OXAI 2026
Call for Contributions
How can we make the behaviour and decisions of AI systems observable, explainable, and accountable?
OXAI 2026 brings together researchers, practitioners, and other stakeholders to advance the state of the art and best practices in observability and explainability of AI systems. Observability provides the measurements, traces, and runtime evidence needed to understand internal states and behaviours. Explainability turns that evidence into understandable accounts of automated decisions, supporting trust, accountability, and compliance.
Building on OXAI 2024, this edition focuses on agentic AI systems that plan, use tools, maintain memory, and adapt under competing objectives. Particular attention will be given to explainable multi-objective optimization and to making trade-offs among quality, latency, cost, safety, privacy, security, sustainability, fairness, and accessibility observable and understandable.
We invite contributions on methods, architectures, tools, evaluations, governance mechanisms, and engineering practices. Our system-level perspective connects models and services with software architecture, cloud computing, data management, testing, operations, and responsible AI.
Topics of interest
Topics include, but are not limited to:
- Observability in AI and Agent Systems: monitoring, logging, tracing, provenance, security telemetry, incident response, AgentOps/LLMOps/AIOps, and analysis of plans, memory, tool calls, interactions, and side effects.
- Explainability in AI Engineering: methods for explaining models and system behaviour, human-centred and accessible explanations, causal diagnosis, and metrics for explanation quality.
- Explainable Multi-Objective Optimization: agent planning, learning, and runtime adaptation under competing objectives; Pareto trade-offs; stakeholder-dependent preferences; and explanations of constraints and decisions.
- Human-Centred Quality Attributes and Governance: integrating equity, diversity, and inclusion (EDI) by design, fairness, and accessibility into stakeholder requirements, data, interfaces, and evaluation, while addressing privacy, regulatory compliance, oversight, recourse, contestability, and societal impacts of opaque AI systems.
- Case Studies, Tools, and Applications: empirical studies, benchmarks, demonstrations, experience reports, and industrial and cybersecurity applications of observable and explainable AI.
Participation and workshop format
We welcome academic scholars, industry professionals, government researchers, and community-facing practitioners, including data scientists, machine learning engineers, software architects and developers, AIOps/DevOps practitioners, and researchers in responsible AI, cybersecurity, and governance. Participants at all experience levels are welcome.
The planned full-day workshop combines invited presentations, community talks and demonstrations, a panel, and a facilitated breakout on competing stakeholder requirements and quality attributes. Participants will connect these requirements to observable system evidence and evaluation criteria. Planned outcomes include a concise engineering roadmap/checklist and a post-workshop activities-and-findings summary.
Important Dates
- Talk abstracts and papers
- (AoE)
- Author notification
- Final presenter materials and camera-ready papers
- Workshop date
- CASCON 2026
- November 9–12, 2026
AoE means Anywhere on Earth (UTC−12). OXAI takes place on the second day of CASCON programming, November 11. CASCON runs November 10–12, preceded by CSER on November 9.
How to Submit
We accept two submission types:
- Talk abstract: up to 2 pages, outlining a proposed talk or demonstration.
- Paper: up to 4 pages, presenting research, work in progress, new ideas supported by evidence, experience reports, or industrial case studies.
Both limits include all text, tables, figures, references, appendices, and other content.
Submit a single PDF in IEEE double-column conference format (24pt title, 10pt body text). For LaTeX, use \documentclass[10pt,conference]{IEEEtran} without compsoc or compsocconf. Submissions are not anonymous: include the title and all authors’ names, affiliations, email addresses, and public profile links in both talk abstracts and papers.
Following IEEE guidance on AI-generated content, disclose such content in the acknowledgments, identifying the system used and affected sections, and cite the system where applicable. At least one author of each accepted contribution must register and present at the workshop.
Submit through Microsoft CMTSign in to CMT or create a free account, then select Create new submission in the Author Console. Upload your PDF (maximum 10 MB), within the page limit for your submission type. For questions, contact the organizers.
Originality: Talk abstracts may describe previously published work; cite and link to the prior publication. Papers must present original, previously unpublished work. Ongoing and industrial work is welcome in either format, subject to these requirements.
Publication: Accepted talk abstracts and papers will be included in the CASCON 2026 proceedings.
Organization
Workshop organizers
- Brock University, Canada
- University of Ottawa, Canada
- Concordia University, Canada
- Concordia University, Canada
Questions and enquiries
For questions about the workshop, please contact the organizers listed above. Please submit contributions through Microsoft CMT.
Past Editions
OXAI 2024 — The 1st International Workshop on Observability and Explainability of AI Systems, held at CASCON on November 12, 2024, in Toronto, Canada.