Fact-checked by Grok 2 weeks ago

Closed-world assumption

The closed-world assumption (CWA) is a in formal logic, , and systems that presumes a is false unless it can be logically deduced as true from the available , effectively treating the absence of evidence as . Introduced by Raymond in his 1977 and 1978 "On Closed World Data Bases," the CWA originated as a semantic framework for deductive databases to enable definite query answering in domains where the is assumed to be complete and exhaustive. Under this assumption, formalized the evaluation of queries by augmenting the database with the negations of all unprovable ground atoms, ensuring that the resulting theory yields precise yes/no answers without gaps. This approach is particularly suited to relational databases and environments, such as , where explicit facts represent all relevant truths, like flight schedules or inventory records. In contrast, the open-world assumption (OWA) treats unknown statements as potentially true, requiring explicit negations to assert falsehood, which is more appropriate for incomplete or evolving knowledge bases like the Semantic Web. The CWA's negation-as-failure semantics can lead to inconsistencies in non-monotonic logics or non-Horn clause sets, but it remains consistent and computationally efficient for definite clause programs and Horn databases. Extensions, such as the local closed-world assumption (LCWA), address limitations by applying the CWA selectively to subsets of the knowledge base, enhancing applicability in distributed or heterogeneous systems. The CWA has profoundly influenced and , underpinning query optimization in relational systems and enabling non-monotonic reasoning in expert systems, though its strictness often necessitates hybrid approaches with OWA in modern knowledge graphs and probabilistic databases.

Fundamentals

Definition and Core Principles

The closed-world assumption (CWA) is a foundational principle in , positing that the contains all relevant facts about the domain, such that any not explicitly entailed as true must be false. This assumption treats the as complete and exhaustive, enabling systems to draw definitive conclusions from the absence of information. At its core, the CWA relies on the completeness of the , where every true is derivable from the given axioms or data, and negation as failure serves as the mechanism for inferring falsehoods: a is considered false if it cannot be proven true through standard procedures. This has direct implications for query answering, as it allows automated systems to resolve queries with by evaluating whether a fact can be derived; if not, the query fails definitively rather than remaining indeterminate. In contrast to everyday human reasoning, which often operates under conditions of incomplete information and avoids assuming falsehood from mere absence of , the CWA imposes a stricter interpretive to facilitate computational tractability. Its initial motivation in systems stems from the need to handle domains with finite, explicitly represented positive knowledge, ensuring decisive inference without requiring exhaustive enumeration of negatives.

Historical Origins

The closed-world assumption (CWA) has roots in the foundational work on relational databases, where Edgar F. Codd's 1970 proposal of the implicitly embodied the idea that all relevant facts are explicitly stored in the database, with absence of information implying negation. This assumption facilitated efficient querying and inference by treating the database as a complete representation of the domain, influencing early database query languages and operations like set difference. The formal articulation of the CWA emerged in 1978 through Raymond 's seminal paper "On Closed World Data Bases," which explicitly defined it in the context of and deductive databases. introduced the CWA for deductive databases assuming the is complete, to address by positing that if a cannot be deduced from the available facts, it is false, thereby enabling sound closed-world evaluation of queries reducible to atomic forms. This work built directly on the relational model's implicit assumptions, providing a logical foundation for handling defaults and nonmonotonic reasoning in computational systems. During the and , the CWA gained prominence in , particularly within the expert systems era, where it underpinned practical rule-based inference in knowledge representation. Systems like and early expert systems adopted the CWA to justify negative inferences from the absence of evidence, supporting efficient decision-making in domains such as and configuration tasks. Reiter's formulation profoundly impacted subsequent research, inspiring developments in nonmonotonic logics and semantics, including the Clark completion and negation-as-failure mechanisms.

Formal Representation

Logical Formalization

The closed-world assumption (CWA) in first-order logic is formally defined for a knowledge base KB comprising a finite set of positive ground literals (atomic facts) as the augmented theory \mathrm{CWA}(KB) = KB \cup \{\neg P \mid P \text{ is a ground atom not derivable from } KB\}. This formulation assumes the domain is fully specified by KB, such that any ground atom absent from or unprovable in KB is explicitly false in the extended theory. The CWA thus introduces explicit negation for all non-entailed atoms, enabling non-monotonic reasoning over incomplete information while maintaining consistency for definite (Horn clause) theories. In the context of logic programming, the CWA is operationalized through the negation as failure (NAF) inference rule. The NAF operator, denoted \mathrm{not}\, P, succeeds (derives \neg P) if and only if there exists no SLD-refutation (proof via linear resolution) for the ground atom P from the program P. Formally, for a normal logic program P, the success set of \mathrm{not}\, P under NAF is defined relative to the failure set of P, where failure occurs after exhaustive search without refutation. This rule aligns with the CWA by treating unprovable atoms as false, but it applies more broadly to programs with negation in rule bodies, provided the program satisfies safety conditions to ensure finite failure. A key semantic foundation for NAF under the CWA is provided by completion semantics, which transforms a logic program into an equivalent theory by converting each rule into a biconditional . For a definite logic program P with a predicate p defined by rules p(\mathbf{x}) \leftarrow b_i(\mathbf{x}), i = 1, \dots, m, the Clark completion \mathrm{comp}(P) includes: \forall \mathbf{x} \left( p(\mathbf{x}) \leftrightarrow \bigvee_{i=1}^m b_i(\mathbf{x}) \right) along with similar completions for other predicates and equality axioms. This equivalence enforces closedness by stipulating that p holds precisely when at least one defining body is true, implying \neg p(\mathbf{x}) otherwise—directly embodying the CWA for the Herbrand universe. For programs without negation, \mathrm{comp}(P) coincides with the least Herbrand model under the CWA, ensuring a unique minimal model. For definite clause programs (Horn clauses without negation), the CWA admits a soundness and completeness theorem with respect to SLD-resolution. Specifically, SLD-derivability is sound and complete for positive ground queries relative to the least Herbrand model M_P of P, meaning a ground atom A is in M_P if and only if there is an SLD-refutation for \leftarrow A. Under the CWA, extended to negative ground literals via NAF, the inference is sound: if \mathrm{not}\, A succeeds (no SLD-refutation for A), then \neg A holds in \mathrm{CWA}(P). Completeness follows for the combined positive-negative queries, as the CWA uniquely determines the full Herbrand base (H_B \setminus M_P for false atoms), and exhaustive SLD-search detects all failures in finite domains. This result relies on the program's stratification and the finiteness of the Herbrand universe, ensuring no infinite resolution paths.

Mathematical and Computational Models

The well-founded semantics provides a declarative foundation for logic programs under the (CWA), assigning truth values (true, false, or undefined) to atoms based on derivability. Introduced for general logic programs with , it constructs the well-founded model through an iterated fixpoint process that operationalizes the CWA by treating non-derivable atoms as false once their unfounded status is confirmed. This semantics ensures a unique minimal model, avoiding the multiple models possible in other nonmonotonic approaches. Central to this construction is the immediate consequence operator T_P, defined for a logic program P and interpretation I as the set of atoms p for which there exists a rule in P whose positive body is satisfied in I and negative body is true under negation as failure relative to I. The well-founded model is obtained by iterating an operator W_P(I) = T_P(I) \cup \{ \neg a \mid a \in U_P(I) \}, where U_P(I) identifies unfounded atoms relative to I (atoms depending on themselves through cycles involving undefined literals). Starting from the empty interpretation, the least fixpoint I^\infty of W_P yields the well-founded model, with true atoms from I^\infty, false atoms as their negations in I^\infty, and the rest undefined; this process terminates in countably many steps and aligns with CWA by falsifying unfounded atoms. In (ASP), the stable model semantics extends the CWA by defining answer sets as stable models that implicitly enforce closure through unfounded sets, ensuring no external support for beliefs. A stable model M of a program P is a subset of the Herbrand base such that M is the least model of the positive program P^M, obtained by deleting rules in P with false negative literals in M and removing negative literals; this reduction captures the CWA by assuming absence of evidence for negated atoms implies their falsity. Unfounded sets play a key role, as a set U is unfounded with respect to M if no atom in U has an external justifying (a whose head is in U, body true in M, and at least one positive body literal outside U); stable models are precisely the unfounded-free fixpoints of T_{P^M}, integrating CWA with skepticism toward circular justifications. This semantics underpins ASP solvers, where multiple stable models represent possible closed worlds. Probabilistic extensions of the CWA incorporate by assigning zero probability to facts, enabling reasoning in partially domains while maintaining closure. In closed-world probabilistic databases, the semantics assumes that facts absent from the database have probability zero, contrasting with open-world variants where they may have positive probability; this directly operationalizes CWA in probabilistic query answering. Such frameworks integrate with Bayesian networks by representing dependencies via programs, where the closed-world prior ensures that the over worlds marginalizes atoms to zero, facilitating exact through lifted or sampling. Complexity analysis reveals that query entailment under CWA in propositional logic is NP-complete for brave reasoning (existence of a model satisfying the query). Under stable model semantics for propositional normal programs—which embody CWA via as failure—deciding if there exists a stable model where a given is true is NP-complete, as it reduces to checking the least model of the reduct against the query, with membership in NP via nondeterministic guessing of the model. Cautious entailment (truth in all stable models) elevates to in this setting, highlighting the computational trade-offs of CWA in nonmonotonic propositional reasoning.

Practical Applications

Examples in Logic Programming

In logic programming, particularly in , the closed-world assumption (CWA) is implemented via negation as failure (NAF), which treats unprovable atomic goals as false, enabling inference over incomplete knowledge bases as if all relevant facts are explicitly stated or derivable. This allows to return definitive "no" answers for queries lacking supporting evidence, contrasting with open-world systems that might leave such queries undetermined. A classic illustration involves querying family relations, where absent facts imply non-existence under the CWA. Consider the following Prolog facts defining parent-child links:
parent(pam, bob).
parent(tom, bob).
parent(bob, ann).
parent(bob, pat).
parent(pat, jim).
With the rule for siblings:
sibling(X, Y) :- parent(Z, X), parent(Z, Y), X \= Y.
The query ?- sibling(ann, pat). succeeds, as both share the parent bob, confirming they are siblings. However, the query ?- sibling(ann, liz). fails and returns "no," implying under the CWA that Ann has no sibling named Liz, since no supporting parent fact exists. This behavior stems from NAF in the unification and backtracking process, where failure to derive a shared parent treats the relation as false. To highlight query outcomes under the CWA, consider a simple ownership example. Define the fact:
has_car(john).
The query ?- has_car(john). succeeds with "yes." In contrast, ?- has_car(mary). fails with "no," assuming under the CWA that Mary does not own a car because no fact or rule derives it—treating the knowledge base as exhaustive. Without the CWA (e.g., in an open-world logic), the same query for Mary would remain undetermined rather than negated. This demonstrates how the CWA supports decisive inference in logic programming but risks incorrect negatives if the base omits true facts. In rule-based inference, the blocks world scenario exemplifies the CWA by assuming unmentioned actions or states are impossible. States are lists of on(Block, Surface) facts, such as on(a, b), on(b, table), on(c, table) for initial configurations. Actions like moving a block are defined with rules incorporating NAF:
action(State, move(Block, Destination)) :- 
    block(Block), 
    \+ Block == Destination, 
    free(State, Block), 
    free(State, Destination).

free(State, Thing) :- 
    thing(Thing), 
    \+ member(on(_, Thing), State).
Here, free/2 uses NAF (\+) to check if a block or table position is clear: it succeeds if no on(_, Thing) appears in the state, assuming under the CWA that absent stacking facts mean nothing is on top. An unmentioned , like moving two blocks simultaneously, fails entirely, as only explicitly definable moves are possible—enforcing a closed domain of operations. This setup enables queries, such as finding a sequence to stack a on c, by iteratively applying valid actions while failing invalid ones. The use of NAF in such clauses leads to closed inference, as seen in a complete example combining facts and rules for a blocks domain:
block(a). block(b). block(c). thing(table).

initial([on(a, table), on(b, table), on(c, table)]).

goal([on(a, c), on(c, table), on(b, table)]).
A planning rule might append new states only if preconditions hold via NAF, ensuring the search assumes : unlisted configurations or actions are impossible, deriving paths like move b away, then a to c. This closed inference succeeds for achievable goals but fails (implying impossibility) for others, like stacking all on an undefined surface.

Use in Database Systems

The closed-world assumption (CWA) forms a foundational in systems, where the absence of a in a relation implies that the corresponding proposition is false. This assumption underpins the introduced by in 1970, which posits that a completely defines the , and all true facts within that domain are explicitly represented as in the relations. Under CWA, the database is treated as a complete and self-contained snapshot of the world it models, enabling efficient querying without needing to account for external or implicit positives. In SQL, the CWA manifests prominently in query semantics, particularly through constructs like NOT EXISTS, which evaluates to true if no qualifying rows are found in the subquery, effectively treating the absence of data as negation. This aligns with negation as failure, where failed proofs (empty results) are interpreted as false, supporting the default behavior of relational queries under the assumption of . For instance, a query checking for non-existent customer orders assumes that unrecorded orders do not exist, streamlining inference in transactional environments. Active and deductive databases extend the CWA through integrity constraints and rule-based s, enforcing consistency by assuming the database's explicit content fully represents the valid state. In active , event-condition-action (ECA) rules rely on CWA to repairs when constraints detect absences that violate domain rules, such as ensuring without external validation. Deductive databases, which integrate logical rules over relational facts, apply CWA to ground atoms not derivable from the facts and rules as false, facilitating sound inference in knowledge-intensive applications. A practical illustration of CWA in database systems is found in inventory management, where the relation lists only available items, and the absence of an item implies it is out of , enabling queries to reliably compute shortages without additional checks. This approach supports real-time decision-making in operations, assuming the database captures all current facts comprehensively.

Variants and Comparisons

Partial Closed-World Assumption

The partial closed-world assumption (PCWA; also known as the local closed-world assumption or LCWA) relaxes the full closed-world assumption by selectively applying the principle that unmentioned facts are false only to designated predicates or s, while treating others under an open-world semantics to accommodate incomplete information. This selective application addresses scenarios where complete knowledge exists for certain aspects of a domain but not others, building on Raymond Reiter's foundational closed-world assumption formalized in the late . Early extensions of Reiter's work in the and , such as those incorporating precedence orders among predicates, enabled this partial closure to maintain consistency in nonmonotonic reasoning. Formally, PCWA can be expressed as PCWA(KB, D), where KB denotes the and D specifies the set of closed predicates or domains. Under this , for any ground atom P(a) where P ∈ D, if KB does not entail P(a), then ¬P(a) holds; atoms outside D remain unaffected by negation-as-failure. This formulation, akin to completeness assertions over subsets of the database, allows for precise control over when is invoked, as seen in query completeness statements that declare specific queries or tables as fully captured by available data. PCWA finds application in knowledge bases blending complete local facts—such as fixed internal records in organizational databases—with incomplete external , like references to dynamic web sources, enabling reliable on the closed portions without overgeneralizing . For instance, in a database, enrollment facts might be treated as closed for local reporting, while affiliation details from external directories remain open to avoid erroneous negations. Flora-2, a modular rule-based , implements PCWA through local closed-world modules that isolate domains for selective closure, facilitating scalable reasoning in hybrid environments.

Contrast with

The (OWA) posits that the is potentially incomplete, such that unknown are neither true nor false, and their truth requires explicit or affirmation. Under OWA, a is considered false only if it is inconsistent with the existing , rather than absent implying falsity. This contrasts sharply with the closed-world assumption (CWA), where absence of for a entails its falsity. A primary difference lies in handling negation and entailment: CWA supports negation as failure, allowing inference that a proposition is false from its non-derivability, which enables compact representations in controlled settings but risks incorrect conclusions if the world is incomplete. In OWA, negation requires classical explicit denial, preserving entailment only for what is positively supported or contradicted, thus avoiding premature falsity claims but demanding more comprehensive data for reasoning. These distinctions affect logical inference profoundly; for instance, in knowledge graph completion tasks, CWA treats all unobserved triplets as negative examples, potentially underestimating model performance in incomplete datasets, whereas OWA accounts for possible missing facts, leading to more realistic but conservative evaluations. CWA is particularly suited to controlled environments where completeness can be reasonably assumed, such as software configurations or relational databases, where unmentioned attributes imply values. Conversely, OWA aligns with dynamic, open environments like web-based data or evolving ontologies, where new information frequently emerges, as seen in applications. Regarding monotonicity, CWA introduces non-monotonic reasoning, where adding new facts can invalidate prior negations (e.g., initially assuming a fact false, then retracting upon ), fostering defeasible but efficient logic. OWA, however, maintains monotonicity, ensuring that entailments persist or expand with additional , which supports stable but less decisive in uncertain domains.

Limitations and Extensions

Key Criticisms

One primary criticism of the closed-world assumption (CWA) is its propensity to generate incorrect negations when applied to incomplete or indefinite information, where the absence of evidence does not reliably imply non-existence. In scenarios involving non-Horn clauses, such as indefinite databases where a query like P_a \lor P_b yields no definitive derivation for either disjunct, the CWA inappropriately assumes the negation of unprovable facts, leading to erroneous conclusions about the state of the world. This risk is particularly acute in real-world applications with partial data, as the assumption of completeness fails, potentially overlooking plausible alternatives and resulting in flawed decision-making. The non-monotonic nature of reasoning under the CWA introduces further instability, as the addition of new facts can invalidate previously drawn , disrupting the reliability of the . For instance, an initial that no exists between two cities—based on the absence in a database—may hold under CWA until a new entry is added, retroactively negating the prior conclusion and requiring re-evaluation of dependent deductions. This retraction mechanism, while enabling , often leads to computational and logical instability in dynamic systems, where incremental updates do not preserve prior entailments as in monotonic logics. The CWA proves unsuitable for distributed or evolving knowledge bases, such as those in the , where data incompleteness and across sources preclude the assumption of a fully specified . Applying CWA in such environments can cause systemic failures, as unstated facts from remote or updating sources are misinterpreted as false, hindering integration and leading to inconsistent ontologies under the prevailing .

Modern Developments and Alternatives

In the semantic web era, particularly following the development of in the early 2000s, researchers have integrated the closed-world assumption (CWA) with to enable hybrid reasoning that combines CWA and (OWA) elements. This allows for local closure over specific fragments while maintaining global openness, addressing limitations in purely OWA-based systems like OWL DL. For instance, local closed-world reasoning under the well-founded semantics extends to support non-monotonic inferences in OWL ontologies, enabling applications such as semantic service matchmaking where certain predicates are assumed closed. Similarly, grounded circumscription provides a semantics for local CWA in OWL, restricting closure to designated individuals or concepts to avoid undecidability in expressive . Extensions incorporating fuzzy and evidential reasoning have adapted the CWA to handle in AI systems, introducing graduated or partial mechanisms. In programming, the CWA is generalized to fuzzy settings by propagating truth degrees under negation-as-failure, allowing for approximate in knowledge bases with vague information. For evidential reasoning, adaptations of Dempster-Shafer theory relax the strict CWA by assigning masses to closed subsets while accommodating open possibilities through intervals, as seen in fusion frameworks for multi-source data under partial knowledge. These approaches enable graduated , where negation strength varies with evidential support, contrasting with binary classical CWA. Recent applications of the CWA in emphasize closed-set classification, where models assume a fixed label space and treat unknowns as errors, underpinning tasks like image recognition under controlled environments. In contrast, open-set recognition extends beyond CWA by detecting novel classes, with methods like classification-reconstruction learning training discriminators to reject outliers while maintaining closed-set accuracy, for example, achieving over 10% improvements in F1-score for unknown detection on MNIST benchmarks with outliers. This shift highlights the CWA's role in foundational ML paradigms while motivating alternatives for real-world deployment. Emerging frameworks in multi-agent systems employ local closed-world reasoning to manage distributed , assuming only within an agent's observable scope to facilitate coordination without global consistency. In the , research formalized this using non-monotonic , enabling semantic in agent interactions where services are queried under partial . For example, well-founded semantics for local CWA in multi-agent ontologies support defeasible inheritance and query answering, as demonstrated in publications on agent-based service coordination. These developments promote scalable alternatives to global CWA in decentralized environments.

References

  1. [1]
    [PDF] ON CLOSED WORLD DATA BASES by RAYMOND REITER ...
    In this paper, we show that closed world evaluation of an arbitrary query may be reduced to open world evaluation of so-called atomic queries. We then show that.Missing: original | Show results with:original
  2. [2]
    On Closed World Data Bases - SpringerLink
    In this paper, we show that closed world evaluation of an arbitrary query may be reduced to open world evaluation of so-called atomic queries.Missing: original | Show results with:original
  3. [3]
    Reconciling description logics and rules | Journal of the ACM
    Achieving seamless integration is nontrivial, since DLs use an open-world assumption, while the rules are based on a closed-world assumption. We overcome ...
  4. [4]
    On the Local Closed-World Assumption of Data-Sources
    In this paper, we introduce an expressive and intuitively appealing method of representing a local closed-world assumption (LCWA) of autonomous data-sources.
  5. [5]
    Open-world probabilistic databases - ACM Digital Library
    Apr 25, 2016 · Our open-world algorithm incurs no overhead compared to closed-world reasoning and runs in time linear in the size of the database for tractable ...
  6. [6]
    5.7 Complete Knowledge Assumption
    Under this assumption, an agent can conclude that an atom is false if it cannot derive that the atom is true. This is also called the closed-world assumption.
  7. [7]
    Incomplete Information in Relational Databases
    This paper concerns the semantics of Codd's relational model of data. ... Under a so-called closed world assumption, it can also handle the operator of difference ...
  8. [8]
    Logic-Based Artificial Intelligence
    Aug 27, 2003 · It is the closed world assumption that justifies a negative answer to a query “Is there a direct flight from Detroit to Bologna?” when the ...
  9. [9]
    [PDF] Principles of Expert Systems
    to the so-called closed-world assumption which is quite common in database applications. In PROLOG, an even stronger assumption, known as negation as finite ...
  10. [10]
    Explicit and default negation in databases and logic programs
    Relational databases adopt the Closed World Assumption (CWA) of Reiter. It ... Codd. A relational model for large shared data banks. Comm. of the ACM ...
  11. [11]
    Negation as Failure | SpringerLink
    The negation as failure rule only allows us to conclude negated facts that could be inferred from the axioms of the completed data base.
  12. [12]
  13. [13]
    [PDF] The Well-Founded Semantics for General Logic Programs*
    Abstract. A general logic program (abbreviated to \program" hereafter) is a set of rules that have both positive and negative subgoals.
  14. [14]
    [PDF] The Stable Model Semantics for Logic Programming
    This paper belongs to the dire tion of resear h whi h attempts to de- fine the de larative meaning of logi programs by means of \ anoni al models". The ...
  15. [15]
    [PDF] Open-World Probabilistic Databases: Semantics, Algorithms ...
    May 3, 2021 · We revisit the semantics of probabilistic databases, and argue that the closed-world assumption of probabilistic databases, i.e., the assumption ...
  16. [16]
    [PDF] Recent Advances in Querying Probabilistic Knowledge Bases - IJCAI
    The closed-world assumption states that any fact that is not entailed by the knowledge base is false, i.e., has probability 0. • The closed-domain assumption ...
  17. [17]
    NEGATION AS FAILURE Keith L. Clark Department of Computer ...
    This suggests a way of reconciling negation as failure with its truth functional semantics. We can assume that the clauses that appear in a logic data base B ...
  18. [18]
    [PDF] PROLOG PROGRAMMING FOR ARTIFICIAL INTELLIGENCE
    Let us discuss one way of carrying out this principle in Prolog. Consider our family example of the previous section again. Each family is a collection of.
  19. [19]
    STRIPS Block World - SWISH - SWI-Prolog
    A set of wooden blocks of various shapes and colors sitting on a table. The goal is to build one or more vertical stacks of blocks. Only one block may be moved ...
  20. [20]
    [PDF] A Relational Model of Data for Large Shared Data Banks
    1970. Communications of the ACM. 381 ... One of the major problems confronting the designer of a data system which is to support a relational model for its.Missing: assumption | Show results with:assumption
  21. [21]
    [PDF] SQL's Three-Valued Logic and Certain Answers
    This is often referred to as the closed-world assumption, or cwa semantics of incompleteness. [12, 21]. Another common semantics uses the open-world assumption, ...
  22. [22]
    Fixpoint semantics for active integrity constraints - ScienceDirect
    Active integrity constraints (AICs) constitute a formalism to associate with a database not just the constraints it should adhere to, but also how to fix ...
  23. [23]
    Logical Integrity Constraints - Teradata Vantage - Analytics Database
    The Closed World Assumption, or CWA, asserts that if an otherwise valid tuple does not appear in the body of a relation, then the proposition representing that ...
  24. [24]
    On indefinite databases and the closed world assumption
    Sep 7, 2005 · In this paper we investigate databases that consist of Horn and non-Horn clauses. We extend the definition of CWA to apply to such databases.
  25. [25]
  26. [26]
    [PDF] Databases under the Partial Closed-world Assumption: A Survey
    In this survey we review foundational and work on the partial closed-world assumption and then discuss work done in our group in recent years on various ...
  27. [27]
    [PDF] Nonmonotonic Reasoning in FLORA-2*
    from) the so called local closed world assumption [13]. In FLORA-2, a module is a container for a concrete knowledge base (or a part of it). Modules isolate.
  28. [28]
    Rethinking Knowledge Graph Evaluation Under the Open-World ...
    Sep 19, 2022 · Abstract page for arXiv paper 2209.08858: Rethinking Knowledge Graph Evaluation Under the Open-World Assumption. ... This closed-world assumption ...
  29. [29]
    On indefinite databases and the closed world assumption
    A database is said to be indefinite if there is an answer to a query of the form. PaVPb where neither Pa nor Pb can be derived from the database.
  30. [30]
    Non-monotonic Logic - Stanford Encyclopedia of Philosophy
    Dec 11, 2001 · Non-monotonic logic (NML) is a family of formal logics designed to model and better understand defeasible reasoning.
  31. [31]
    Saturation, nonmonotonic reasoning and the closed-world assumption
    To deal with the problem of implicit negative information in databases, it is necessary to use a nonmonotonic form of reasoning. The form of reasoning described ...
  32. [32]
    None
    Summary of each segment:
  33. [33]
    [PDF] From expert systems to machine-learned diagnosis models
    This closed world assumption and the structure of knowledge base lends itself to a simple but ... Medical diagnosis has been a prototypical application of AI that ...
  34. [34]
    [PDF] Local Closed World Semantics: Grounded Circumscription for OWL
    Closed World Assumption (LCWA). Most of these combinations are derived ... many decidable description logics around OWL. We will point out restrictions.
  35. [35]
    [PDF] Fuzzy Prolog with Default Knowledge - Sedici UNLP
    facts and rules of a program. For example, the Closed World Assumption (CWA) asserts that any atom whose truth-value cannot be inferred from the facts and ...
  36. [36]
    [PDF] A Logic-Based Analysis of Dempster- Shafer Theory - CORE
    9 The closed-world assumption has been formalized logically by Reiter [48]. ... Gordon, J., and Shortliffe, E., A method for managing evidential reasoning in a.<|control11|><|separator|>
  37. [37]
    [PDF] Classification-Reconstruction Learning for Open-Set Recognition
    closed-world assumption, that is, the training datasets are assumed to ... Although near-100% accuracy has been achieved in closed-set classification [4], the ...