Reproduce “CogniPair: From LLM Chatbots to Conscious AI Agents -- GNWT-Based Multi-Agent Digital Twins for Social Pairing -- Dating & Hiring Applications” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
Wanghao Ye, Sihan Chen, Yiting Wang, Shwai He, Bowei Tian, Guoheng Sun, Ziyi Wang, Ziyao Wang and 13 more
Researchers at the University of Maryland and collaborators developed CogniPair, a social-influence decision system featuring GNWT-Agents that computationally implement Global Workspace Theory. This framework enables AI agents to simulate evolving human psychological and social behaviors, achieving 77.8% match prediction accuracy in speed dating scenarios and demonstrating high behavioral fidelity for "digital twin" agents in human validation studies.
Key insights
- The computational operationalization of Global Workspace Theory enables AI agents to develop dynamic internal psychological states and adaptive social behaviors.
- A parallel cognitive architecture with specialized modules, integrated through a global workspace, facilitates genuine personality evolution and coherent, context-aware responses.
- Simulating social dynamics through agent interactions can lead to adaptive changes in preferences and self-perception that mirror observed human behavioral shifts.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.