[Summary]    [Vision]    [Call for papers]    [Organization]    [Submission]    [Programme]    [Registration]

Workshop on Ontology and Semantics of Explanations and Trustworthy AI

OSETAI 2026 is a workshop dealing with the formal characterization, representation, and evaluation of explanations and trust. The workshop is part of JOWO 2026. Explainable and trustworthy AI raises foundational questions about the nature, structure, and semantics of explanations and trust claims, which fall squarely within applied ontology and formal semantics. Trust and trustworthiness are not solely properties of models or explanations; they are socially grounded, emerging from institutional contexts, shared norms, accountability structures, and sustained relationships between agents. Explanations and validity statements or certificates are metadata associated with models and data.

Limitations on the range of applicability, uncertainty documentation, and expected deviations from experimental data or ground truth must be documented as well, and it needs to be established why such metadata are trustworthy, and to what extent. A related applied ontology issue concerns the epistemic status of explanations. In safety-critical, regulatory, and legal contexts, explanations and trust claims must be auditable, reproducible, and formally grounded, so that models, decisions, and derived evidence can withstand scrutiny in settings such as digital forensics, certification, and judicial admissibility. OSETAI addresses opacity, transparency, and interpretability not merely as technical properties of models, but as relational and semantic notions, opaque or interpretable with respect to specific agents, purposes, contexts, and forms of interpretation.

Topics of interest

  • Abductive reasoning and the generation of explanations and justifications
  • Applications of formal ontology to explainability and trustworthiness
  • Causal thinking, reasoning, and modelling
  • Connecting theory and ontologization of explanations to regulatory compliance and industrial impact
  • Digital forensics and formal documentation of evidence admissible in court
  • DIKW hierarchy and epistemic levels of explanation: Data, knowledge, information, and wisdom
  • Explainability in symbolic reasoning and documentation of formal proofs
  • Explanation formats exploiting domain knowledge
  • Factual and counterfactual explanations
  • Large language models from the perspective of explainability and trustworthiness
  • Nature of the entities to be explained (explananda)
  • Ontological and semantic perspectives on holistic versus component-wise validation of models and systems
  • Opacity and transparency, interpretability and interpretation
  • Reasoning with explainable neural models
  • Responsibility, trust, trustworthiness and reasons/motivations for trusting
  • Social aspects and sociology of trustworthiness and relationships of trust
  • Taxonomies and ontologies of explanations
  • Use of explanations for debugging knowledge-based systems and AI models

The scope of the workshop is also described in the OSETAI 2026 vision statement.

Manuscript types

We welcome original submissions of the following types.

  • Full paper: 12 to 18 pages
  • Short paper: 6 to 9 pages
  • Abstract only: 2 to 3 pages

Submissions must adhere to the one-column LaTeX CEURART style. All page limits include references. The short papers can e.g. also be position papers, summaries on research in progress, or practitioner reports. Use the EasyChair set up by FOIS 2026 to submit your manuscript. Full papers and short papers will be included in the proceedings, while contributions submitted as abstract only will not.

DigiPass CSA is funded from the EC's Horizon Europe research and innovation programme under GA no. 101137725.