Glossary
Open-world assumption
Under the open-world assumption, a missing statement does not imply that the statement is false.
Knowledge fusion
The open-world assumption treats a knowledge base as potentially incomplete. If no statement says an entity has a certain relationship, that absence alone cannot establish that the relationship does not exist. The assumption is common in web knowledge representation, where no dataset can be expected to enumerate every fact. Explicit negative information requires its own representation and evidential basis.
In practice, open-world reasoning prevents an empty query result from becoming an unsupported denial. A source may omit a trial status because it has not updated, lacks scope or uses another term. Applications still need operational choices, but they should distinguish unknown from false. A completeness claim about a specific source, property and period may justify a narrower conclusion.
For organizations defending decisions, this distinction is central: absence of evidence and evidence of absence have different meanings. Closed-world reasoning can be appropriate within a declared complete register or transaction set. The two assumptions are policies about interpretation of missing facts, not competing claims about which individual source is more reliable.
Example
No Halden Therapeutics AG (fictional) trial appears in one publication index; under open-world reasoning that does not prove no trial exists.
What the assumption means in practice
Under the open-world assumption a system cannot conclude that a statement is false merely because it does not know it to be true. Absence of a statement leaves three possibilities open: the statement is true and unrecorded, it is false, or it is unknown. The OWL 2 Primer describes OWL's reasoning in these terms, and the assumption is natural for the web, where no collection claims to describe everything. A consequence is that adding a statement can never make a previous conclusion invalid, which is what makes reasoning monotonic.
The practical effect is easiest to see in a question that a database user would find trivial. Asked whether a study has a sponsor, a closed-world system with no sponsor recorded answers no. An open-world system answers that it does not know, unless the model states that every study has a sponsor, in which case it infers that one exists though not which. The article on why unknown is not false works through such questions and the design errors that follow from treating the answers as equivalent.
Where an organisation needs closure
Some decisions require a closed answer. A compliance check that asks whether a counterparty appears on a sanctions list needs to know that the list was complete as of a stated date and that the search covered all of it. The open-world assumption does not prevent this; it requires that the completeness be stated as a fact about a defined dataset, in the way Reiter's closed-world assumption is limited to a database whose contents are declared complete. In practice this means attaching to a collection a statement of its scope and date, and writing the query to say not found in this collection as of this date.
Constraint languages such as SHACL take the closed view of the data actually present, so that a required property that is missing is reported as a violation, even though an OWL reasoner would simply infer that a value exists somewhere. Using the two together is common and deliberate, and the article on OWL versus SHACL explains how to decide which job belongs to which.
A worked example
A register of authorised representatives for Halden Therapeutics AG (fictional) lists three individuals. A user asks whether a fourth person is an authorised representative. If the register is described as a complete list as of 1 March, the answer from that register on 1 March is no. If the register is described only as a list of known representatives, the answer is not known. If the question is asked in May, the answer depends on whether changes since March are known. The record that serves the user keeps the scope and date of the register beside the answer, and shows an explicit unknown when the register's scope does not cover the question, instead of a bare negative.
Related terms and boundaries
The assumption underlies most ontology reasoning and the knowledge graph systems built on it, and it sets the meaning of a blank in null semantics. It is the reason a system should be able to abstain and should store unknown as a value that is distinct from false. Its counterpart, the closed-world assumption, is the right reading inside a dataset whose completeness has been established.
Checks a reviewer can run
Ask how a missing statement is answered: false, unknown or not recorded. Test a question whose answer is absent from the data. Ask whether collections state their scope and date, so that closure can be applied where it is justified. Check that unknown is stored as a distinct value and is not collapsed into false in exports.
Common misreadings
- Turning a failed search into a negative factual claim.
- Ignoring explicit completeness guarantees.
- Treating unknown and false as interchangeable.
Appears in
- Blank cell vs declared unknown
A blank leaves its reason unstated; a declared unknown records what was checked and why no value is available.
- Declaring unknowns
Distinguish not found, blocked by licence and not applicable, with a dated search boundary.
- Resource Description Framework (RDF)
A graph data model for statements about identified resources.
Found an error on this page? See corrections for what counts as one, how to report it and the dated log.
Enquiry
Where does "Open-world assumption" bear on a decision of yours?
This entry sits under knowledge fusion. If a decision in your organization turns on it, send the entity concerned, the sources you fuse to describe it and who may read which of them. We reply with what a first record would cover.
Or write to [email protected] in your own words.