Research
Five questions the lab is working on
Our work runs from the mathematics of ranking objectives to the social consequences of deploying them. These five themes are the taxonomy the whole site uses — every publication and every lab member is tagged into it.
-
Responsible Information Access
When a ranking model learns from human data, whose assumptions does it carry forward — and can we take them back out?
Measuring and mitigating the social harms that search and recommendation systems reproduce.
-
Robustness & Adversarial IR
If a search ranking can be moved by an adversary, what is a retrieval score actually worth?
How neural rankers can be attacked, why they fail, and what it takes to defend them.
-
AI & Scientific Integrity
Peer review is the quality control of science. What happens when the reviewer is a language model?
Whether large language models can be trusted inside the peer review process — and how we would know.
-
Neural Ranking & Retrieval Models
Most retrieval systems optimize a surrogate for the metric they are judged on. What if they did not have to?
The core machinery — ranking objectives, query reformulation, and generative approaches to retrieval.
-
Information, People & Society
Information systems are studied as engineering. What do they look like when studied as social infrastructure?
What information systems do to the communities that depend on them — immigrants, workers, courts.