Conditioning language models on social values improves human-like choices
Conditioning LLMs on Social Value Orientation improves behavioural alignment in a sequential social dilemma
Computer Science and Game Theory
Summary
Large language models often struggle to mimic the variety of human decision-making in social situations. The authors found that by telling these models about a person's social values—how much they care about themselves versus others—they can make the models behave more like real people in a game setting. This approach made the models' choices much closer to actual human behavior, especially showing how more caring individuals tend to cooperate more. The results suggest the models don’t need fixed social feelings but can match social values to actions through prompts.
What this means in practice
- •For game designers: Generate more realistic player simulations by conditioning AI agents on social value profiles to better reflect human strategic diversity.
- •For social robotics teams: Create robots that adapt their social decisions based on modeled human social preferences, improving interaction quality.
Authors
Marco Saponara, Axel Abels, Ann Nowé, Tom Lenaerts
Abstract
Large language Models (LLMs) are increasingly used to simulate human decision-making, yet their outputs often under-represent human behavioural heterogeneity. We investigate whether conditioning LLMs on Social Value Orientation (SVO, a measure of how individuals value their own outcomes relative to others') can better reproduce human behaviour in a sequential social dilemma. Using experimental data from two variants of the Centipede Game (CG) as reference, we compare the default behaviour of eight LLMs with behaviour generated after conditioning them on SVO profiles drawn from the human sample. We find that the default strategies vary substantially across models, but are generally distant from the reference distributions. However, conditioning them on SVO profiles systematically steers their strategic behaviour, improving alignment with the human reference by up to 70% with respect to the default. Across models, higher induced SVO values decrease the probability of stopping the game, reproducing the relationship between prosociality and cooperation observed in human behaviour. Together with the sensitivity of LLMs' elicited SVO to prompt and order effects, these results suggest that the usefulness of SVO for behavioural simulation does not depend on LLMs possessing stable social preferences, but rather on their ability to map social preferences onto corresponding strategic choices.