StARS: Socially Appropriate Robot Actions via a Recommender System-Driven Approach

Researchers introduced StARS, a model-agnostic framework for generating socially appropriate robot actions in human-robot interaction.

StARS uses recommender-system style preference modeling to predict user-specific appropriateness scores for robot behaviors.

In tests on MannersDB+ and SocNav1, the approach improved agreement with human annotators.

The results suggest social norms in robotics are subjective, and personalization can help robots choose better actions across users and settings.