Summary of Interpretable Concept Bottlenecks to Align Reinforcement Learning Agents, by Quentin Delfosse et al.
Interpretable Concept Bottlenecks to Align Reinforcement Learning Agents
by Quentin Delfosse, Sebastian Sztwiertnia, Mark Rothermel, Wolfgang Stammer, Kristian Kersting
First submitted to arxiv on: 11 Jan 2024
Categories
- Main: Machine Learning (cs.LG)
- Secondary: Symbolic Computation (cs.SC)
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 paper introduces Successive Concept Bottleneck Agents (SCoBots), a novel approach to deep reinforcement learning (RL) that addresses common issues such as goal misalignment, reward sparsity, and difficult credit assignment. SCoBots integrate consecutive concept bottleneck layers, representing concepts as relations between objects, crucial for many RL tasks. The results demonstrate competitive performances and the potential for domain experts to understand and regularize SCoBots’ behavior. This leads to more human-aligned RL agents. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary Deep learning is helping robots learn how to play games and make decisions, but it’s not always easy for people to understand why the robot made a certain choice. A team of researchers has created a new way to help these robots learn by showing them how different things are related. They call this approach Successive Concept Bottleneck Agents or SCoBots. These agents can learn to play games and make decisions just as well as other methods, but they also give people insight into why the robot made certain choices. |
Keywords
* Artificial intelligence * Deep learning * Reinforcement learning
