liveBook · Manning

Decision making

As explained in Timeless Algorithms, The Foundational Ideas

Decision-making is the process of selecting an action or outcome under available information, constraints, and uncertainty.

Decision-making extends Bayesian inference from revising beliefs to choosing among acts. Probability describes what may be true; choice also depends on the consequences of each act in each possible state. Expected utility combines them: EU(a) = Σₛ P(s)U(a,s). Prefer the act with the highest expected utility, while recognizing that uncertainty can still produce an unfavorable outcome. section 7.6

A decision under uncertainty can be organized around three elements:

  • Acts are the available choices; states are unknown conditions outside the decision-maker’s control; experiments are ways to learn before acting. section 7.6
  • Decision trees make the reasoning explicit: decision nodes show available acts, chance nodes show uncertain events, and terminal nodes show outcomes. Rollback evaluates the tree from its outcomes toward the preferred act. section 7.6
  • Value-of-information analysis compares the expected benefit of learning with its cost. EVPI bounds the value of complete information; EVSI measures the expected improvement from imperfect information before acquisition cost. section 7.6
  • Opportunity loss measures an outcome against the best act that could have been chosen with the true state known. Minimizing expected opportunity loss selects the same act as maximizing expected utility. section 7.6

The framework improves the coherence of a choice, not its guarantee of success: uncertainty remains after acting, so a rational decision can still have a poor outcome. section 7.6

How Timeless Algorithms, The Foundational Ideas covers it

How deep it goes
Defines expected utility, decision-tree rollback, and information-value analyses, while noting that rational choice cannot guarantee a favorable outcome.
Examples
Uses decision trees, EVPI and EVSI, and opportunity loss to make choices under uncertainty explicit.
What it assumes
Builds on Bayesian reasoning while defining acts, states, and experiments as the structure for decisions under uncertainty.
What you can do after
Evaluate competing acts by expected utility or expected opportunity loss and assess whether additional information is valuable.
How it connects
Connects Bayesian inference, probability, expected utility, decision trees, and value-of-information analysis.

Also explained in

Compare how 9 books explain decision making →

Related concepts

This explainer was generated by Manning's knowledge graph from the text of Timeless Algorithms, The Foundational Ideas. The section links point to the parts of the book it draws on.