Summary of Explaining Learned Reward Functions with Counterfactual Trajectories, by Jan Wehner et al.
Explaining Learned Reward Functions with Counterfactual Trajectories
by Jan Wehner, Frans Oliehoek, Luciano Cavalcante Siebert
First submitted to arxiv on: 7 Feb 2024
Categories
- Main: Artificial Intelligence (cs.AI)
- Secondary: Machine Learning (cs.LG)
GrooveSquid.com Paper Summaries
GrooveSquid.com’s goal is to make artificial intelligence research accessible by summarizing AI papers in simpler terms. Each summary below covers the same AI paper, written at different levels of difficulty. The medium difficulty and low difficulty versions are original summaries written by GrooveSquid.com, while the high difficulty version is the paper’s original abstract. Feel free to learn from the version that suits you best!
| Summary difficulty | Written by | Summary |
|---|---|---|
| High | Paper authors | High Difficulty Summary Read the original abstract here |
| Medium | GrooveSquid.com (original content) | Medium Difficulty Summary The proposed Counterfactual Trajectory Explanations (CTEs) aim to interpret reward functions in reinforcement learning by contrasting an original with a counterfactual partial trajectory and their respective rewards. The method optimizes six quality criteria using a Monte-Carlo-based algorithm, generating CTEs that increase the similarity between predictions and the reward function on unseen trajectories. This approach demonstrates informative explanations for a proxy-human model, enabling accurate judgment of reward differences and generalization to out-of-distribution examples. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary This paper is about making artificial intelligence (AI) systems better understand what they’re supposed to do. Right now, AI systems can learn from human behavior or feedback, but it’s hard to know if they’re doing the right thing. The researchers propose a new way to explain why an AI system chose a certain action. They call this “Counterfactual Trajectory Explanations” (CTEs). CTEs help us understand how the AI system learned its rewards and what makes them good or bad. This can help us make better decisions about what tasks we should give to AI systems. |
Keywords
* Artificial intelligence * Generalization * Reinforcement learning
