Decisiongraph

Decisiongraph is a term used in multiple disciplines to describe a graphical representation of decisions and their consequences. Broadly, a decisiongraph is a directed graph in which nodes denote decisions, events, or outcomes, and edges denote actions or transitions between states. It is used to model sequential decision processes, sometimes under uncertainty, and can serve as an alternative or generalization of decision trees and flowcharts.

Compared with traditional decision trees, decision graphs may allow sharing of subgraphs, cycles, or converging paths, enabling more compact or reusable representations. Depending on the context, edges may carry probabilities, costs, or other annotations, turning a decision graph into a probabilistic, costed, or qualitative model; in other cases it remains purely deterministic.

Nodes are typically categorized as decision nodes (where an agent selects an action), chance or event nodes (representing uncertainty), and terminal nodes (outcomes). Some formulations include information nodes that carry sensory data or observations used to guide decisions.

Applications span decision analysis, operations research, artificial intelligence, risk assessment, and medical decision support. In AI, decision graphs appear in planning systems, policy representations, and rule-based or graph-based classifiers, and are related to, yet distinct from, influence diagrams, decision diagrams, and Bayesian networks.

Variations exist across disciplines; some authors use decisiongraph to describe a class of models whose subgraphs can be shared to reduce redundancy, while others apply the term to specific graph structures for learning or decision support.

See also: decision tree, influence diagram, decision diagram, probabilistic graphical model, planning graph.