One-day bridge event AAAI 2027

ClimateSignal

Bridging Climate Science and AI to Counter Mis- and Disinformation

How can AI systems verify and communicate climate claims with scientifically appropriate evidence, provenance, context and uncertainty? A cross-disciplinary Bridge built around learning, hands-on auditing and co-design.

Montréal, Canada · February 16–17, 2027 (one day, exact date TBC) · In person

Date16–17 February 2027One day within this window · TBC
LocationMontréal, CanadaAAAI-27 · room to be assigned
FormatOne full dayFully in person · 09:00–17:00
Community~40–60 participantsResearchers, students & practitioners
Planning status: Confirmed
About the Bridge

A shared language for climate information integrity.

Climate evidence is complex. AI can help people understand it — but it can also reproduce misleading narratives, omit important context or make uncertain conclusions sound definitive.

ClimateSignal brings together the people who produce climate evidence, develop AI systems, fact-check claims and communicate scientific findings. The goal is to turn an urgent challenge into a shared research agenda with common concepts, evaluation methods and lasting collaborations.

Unlike a paper-centred workshop, the Bridge combines accessible disciplinary primers, a guided claim-auditing lab and collaborative studios. Participants learn together in the morning and build practical research directions together in the afternoon.

Learn together. Test together. Specify together. Maintain together.

Climate & Earth-system science

Observations, projections, attribution, baselines, spatiotemporal scope and uncertainty.

AI & computer science

NLP, multimodal systems, retrieval, foundation models, robust evaluation and AI safety.

Information integrity & society

Fact-checking, communication, misinformation research, social science, economics and policy.

Research agenda

Questions we want to tackle.

The Bridge centres on challenges that need both climate-science expertise and AI or information-integrity perspectives.

01 / REPRESENTATION - Claims, context & uncertainty

How do we represent climate claims together with their geographic and temporal scope, provenance, evidence and causal attribution?

02 / VALIDATION - Grounding in evolving science

How should AI retrieval and fact-checking systems handle changing evidence, scientific sources, projections and generated summaries?

03 / ROBUSTNESS - Multimodal mis- and disinformation

How can we evaluate misleading plots, synthetic media, selective baselines, outdated findings and adversarial framing?

04 / COMMUNICATION - Human-centred explanations

How can AI communicate limitations, calibrated uncertainty and useful context without further amplifying misleading claims?

Programme

TBD

What we hope to create

Outputs designed to outlast the event.

These are proposed community-created drafts — not completed standards, datasets or tools.

A shared claim–evidence schema

Common fields for provenance, evidence, spatiotemporal scope, uncertainty and interpretation.

A living benchmark blueprint

Requirements for multilingual, multimodal climate-claim evaluation with time-aware evidence.

A practical reporting checklist

Guidance for evidence traceability, limitations and responsible communication in climate-facing AI.

Participation

Bring your perspective to the table.

The Bridge Event is designed for climate and AI researchers, fact-checkers, social scientists, communication and policy specialists, practitioners, and students at different career stages. No specialist knowledge of every field is expected.

Participants will be invited to share short non-archival position statements, climate-claim cases or demonstrator abstracts. The aim is constructive cross-disciplinary exchange, not a collection of finished papers.

Participation instructions and any Bridge-specific deadlines are not yet confirmed.

Our team

Organising committee.

Bringing together expertise in trustworthy AI, climate NLP, misinformation and societal impacts.

Lead organiser

Eleonore Vissol-Gaudin

Fujitsu Research of Europe · UK

AI security and anti-disinformation; coordination, cross-domain teaching and provenance-focused research.

Organiser

Rahul Mishra

Fujitsu Research of India · India

Secure and trustworthy AI, NLP and misinformation; climate-claim clinic and robust evaluation.

Organiser

Maged Ali

Essex Business School, University of Essex · UK

AI, sustainability and social impacts; practitioner engagement and responsible communication.

Organiser

Markus Leippold

University of Zurich · Switzerland

Climate NLP, sustainable finance and climate evidence; case annotation and benchmark design.