Do Large Language Models Encode Institutional Experience? Evidence from Cross-Linguistic Moral Reasoning Under Ambiguity

1. 1Test the hypothesis that languages encode institutional environments using cross-linguistic analysis
2. 2Run experiments across multiple languages and LLMs to identify systematic differences in moral reasoning
3. 3Configure LLMs to handle ambiguity in moral reasoning tasks
4. 4Apply institutional theory to interpret results and understand the source of variation in LLM moral reasoning
5. 5Build a framework to evaluate the impact of institutional experience on LLM decision-making

AI engineers and data scientists benefit from understanding how LLMs inherit institution-specific moral priors, which can impact model performance and fairness

💡 LLMs can encode aspects of institutional environments, leading to systematic differences in moral reasoning across languages

🤖 LLMs may inherit institution-specific moral priors through training, impacting decision-making #AI #LLMs