Knowledge Graph-guided Agentic AI:
Bridging Reasoning and Explainability
in Healthcare
Montréal, Canada.

February 16 or 17, 2027  |  held in conjunction with AAAI-27

AAAI-27 Bridge Program

Bridge Program Overview

Knowledge Graph-guided Agentic AI: Bridging Reasoning and Explainability in Healthcare is an AAAI-27 Bridge Program exploring knowledge graphs as a shared interface between agentic reasoning and explainability, with healthcare as a central application. We bring together researchers and practitioners in AI, knowledge representation, causal reasoning, human–computer interaction, and healthcare.

Agentic systems plan, retrieve information, use tools, coordinate with other agents, and adapt. Understanding their decisions means connecting those actions to the evidence behind them. This bridge asks how we can make agent trajectories traceable, testable, and understandable to the people who rely on them. Graphs alone do not guarantee faithful explanations: their claims must be grounded in recorded events and tested against evidence.

The bridge combines tutorials, hands-on labs, research presentations, and interdisciplinary discussions to develop a shared graph vocabulary, evaluation and benchmark designs, and a community research roadmap. It is planned as a one-day, in-person event in Montréal during the February 16–17, 2027 Bridge Program window. The exact day will be announced.

Topics of Interest

We welcome contributions on the following six themes, with healthcare as a central application.

  • Graph semantics and scope. Represent agent goals, observations, plans, memory, tool calls, delegation, human approvals, safeguards, outcomes, and rejected alternatives. Define graph relationships that capture time, information flow, and causation.
  • Graph construction and provenance. Build reasoning graphs within agent architectures, reconstruct them from execution traces, or combine both approaches. Link explanatory claims to verifiable records while protecting private data and proprietary reasoning.
  • Causal and counterfactual validity. Test whether a graph captures causal evidence. Use interventions, replays, ablations, and alternative-plan tests to answer “why,” “why not,” and “what if” questions.
  • Multi-agent composition and attribution. Combine individual agent traces into a system-level explanation. Attribute decisions and outcomes across agents, humans, models, information sources, and tools when reasoning paths merge or conflict.
  • Human-facing explanations. Create consistent views of the same evidence for clinicians, users, developers, operators, and auditors. Design visual and conversational interfaces that help people understand decisions and take corrective action.
  • Evaluation and applications. Develop benchmarks for graph completeness, provenance integrity, causal faithfulness, stability, usability, calibrated trust, debugging value, latency, and storage cost. Study these measures in clinical decision support, scientific agents, and other consequential applications.

Important Dates

DateEvent
Call for papers opens
Submission deadline — 11:59 PM Anywhere on Earth (UTC−12)
Author notifications
Participant materials — tutorial and lab resources
February 16 or 17, 2027AAAI-27 Bridge Program — Montréal, Canada. One-day bridge; exact day to be announced.

All deadlines use AoE (UTC−12). See the AAAI-27 Bridge timetable. Add the submission deadline to your calendar.

AAAI-27 Bridge AI-Health Submission Link

Call for Papers & Participation

We invite position papers, extended abstracts, challenge statements, demonstrations, dataset or benchmark descriptions, research papers, and early-stage research contributions.

Submissions should describe the problem, its relevance to the bridge, the proposed contribution or perspective, and how the authors expect to participate in bridge activities.

Submissions will be evaluated based on relevance, clarity, interdisciplinary value, clinical or technical significance, and potential contribution to bridge discussions and shared artifacts.

At least one author of each accepted submission must register for AAAI 2027 and attend the bridge in person.

  • Position statements and open challenges
  • Demonstrations, datasets, and tools
  • Early-stage research and interdisciplinary ideas

Selection will emphasize disciplinary breadth and contributions to shared bridge outcomes. Selected contributions will seed short talks, posters, and collaborative working sessions.

Formatting, reference-page limits, review policies, and publication details will be announced with the submission link.

Tentative Program Schedule

This proposed one-day program will take place within February 16–17, 2027. The final date, room, speakers, and session times will be announced.

SessionActivity
Morning · Session 1A common foundation
Opening vision and tutorials on agent architectures, graph representations, and causal and human-centered explanation.
Morning · Session 2From traces to graphs
A hands-on lab, selected position talks, and a poster and mentoring session to connect ideas across disciplines.
Afternoon · Session 3Put explanations to the test
Counterfactual evaluation, multi-agent attribution, interactive graph explanations, and application roundtables.
Afternoon · Session 4Turn questions into shared work
An Oxford-style debate, benchmark and schema design sprints, working-group reports, and a community research roadmap.

Shared outcomes: a graph vocabulary, evaluation and benchmark designs, and a community research roadmap.

Bridge Chairs

Bridge Organizers

Participation Information

Who is this bridge for?

Researchers, practitioners, graduate students, and advanced undergraduates working in agentic AI, planning, knowledge graphs, explainable and causal AI, HCI, software assurance, and healthcare. Tutorials assume general AI literacy; you do not need expertise in every participating field.

What should I submit?

A two-page position statement, challenge, demonstration, dataset, or early-stage contribution related to the bridge themes. Detailed formatting instructions, reference-page limits, review and publication policies, and the submission portal will be announced here.

Will the bridge be in person?

Yes. The bridge is planned as a one-day, in-person event in Montréal during the AAAI-27 Bridge Program on February 16–17, 2027. The exact day and room will be announced. AAAI requires bridge presenters to attend in person.

Where can I find registration and materials?

Conference registration information will be available through the official AAAI-27 website. Bridge-specific participation instructions and links to tutorial slides, lab materials, and community resources will be posted here as they become available.

Contact

For questions about submissions, participation, or the bridge program, contact Zhong Chen at zhong.chen@siu.edu.