Summary of Formal Ethical Obligations in Reinforcement Learning Agents: Verification and Policy Updates, by Colin Shea-blymyer et al.
Formal Ethical Obligations in Reinforcement Learning Agents: Verification and Policy Updates
by Colin Shea-Blymyer, Houssam Abbas
First submitted to arxiv on: 31 Jul 2024
Categories
- Main: Artificial Intelligence (cs.AI)
- Secondary: Logic in Computer Science (cs.LO)
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 Expected Act Utilitarian deontic logic enables designers to reason about an agent’s strategic obligations, including ethical and social constraints, when operating in uncertain environments. The logic allows for specifying and verifying these obligations at design time, facilitating the modification of policies to meet these constraints. This approach differs from reward-level modifications by increasing transparency into trade-offs. Two algorithms are introduced: one for model-checking whether an RL agent has the right strategic obligations, and another for modifying a reference decision policy to meet these obligations expressed in the logic. The methods are demonstrated on DAC-MDPs, which accurately abstract neural decision policies, as well as toy gridworld environments. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary When designing agents that operate in uncertain environments, it’s essential to consider what they should do, how their actions align with expectations, and how policies can be adjusted to meet these requirements. This paper proposes a new way of thinking about this problem using something called “expected act utilitarian deontic logic.” Essentially, it’s a tool for designers to specify and verify the rules that guide an agent’s behavior, taking into account ethical and social considerations. Two algorithms are developed to work with this logic: one checks if an RL agent is following the right rules, and another adjusts a policy to meet these rules. The approach is demonstrated on simple scenarios. |
