Exploring the opportunities and challenges of using large language models to represent institutional agency in land system modelling DOI Creative Commons
Yongchao Zeng, Calum Brown, Joanna Raymond

et al.

Earth System Dynamics, Journal Year: 2025, Volume and Issue: 16(2), P. 423 - 449

Published: March 13, 2025

Abstract. Public policy institutions play crucial roles in the land system, but modelling their policy-making processes is challenging. Large language models (LLMs) offer a novel approach to simulating many different types of human decision-making, including choices. This paper aims investigate opportunities and challenges that LLMs bring system by integrating LLM-powered institutional agents within an agent-based use model. Four LLM are examined, all which, examples presented here, taxes steer meat production toward target level. The provide simulated reasoning action output. agents' performance benchmarked against two baseline scenarios: one without interventions another implementing optimal actions determined through genetic algorithm. findings show that, while perform better than non-intervention scenario, they fall short achieved actions. However, demonstrate behaviour marked consistency transparent reasoning. includes generating strategies such as incrementalism, delayed action, proactive adjustments, balancing multiple stakeholder interests. Agents equipped with experiential learning capabilities excel achieving objectives progressive order which proposed output has notable effect on performance, suggesting enforced both guides explains decisions. here points promising significant challenges. include, exploring naturalistic handling massive documents, human–AI cooperation. Challenges mainly lie scalability, interpretability, reliability LLMs.

Language: Английский

Exploring the opportunities and challenges of using large language models to represent institutional agency in land system modelling DOI Creative Commons
Yongchao Zeng, Calum Brown, Joanna Raymond

et al.

Earth System Dynamics, Journal Year: 2025, Volume and Issue: 16(2), P. 423 - 449

Published: March 13, 2025

Abstract. Public policy institutions play crucial roles in the land system, but modelling their policy-making processes is challenging. Large language models (LLMs) offer a novel approach to simulating many different types of human decision-making, including choices. This paper aims investigate opportunities and challenges that LLMs bring system by integrating LLM-powered institutional agents within an agent-based use model. Four LLM are examined, all which, examples presented here, taxes steer meat production toward target level. The provide simulated reasoning action output. agents' performance benchmarked against two baseline scenarios: one without interventions another implementing optimal actions determined through genetic algorithm. findings show that, while perform better than non-intervention scenario, they fall short achieved actions. However, demonstrate behaviour marked consistency transparent reasoning. includes generating strategies such as incrementalism, delayed action, proactive adjustments, balancing multiple stakeholder interests. Agents equipped with experiential learning capabilities excel achieving objectives progressive order which proposed output has notable effect on performance, suggesting enforced both guides explains decisions. here points promising significant challenges. include, exploring naturalistic handling massive documents, human–AI cooperation. Challenges mainly lie scalability, interpretability, reliability LLMs.

Language: Английский

Citations

0