Ai learns to be more creative in scientific idea generation
Reinforcing Agentic Creativity in Scientific Ideation with Night Science
Artificial IntelligenceComputation and Language
Summary
Large language models often give safe, predictable answers, which limits their creativity in science. The authors show that creativity can be trained by teaching AI when to break from usual patterns in thinking. Their approach uses ideas from cognitive science to guide AI along three aspects of creativity: what to do, when to explore new ideas, and how novel and useful the ideas are. This makes the AI suggest more diverse and original scientific ideas, which could help researchers find new directions beyond usual methods. Simply changing randomness in model output was not enough; directing the model’s creative effort was key.
What this means in practice
- •For ai development teams: Improve AI-based scientific ideation tools by training models to produce more diverse research ideas using guided creativity metrics.
- •For innovation managers: Use AI that better explores unusual research directions to support brainstorming and early-stage product development.
Authors
Priyanka Kargupta, Silviu Cucerzan, Shweti Mahajan, Allen Herring, Jiawei Han, Ryen W. White, Sujay Kumar Jauhar
Abstract
Large language models (LLMs) excel at structured, verifiable tasks, but their low-entropy bias can produce homogeneous and predictable outputs, limiting their utility for open-ended scientific ideation. Effective discovery, however, spans a broader creative spectrum: from structured day science to loosely structured, serendipitous night science that reaches ideas beyond those typically considered. We introduce AI Night-Scientist, an agentic framework that uses reinforcement learning to teach models when and how to depart from predictable reasoning. Grounded in cognitive science, we model creativity along three axes: action (what to do and how creatively), process (when to explore versus exploit), and outcome (the novelty and usefulness of the resulting idea). We use these axes to train models with GRPO, exposing them to varying degrees and forms of creativity throughout training. This produces substantially more diverse scientific proposals, expanding the range of research directions by 27.8% and contribution types by 14.9% over the base model. It also improves predicted citation impact by up to 32.0 percentage points and originality by 66.2 points. These gains cannot be reproduced by simply increasing decoding temperature; instead, we find that semantic guidance specifying what kind of creativity to pursue is critical. Overall, our results suggest that creativity is a learnable, multi-level ability that can be shaped to help researchers reach ideas beyond those typically explored by LLMs.