AI Glossary / Graphs and symbolic AI / Intermediate
Closed-world assumption
Definition
The closed-world assumption is a reasoning convention that treats relevant facts not established by the knowledge base as false. It lets a system draw negative conclusions from the absence of positive information within a defined scope. The assumption is useful when that scope is sufficiently complete, but it can turn missing data into incorrect conclusions when completeness is not justified.
What it means
Suppose an authoritative table lists every active room reservation for tomorrow. If no reservation exists for a room and time, the application may treat that slot as unreserved. It is relying on the table being complete for that purpose. A partial collection of public announcements would not support the same inference.
The contrasting open-world assumption allows an unrecorded fact to remain unknown. Neither convention says that all data is accurate. They specify how to interpret absence. The chosen convention should be attached to a particular relation, time period, or knowledge source rather than silently applied to everything an organization knows.
How it works
Identify the facts and rules the system uses and the domain over which negative conclusions are allowed. Establish whether a proposition is recorded or can be derived. Under an appropriate closed-world convention, failure to establish that proposition supports treating it as false for the modeled task.
In a simple planning state represented by a set of positive facts, this can mean that any relevant fact outside the set is false. More general knowledge bases require care: applying a naive closure to arbitrary disjunctive information can create inconsistency. The assumption is not a substitute for a sound reasoning method, complete search, or an explicit policy about which parts of the knowledge base are closed.
Three practical examples
- A booking system treats an absent reservation as no booking when its reservation table is authoritative for the specified date and resource.
- A classical planning model treats a location fact missing from its complete state description as false when checking action preconditions.
- A configuration tool treats an option not selected in a complete configuration record as disabled, while keeping incomplete draft records separate.
Think of it this way
Think of checking a finalized guest list at an event where every invited guest must be recorded. Absence has meaning because the list is intended to be complete. The analogy fails for a rough note of people someone happens to remember. The list's authority and scope, not the mere existence of a list, justify the negative conclusion.
Worked example
Consider three rooms, A, B, and C, and one specified time slot. The finalized reservation relation contains Reserved(A) and Reserved(C), with no rules that derive additional reservations. If the relation is complete for these rooms and this slot, the closed-world interpretation makes Reserved(B) false. The set of unreserved rooms is therefore {B}.
Now suppose the records are only a partial import from another system. The same two positive facts still establish that A and C are reserved, but B's absence no longer establishes that B is unreserved under an open-world interpretation. Its status is unknown until the missing information is resolved.
If a late authoritative record Reserved(B) is added, the earlier closed-world negative conclusion must be withdrawn. The data update changes what absence supports. Also, unreserved is not automatically usable: the room could be unavailable for a separate reason. Inferring usability would require additional facts and rules.
Illustrative case study
At fictional Linden Events, an assistant uses an incomplete spreadsheet to suggest available rooms. A missing row is initially interpreted as an available slot. A reviewer discovers that the sheet contains only one department's bookings.
The team limits negative conclusions to a consolidated reservation source and labels other records as incomplete. It also separates booking status from maintenance status. This fictional case illustrates how an apparently simple missing-data rule depends on provenance and scope. It does not claim a real booking incident or a quantified reduction in errors.
Common misconceptions and mistakes
Not found is not universally equivalent to false. A search may be incomplete, a source may be stale, or the relevant fact may use a different representation. Closed-world reasoning deliberately adds an assumption about such absence; it does not prove that the world contains no missing facts.
It also differs from the unique-names assumption, which concerns whether different names identify different entities. SQL null values and their comparison rules introduce further semantics that should not be collapsed into this definition. Finally, a failed attempt to prove something within a resource limit is not necessarily logical failure. An operational search timeout should not silently become a negative fact.
Business use
Businesses should document which datasets are complete enough to support negative answers and for what period. This is especially relevant when an AI assistant retrieves records: no retrieved evidence may reflect retrieval limits rather than actual absence. The interface should distinguish a negative result from an unknown result whenever the source does not justify closure. Test missing, delayed, and partially imported records, because those cases reveal whether the system is applying the assumption too broadly.
People also ask
Does closed-world mean the database contains every fact in existence?
No. Closure concerns a defined domain or relation needed for a task.
Can new information change a negative conclusion?
Yes. A conclusion based on absence may no longer hold after a positive fact is added.
Does open-world reasoning prevent negative conclusions?
No. Explicit negative facts or logical consequences can still establish them; absence alone is insufficient.
Check your understanding
Why can the partial room import not establish that B is free?
Answer: It has not been shown to contain all reservations for B. The missing record may reflect incomplete coverage, so the closed-world assumption is not justified by that source.
Related terms
- Knowledge base
- Classical planning
- Circumscription
- Unique-names assumption
- Open-world assumption
- negation as failure
- data completeness
- nonmonotonic reasoning
Learn next
Learn open-world reasoning next. Revisit a small dataset and label each answer as established true, established false, or unknown. Then identify which explicit completeness claim would justify changing an unknown answer into a closed-world negative.
More in Graphs and symbolic AI
- Circuit verification - Circuit verification checks whether a circuit design satisfies a stated specification.
- Clause - A clause, in the logic used by many symbolic AI systems, is a disjunction of literals: statements joined by logical OR, where each statement is an atom or...
- Closed-loop - Closed-loop describes a system in which information about the system's behavior feeds back into decisions that influence its subsequent behavior.
- Closeness - Closeness, or closeness centrality, measures how near a node is to other nodes in a graph according to shortest-path distances.
- Clutter - Clutter, in tracking and data association, means detections that do not come from the target or targets of interest and instead create nuisance observations.
- co-NP - co-NP, usually written coNP, is the class of decision problems whose complementary problems belong to NP.
Sources
Primary sources: Russell and Norvig, original planning chapter and state semantics, https://aima.cs.berkeley.edu/newchap11.pdf ; W3C OWL 2 Primer, database and ontology assumptions, https://www.w3.org/TR/owl-primer/ ; original knowledge-graph evaluation research, https://arxiv.org/abs/2209.08858 . The reservation example and fictional case are original illustrations.
Previous: Closed-loop Next: Closeness