Propositional Constraint Graphs: An Intuitive, Domain-General Tool for Diagramming Knowledge, Assumptions, and Uncertainties

Abstract

A diagramming method called Propositional Constraint (PC) graphing was developed as an aid for tasks involving argumentation, planning, and design. Motivated by several AI models of defeasible (or non- monotonic) reasoning, PC graphs were designed to represent knowledge according to an analogical framework in which constraints (e.g., evidence, goals, system constraints) may elicit or deny possibilities (e.g., explanations, decisions, behaviors). In cases of underspecification, an absence of constraints yields uncertainty and competition among plausible outcomes. In cases of overspecification, no plausible outcome is yielded until one of the constraints is amended or forfeited. This framework shares features with theoretical models of reasoning and argumentation, but despite its intuitiveness and applicability, we know of no modeling language or graphical aid that explicitly depicts this defeasible constraint structure. We describe the syntax and semantics for PC graphing and then illustrate potential uses for it.

Nodes Afford Connection: A Pilot Study Examining the Design and Use of a Graphical Modeling Language

“…Propositional constraint (PC) graphs are directed acyclic graphs representing how sets of propositions (factual true-false statements) are related by compatibility, possibility, and inference. They are designed to visualize sequences of events and chains of inference (Pugh & Gillan, 2020). For instance, Figure 2 represents a chain of inference that further branches into multiple independent inferences.…”