Scroll →
At the University of Chicago, we're exploring new forms of socio-cognitive AI that complements the individual and collective human intelligence with which the world is familiar.
Intelligence is not a linear spectrum, but a multidimensional landscape where diverse forms of thinking coexist, whether among humans, animals, or artificial agents.
Our initiative places social and cognitive dimensions at the core of AI, moving beyond the quest to automate human tasks or build superintelligent machines.
We are discovering new forms of socio-cognitive AI: systems capable of genuine collective thinking, social reasoning, and collaborative discovery.
This project tests whether LLMs can generate end-to-end scientific discoveries—from forming hypotheses to collecting and interpreting new data—both alone and as coordinated “AI lab” teams.
The team will build “explainable AI through modularity” so AI systems can make interpretable, feature-aware decisions and audit AI-driven hypothesis generation and automated experimentation.
This project strengthens LLMs as scientific collaborators by improving their hypothesis generation, critique, and teamwork via automated research platforms and pipelines.
Using Prophet Arena, the project develops theory and back-tests strategies for converting AI forecasts plus market consensus into risk-aware, profit-maximizing decisions.
The team investigates why LLMs struggle to notice missing information and tests whether targeted interventions can fix this “absence blindness” or if it is architectural.
This research builds a mechanistic account of how people attribute beliefs and intentions to robots, explaining when and why “mind” is inferred from robot behavior.
The project characterizes the non-Euclidean geometry and topology of LLM representation spaces to improve interpretability, diagnostics, and activation-level interventions.
This study tests how different AI writing/support conditions affect short- and long-term human memory, identifying designs that enhance learning rather than diminish it.
The project compares how humans and LLMs encode “surprise” in narratives to reverse-engineer event expectation building and updating in people.
This project creates embodied AI that teaches by guiding learners’ movements—using gesture as a mechanism for reasoning and conceptual change.
The team models how hierarchy can emerge endogenously in multi-agent systems under uncertainty and communication costs, and how that structure shapes collective exploration vs. exploitation.
Project SENSUS explores how AI/robots can develop new sensory capacities, drawing on evolutionary biology, cognitive science, and observational practices.
The project tests whether real-time AI writing interventions can reduce gender bias in hiring by shifting evaluators’ language and underlying judgments during decision-making.
This project studies and prototypes AI creativity that goes beyond remixing by modeling how social learning and network structure enable exploration into genuinely novel conceptual space.
This project maps and manipulates the personas LLMs can adopt to understand how perspective shapes representations, concepts, and the tendency toward uniform “persona collapse.”
Collaborative AI is at the heart of our vision—a new framework for artificial intelligence systems that think, reason, and create alongside humans.
Rather than seeking to replicate individual human intelligence, we focus on developing AI that can actively participate in the social and cognitive dynamics of groups, institutions, and collective intellectual enterprises. Our research explores:
Connect with Socio Cognitive AI.
We welcome inquiries, partnerships, and new ideas from researchers, students, and organizations redefining intelligence.
Whether you're interested in joining our team, sharing your perspective, proposing a project, or learning more about our work, your message is the start of something novel.