Dempster–Shafer theory¶

Assign evidential mass to subsets of a frame of discernment, derive belief and plausibility bounds for propositions, and combine appropriately independent evidence with a declared conflict rule.

Version: v3 · 2026-09-19 · History
Domain-specific #: 1641
Domain group: Applied Sciences & Engineering
Origin domain: Computer Science & Software Engineering
Subdomains: Uncertain Reasoning, Belief Functions and Evidence Fusion → Computer Science & Software Engineering

Core Idea¶

In Dempster–Shafer theory a basic belief assignment \(m:2^{\Theta}\to[0,1]\) places support on subsets of a frame \(\Theta\), usually with \(m(\varnothing)=0\) and \(\sum_A m(A)=1\); \(Bel(A)=\sum_{B\subseteq A}m(B)\) and \(Pl(A)=\sum_{B\cap A\ne\varnothing}m(B)\) bound support for (A). Mass assigned to a nonsingleton focal set records support that does not discriminate among its members; Möbius transforms connect mass, belief and plausibility, while a typed fusion rule intersects focal sets and either exposes, redistributes, or retains conflict according to its assumptions.

How would you explain it like I'm…

One-of-the-Twins Clues

Imagine a cookie went missing and you know one of the twins took it, but you can't tell which. Instead of guessing half-and-half, you can say "my belief goes on the twins as a pair." Then for each twin you get a smallest amount you're sure of and a biggest amount it could be. That way of keeping track of what you do and don't know is Dempster–Shafer theory.

Belief and Plausibility Bounds

Dempster–Shafer theory is a way to handle evidence when you are not sure. Instead of giving a chance to each single answer, you can give support to a group of answers when your evidence cannot tell them apart. From that, you get two numbers for any answer: how much the evidence surely supports it (belief) and how much the evidence at least allows it (plausibility). The gap between them shows what you just do not know. There are also rules for combining evidence from different sources, including what to do when the sources disagree.

Evidence on Sets of Answers

In Dempster–Shafer theory you start with a frame, the set of possible answers, and a mass function that assigns amounts of support to subsets of that frame, with masses adding up to 1 and usually none on the empty set. Putting mass on a set like {A, B} means the evidence supports 'A or B' without favoring either. From the masses you compute belief in a set (the total mass on subsets inside it) and plausibility (the total mass on sets that overlap it). Belief is a lower bound and plausibility an upper bound. This lets the theory tell apart 'I have no idea' from 'I think each option is equally likely,' which ordinary probability treats the same way. Combination rules merge evidence from independent sources and reveal conflict when sources disagree.

Dempster–Shafer theory represents evidence with a basic belief assignment m: 2^Θ → [0, 1] over subsets of a frame of discernment Θ, usually with m(∅) = 0 and masses summing to 1. Subsets with positive mass are focal elements, and mass on a nonsingleton set records support that does not discriminate among its members, making ignorance explicit rather than spreading it evenly as equal probabilities. Belief Bel(A) = Σ_{B⊆A} m(B) and plausibility Pl(A) = Σ_{B∩A≠∅} m(B) give lower and upper bounds on support for A; they are dual (Pl(A) = 1 − Bel(not-A)), monotone, and connected to m by Möbius inversion. Belief is not additive in general. Fusion rules, Dempster's rule and alternatives, intersect focal sets across sources and then expose, redistribute or retain the mass that falls on the empty set (conflict), according to their assumptions; Dempster's normalization is undefined under total conflict, and the rules assume independent sources. Common errors include putting all mass on singletons (reducing to ordinary probability), treating belief as additive, combining dependent reports as independent, and conflating ignorance with equiprobable uncertainty. Applications include sensor and expert evidence fusion, conflict analysis and producing lower/upper support bounds.

Scope of Application¶

Dempster–Shafer theory applies when the analyst can specify a finite frame of mutually exclusive possibilities and its power set, together with evidence sources represented by basic belief assignments and establish that support is allocated to subsets rather than only singleton outcomes, belief and plausibility are derived consistently from that mass assignment, and any multi-source fusion declares the independence and conflict-normalization semantics it requires. The entry presents a family of related belief-function formalisms, not one universally accepted semantics. Every application must declare its frame, source interpretation, empty-set convention, and combination rule rather than invoking the label as a generic badge of uncertainty handling.

Clarity¶

A clear claim names the carrier, governing rule, assumptions, and recognition test. This matters because belief, confidence, support, mass and evidence also have ordinary-language and Bayesian meanings, and the wider evidence-theory literature contains competing interpretations and conflict rules. The disciplined statement is that the object counts as Dempster–Shafer theory exactly when support is allocated to subsets rather than only singleton outcomes, belief and plausibility are derived consistently from that mass assignment, and any multi-source fusion declares the independence and conflict-normalization semantics it requires

Manages Complexity¶

The abstraction compresses Dempster's original multivalued mappings, Shaferian belief functions, normalized and unnormalized conjunctive rules, disjunctive and cautious fusion, transferable-belief interpretations, random-set interpretations, finite and extended frames, and sensor-fusion applications into a stable carrier, rule, invariant, and failure boundary. It makes comparison tractable while retaining the variables that control validity.

Abstract Reasoning¶

Type the carrier. Establish a finite frame of mutually exclusive possibilities and its power set, together with evidence sources represented by basic belief assignments and reject examples from a different problem.
Lock the rule. Express that support is allocated to subsets rather than only singleton outcomes, belief and plausibility are derived consistently from that mass assignment, and any multi-source fusion declares the independence and conflict-normalization semantics it requires independently of one notation or implementation.

Knowledge Transfer¶

Transfer within uncertain reasoning is strong when new cases preserve the same carrier, mechanism, and diagnostic. The move from Two conditionally independent sensors assign mass to overlapping subsets of a finite fault frame; their conjunctive intersections reveal agreement and conflict before a declared Dempster normalization produces the fused belief function to A classifier abstains between several related labels by placing mass on their union, then reports belief and plausibility for a coarser operational category rather than fabricating precise singleton probabilities demonstrates that continuity.

Relationships to Other Abstractions¶

Current abstraction Dempster–Shafer theory Domain-specific

Parents (1) — outgoing typed relations

Dempster–Shafer theory is a kind of Statistical Inference Prime

The framework reasons from evidence to uncertain propositions under a formal model and therefore instantiates statistical inference broadly construed.

Hierarchy paths (3) — routes to 3 parentless roots

Dempster–Shafer theory → is a kind of Statistical Inference → is a kind of Inductive Reasoning

Show alternative paths (2)

Neighborhood in Abstraction Space¶

Dempster–Shafer theory sits in a moderately populated region (50^th percentile for distinctiveness): it has near-neighbors but no dense thicket of look-alikes.

Family — Metaphysics of Mind & Knowledge (23 abstractions)

Nearest neighbors

Computed from structural-signature embeddings · 2026-10-25