arXiv:2606.17657v1 Announce Type: new Abstract: People make decisions differently in strategic interactions. Some update beliefs like a Bayesian; others exhibit biases like motivated reasoning. Although creators of large language models use simulated humans for safety evaluations and training, they often fail to cover this breadth of human behavior. We argue that cognitive science and economics provide a convenient tool for doing so, making use of mathematical models of human decision-making. We propose an approach that we call Equation-to-Behavior Prompting for guiding large language models t
The increasing sophistication and widespread use of large language models for simulation and decision-making necessitates more realistic human behavior modeling to ensure safety and effectiveness.
Improving how AI models simulate human behavior, particularly biases and varying decision-making strategies, is crucial for developing safe, robust, and ethically aligned AI systems used in strategic interactions.
The proposed 'Equation-to-Behavior Prompting' offers a new methodological framework for guiding LLMs to exhibit a broader and more accurate spectrum of human cognitive processes, moving beyond simplistic simulation.
- · AI Safety Researchers
- · Cognitive Scientists
- · Economists
- · Developers of AI agents
- · Creators of simplistic LLM simulations
- · Users relying on biased AI evaluations
More sophisticated and nuanced AI simulations of human interaction will become possible, improving model evaluation and training.
This could lead to AI systems that are better at predicting and navigating complex human social and strategic environments, from negotiation to policy impact assessment.
Enhanced human-like AI decision-making could accelerate the integration of autonomous agents into roles requiring deep understanding of human psychology, potentially disrupting white-collar work further.
This signal links to a primary source. Continuum Brief monitors and indexes it as part of the live intelligence stream — we do not republish source content.
Read at arXiv cs.AI