Summary of Can Cdt Rationalise the Ex Ante Optimal Policy Via Modified Anthropics?, by Emery Cooper et al.
Can CDT rationalise the ex ante optimal policy via modified anthropics?
by Emery Cooper, Caspar Oesterheld, Vincent Conitzer
First submitted to arxiv on: 7 Nov 2024
Categories
- Main: Artificial Intelligence (cs.AI)
- Secondary: Computer Science and Game Theory (cs.GT)
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 explores a fundamental challenge in decision theory, specifically Newcomb’s problem. Here, causal decision theory (CDT) recommends two-boxing, diverging from evidential decision theory (EDT) and ex ante policy optimisation (which suggest one-boxing). The authors argue that CDT might recommend one-boxing if one believes they are in a simulated world designed by the predictor to determine whether to fill the opaque box. To resolve this paradox, the paper studies generalisations of this approach, considering Newcomblike problems and self-locating beliefs. The authors propose two approaches: simulating the agent’s world and ‘Generalised Generalised Thirding’ (GGT). For each method, they characterise the resulting CDT policies and prove that under certain conditions, these include the ex ante optimal policies. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary Newcomb’s problem is a classic puzzle in decision theory. Imagine you’re given a choice: take one box with some money or two boxes, where one box has a million dollars if the predictor thought you would choose the single box. Most theories say to take both boxes (two-boxing). But what if you think you might be living in a simulated world designed by this predictor? This paper looks at how we can solve Newcomb’s problem and make decisions when we’re not sure if we’re in a simulation or real life. |
