Tag: #decision-theory6 results
YUKTI introduces a typed-proposition graph to move beyond naive point-valued optimization in language models. By incorporating uncertainty, provenance, and assumption-robust Pareto frontiers, it significantly reduces decision regret in complex, real-world scenarios.
This post operationalizes Functional Decision Theory (FDT) using logical causal graphs, defining the logical do-operator and exploring its options. It uses Parfit's hitchhiker to illustrate why specific do-operator definitions enable FDT agents to pay. The analysis compares FDT to CDT and EDT, emphasizing logical counterfactuals and un-updating.
The article proposes a 'pragmatic' version of Functional Decision Theory (FDT) to address theoretical ambiguities. It explores the intersection of decision theory and game theory, specifically regarding how predictors influence agent behavior.
Article analogies Euclid's contested parallel postulate to decision theory's independence axiom in vNM framework. Bolyai's non-Euclidean geometry shows axioms aren't always necessary; similarly, dropping independence may yield consistent alternatives to expected utility maximization. Questions if rational agents must strictly follow EU to avoid Dutch books.
This post refines a thought experiment to critique expected utility maximization, emphasizing updatelessness. Omega flips a coin, reveals the result, asks for $1, and punishes $1M if it predicts you wouldn't pay in the counterfactual world. It argues ignoring counterfactuals leads to value destruction, challenging standard decision theories.
Soft set theory offers a framework for parameterized decision modeling by assigning subsets to attributes, handling uncertainty effectively. The survey covers major extensions including hypersoft sets, superhypersoft sets, TreeSoft sets, bipolar soft sets, and dynamic soft sets. It highlights connections to topology and matroid theory, with core definitions and current developments.