Reproduce “Human-Modeling in Sequential Decision-Making: An Analysis through the Lens of Human-Aware AI” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
Silvia Tulli, Stylianos Loukas Vasileiou, Sarath Sreedharan
This work establishes a clear definition and multi-dimensional framework for "human-aware AI" systems, applying it to analyze 46 recent papers in sequential decision-making. The analysis reveals current trends and identifies significant gaps in explicit human modeling, the integration of social science theories, and empirical validation through user studies.
Key insights
- The proposed human-aware AI framework proved robust for analyzing diverse human-AI interaction papers, effectively categorizing works even without explicit self-labeling.
- While implicit human assumptions are common, explicit modeling predominantly focuses on the human's knowledge state (M_H) regarding goals and preferences, with significant gaps in modeling the human's perception of the AI (M_R_h), inferential capabilities, and vocabulary.
- There is a notable underutilization of explicit social science theories and a scarcity of user studies for empirical validation across much of the reviewed human-AI interaction literature.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.