evidential versus causal decision theory
Evidential Decision Theory is a branch of decision theory that focuses on observational evidence: given a decision problem, Evidential Decision Theory prescribes the action that we observe to have done in the best possible outcome [Ahm14]. This contrasts with Causal Decision Theory, which prescribes the action that causes the best possible outcome [GH78]. In Evidential Decision Theory, no direct causal connection is required for the action to affect the outcome: it suffices that the observation of the action alters the conditional probability of the outcome via Bayesian update [YS17]. Characterizing, comparing and formalizing decision theories, such as Evidential Decision Theory, remains an open problem in artificial intelligence research [HBH88], [ELH15].
References
- Evidence, Decision and Causality (Ahmed, 2014)
- Counterfactuals and Two Kinds of Expected Utility (Gibbard, Harper, 1978)
- The Lesson of Newcomb’s Paradox (Wolpert, Benford, 2013)
- Functional Decision Theory. A New Theory of Instrumental Rationality (Yudkowsky, Soares, 2018)
- Timeless Decision Theory (Yudkowsky, 2010)
Outline
Context: decision theory, Newcombs problem.