Personalized AI Co-Scientists Tailor Research to Individual Researchers.

The paper proposes a paradigm shift in AI co-scientist development: personalized auto-research. Current AI systems designed to assist with scientific inquiry often operate generically, optimizing for universal metrics like novelty or validity without considering the specific researcher. This overlooks the fundamental truth that research is deeply personal, influenced by an individual's unique background, expertise, collaborators, and research community. The authors argue that true AI co-scientists must integrate this individual context to be genuinely valuable. To achieve this, the research outlines a flexible framework that embeds a "graph-grounded researcher context" throughout the entire research pipeline, from hypothesis generation and literature retrieval to experimentation, writing, and peer review. This personalization is not merely a convenience but a core property enabling AI to act as a true co-scientist rather than just a tool. The framework emphasizes three components: graph-grounded researcher representations, personalization across all research stages, and individual-centric evaluation. The paper highlights a "one-size-fits-all" failure mode where generic AI outputs stifle novel ideas by ignoring tacit knowledge, and discusses open challenges in realizing this vision.