Human Choice Prediction in Language-based Persuasion Games: Simulation-based Off-Policy Evaluation

View Full Paper

Ask AI

Bookmark

The proposed simulation technique enhances the prediction accuracy of human decisions by 7.1%.
A dataset of 87K decisions was collected from human interactions with artificial agents in persuasion games.
The off-policy evaluation framework uses a prediction model trained on data from one expert set to evaluate another.
This approach suggests that interactions across the agent space significantly boost performance in challenging case predictions.

Discussion

No takes yet. Share an insight, caveat, or question.

Simulation techniques improve off-policy evaluation performance in predicting human decisions in persuasion games.

Shapira et al. (2025) studied this question.

View Full Paper

Ask AI

Bookmark