Introduction
The Open World Assumption (OWA) is a fundamental principle in ontology and semantic web technologies that states "unknown is not false." This means that if something isn’t explicitly stated in the data, it’s not considered false, but rather unknown.
Overview
The Open World Assumption (OWA) is a key principle used in Description Logic (DL) reasoning for ontologies. Under OWA, anything not explicitly stated in the data or inferable from explicit axioms is assumed to be unknown rather than false or non-existent. This principle is important for the purposes of extensibility, as it allows ontologies to be extended without invalidating prior inferences.
The OWA contrasts with the Closed World Assumption (CWA), which assumes that anything not explicitly stated is false. The OWA is particularly important in the context of Semantic Web Technologies (SWT) and ontologies used in Digital Engineering (DE) frameworks like IoIF.
| The OWA is not a property of the data itself, but rather a property of the reasoning system that operates on the data. This means that the same data can be reasoned about using either OWA or CWA, depending on the reasoning profile. |
Position in Knowledge Hierarchy
Broader concepts: - SSVL (is-a)
Details
The OWA is a fundamental aspect of how Description Logic (DL) reasoners operate. It’s important to understand how it differs from the Closed World Assumption (CWA) and how it affects ontology development and reasoning.
OWA vs. CWA
The following table highlights the key differences between OWA and CWA:
Assumption |
Implication for Reasoning |
Open World Assumption (OWA) |
Anything not explicitly stated is assumed to be unknown, not false. Reasoners cannot infer the absence of information. |
Closed World Assumption (CWA) |
Anything not explicitly stated is assumed to be false. Reasoners can infer the absence of information. |
| The OWA is a key principle for ensuring extensibility in ontologies. If logical reasoning is done under a closed world assumption, new data or extended ontologies might render prior inferences false. Under the OWA, the set of inferences made will generally hold unless the ontology itself is modified, and can be extended if new restrictions, classes, data, or the like are added to the ontology. |
OWA in Practice
In practice, the OWA means that a reasoner operating under OWA will not flag a violation for missing information. For example, if an ontology states that every sedan has exactly four wheels, a sedan with three wheels would not be considered inconsistent under OWA because the reasoner assumes a fourth wheel exists (but it’s not explicitly stated).
This is in contrast to the CWA, where a sedan with three wheels would be considered inconsistent because the ontology explicitly states that every sedan has exactly four wheels.
| When developing ontologies for Digital Engineering applications, it’s important to consider the implications of OWA and nUNA on your reasoning requirements. The OWA can make certain verification tasks more difficult, particularly those requiring completeness checking. |
OWA and nUNA
The non-Unique Naming Assumption (nUNA) can also affect the expected output of a reasoning task under OWA. The Unique Naming Assumption (UNA) states that individuals identified with different names are assumed to be unique. The nUNA is used by default in DL-reasoning, so unless the wheels have been explicitly identified as different from each other, a DL-reasoner would assume that at least some of the wheels listed are actually the same wheel being called two different things.
| In ontology development, the combination of OWA and nUNA can lead to unexpected reasoning results. It’s important to be aware of these assumptions when designing ontologies for Digital Engineering applications. |
Practical applications and examples
The OWA has significant implications for how ontologies are used in Digital Engineering (DE) and Semantic Web Technologies (SWT) frameworks like IoIF.
Example: Vehicle Ontology
Consider an ontology describing vehicles that defines a subclass SportsUtilityVehicle with the axiom that an instance of SportsUtilityVehicle has_component exactly 4 wheel.
Under OWA: - A SportsUtilityVehicle with 5 wheels would be inconsistent (violates the cardinality constraint) - A SportsUtilityVehicle with 3 wheels would not be inconsistent (the reasoner assumes a fourth wheel exists but isn’t explicitly stated)
Under CWA: - Both a SportsUtilityVehicle with 5 wheels and one with 3 wheels would be inconsistent (the cardinality constraint is violated)
| This example illustrates why the OWA is important for extensibility in ontologies. If you’re developing an ontology for a DE application, you’ll want to consider whether OWA or CWA better fits your use case. |
Using OWA in IoIF
The IoIF (Armaments Interoperability and Integration Framework) uses OWA for its Semantic System Verification Layer (SSVL). The SSVL uses three different verification approaches: 1. Open World Analysis (Description Logic Reasoning) - uses OWA 2. Closed World Analysis (SHACL) - uses CWA 3. Graph-Based Analysis
| The OWA is used in the Open World Analysis prong of the SSVL, while Closed World Analysis (using SHACL) is used to handle cases where the absence of data needs to be verified (e.g., "this property must be present"). |
Let’s consider a specific example from the context: a requirement that "All value properties shall be tagged with a value in the loaded ontologies" (Requirement 7).
Under OWA, a value property that is not tagged with an ontology term would not be flagged as inconsistent (since the ontology doesn’t state that all value properties must be tagged). However, under a Closed World Analysis using SHACL, the same value property would be flagged as a violation.
This demonstrates why the SSVL uses both OWA and CWA approaches to provide a comprehensive verification framework.
Related wiki pages
References
Knowledge Graph
Visualize the relationships between OWA and related concepts