthropology of Machines: Mechanistic Interpretability for Social Scientists is a draft that grows chapter by chapter, with comments open on every page. Social scientists have always measured what they cannot see directly; language models are a new object for that same effort. The book works through three ways of studying one:
- Observe — descriptive inference about a model's internals: probes, the logit lens, sparse autoencoders
- Intervene — causal inference: activation patching, ablation, circuits
- Validate — knowing when a result is real: benchmarks, metrics, reliability
Read the working draft
Free · comments open · new chapters as they're written
People
About
BLISS exists because language models arrived in social science faster than the training did. We bring Boston-area researchers together — across disciplines and institutions — to build practical skill and critical judgment side by side: how these models work, where they fail, and what responsible use looks like in real research.
Workshops
Hands-on training, by hand first — from paper-and-pencil model mechanics to Hugging Face and APIs.
Reading group
A biweekly conversation on new work at the intersection of LLMs and social science.
Panels & talks
Public conversations with leading researchers on what generative AI means for the field.
Open materials
What we teach, shared in the open — workshop materials and the organizers' books-in-progress.
Stay in the loop.
One email when there's something worth knowing — a new event, a new chapter. No noise.