# Generative AI for climate governance and acceptability-constrained policy design

## TL;DR

This paper introduces a new approach to climate policy design that uses AI to simulate public responses and ensure policies align with cultural values and social norms.

## Contribution

The novelty lies in using large language models as 'cultural world models' to enhance policy acceptability alongside climate efficacy.

## Key findings

- LLMs can simulate public responses to climate policies, improving their social legitimacy.
- Embedding LLMs in agent-based models allows co-optimization of policy efficacy and acceptability.
- The approach highlights limitations in LLM representation and transparency.

## Abstract

Climate policies often fail when they clash with cultural values, social identities, and fairness perceptions. We propose Acceptability-Constrained Climate Policy Design (ACCPD), using large language models as “cultural world models” to simulate public responses before implementation. By embedding LLMs in generative agent-based models and physical system simulators, ACCPD aims to enable policymakers to co-optimize for climate-policy efficacy and social legitimacy. We discuss methodological limitations regarding representation and LLM opacity.