Raccoon portrait — CC BY 2.0 GregTheBusker via Wikimedia Commons

Evidence Jam 2026
Causapalooza!

Using AI for causal synthesis of ecological knowledge

📍 University of Waterloo · St. Jerome's University 📅 November 13–16, 2026 · Friday to Monday 🌿 Nibi HPC · GPU Workstations

About

What is the best way to manage invasive buckthorn? How could species reintroduction change ecosystems? What improves ecosystem resilience? These are crucial questions in the restoration and conservation of biodiversity. Typically they would take years to answer through an exhaustive synthesis of the literature. We are going to try to answer them in a weekend.

Evidence Jam is a blitz-style hackathon that challenges teams to synthesize evidence using cutting-edge technology. This year we focus on causality. Six teams will be tasked with mapping the causes and effects in ecology to address three challenges: the use of LLMs for information extraction; the use of LLMs for question-answering; and the design and details of presenting such rich information.

Teams receive fully stocked GitHub repositories and real ecological data to work with. Roving mentors will be available throughout the weekend to break logjams and encourage collaboration across teams. We have access to the Nibi supercomputer and desktops with GPUs to run local LLM models. Most importantly: we have food.

Who Should Apply

  • Software developers & ML engineers
  • Restoration & conservation ecologists
  • UX/UI designers
  • Data scientists & knowledge engineers
  • Philosophers of science

Team Size

Six teams of 2–6 people. You may apply as an individual and be matched to a team, or apply as a group. Interdisciplinary teams are strongly encouraged and will be prioritized.

Tracks

Teams pick one track. Each is a single open problem. Two cross-team check-ins over the weekend encourage collaboration and knowledge-sharing between tracks.

Track 1

Turn papers into a causal graph

Given a corpus of restoration ecology journal articles, extract their causal claims into a graph. How much of the CAMO schema's richness you capture, and how you get there, is your call.

You get: the corpus, the schema, a few hand-annotated examples.

Judged on: accuracy, richness, cleverness, scalability
Track 2

Ask the graph

Build something that lets a restoration practitioner ask a question and get a trustworthy, evidence-grounded answer from a CAMO graph — without inventing anything the graph doesn't contain.

You get: a pre-built CAMO graph, the schema.

Judged on: usefulness, traceability, confidence, graceful "I don't know"
Track 3

Help a practitioner decide

Build an interface that helps someone planning a restoration make a decision from this evidence. What they care about: will it work, how sure are we, can it be undone, does it fit my site.

You get: evidence data, existing wireframes (use them or don't).

Judged on: real-world usefulness, honest confidence display

What is the Causal Mosaic Schema?

The core of this challenge is the Causal Mosaic Schema (CAMO) — a data schema written in LinkML that works like a fill-in-the-blanks form for documenting causal claims. If a researcher says "X causes Y," CAMO provides ways to document exactly how X causes Y and how certain the researcher is that the relationship is causal rather than correlational.

The schema has been collaboratively developed with philosophers of causality, including Dr. Phyllis Illari, who wrote the book on causality. All teams receive a half-page primer and a worked example graph as part of their starter repository.

Track 1 extracts causal claims into CAMO. Track 2 queries a CAMO graph. Track 3 helps practitioners act on CAMO-structured evidence. Each track builds on the same schema — making collaboration across teams both possible and valuable.

What's in the repos

  • The CAMO schema
  • One worked sample graph
  • A half-page primer
  • Environment setup

Track extras

  • Track 1: paper corpus, hand-annotated examples, held-out judging set
  • Track 2: pre-built CAMO graph
  • Track 3: practitioner evidence data, wireframes

Awards

Three award categories recognise the best solution in each track, sustainable AI practices, and outstanding collaboration.

🦝
3 AWARDS · ONE PER TRACK · AWARDED BY JUDGING TEAM

Gold Raccoon — Best Solution

Awarded to the team delivering the strongest solution in each track. Prize details to be announced.

🌿
1 AWARD · EVENT-WIDE · AWARDED BY JUDGING TEAM

Green Raccoon — Most Sustainable Use of AI

Awarded to the team best demonstrating efficient, responsible AI use — thoughtful model selection, minimal compute footprint, and a sustainable approach throughout the weekend.

1 AWARD · EVENT-WIDE · VOTED BY ALL PARTICIPANTS

Most Valuable Possum — Best Collaborator

Voted on by all individual project participants. Awarded to the person who most enriched the weekend through cross-team collaboration, knowledge-sharing, and generosity with their work.

Schedule

Physical access to the DRAGEN Lab runs from 9 a.m. to 9 p.m. on Saturday and Sunday. Teams may submit jobs to the Nibi supercomputer at four checkpoints across both days. Two cross-team check-ins encourage collaboration between tracks.

Friday · November 13

Introductory gathering, ice-breakers, overview of weekend format
Tour of Nibi supercomputer & GPU workstations · HPC training
Delivery of private brief
End of day

Saturday · November 14

Work rooms open · Nibi checkpoint 1
Lunch + Cross-team check-in 1
Work rooms open · Nibi checkpoint 2
End of day

Sunday · November 15

Work rooms open · Nibi checkpoint 3
Lunch + Cross-team check-in 2
Work rooms open · Nibi checkpoint 4
End of day

Monday · November 16

Presentations, judging, prizes awarded

The Space

Evidence Jam takes place in the DRAGEN Lab — the Medieval Digital Research in Arts and Graphical Environmental Networks Laboratory — at St. Jerome's University on the University of Waterloo campus. Founded in 2016, the DRAGEN Lab is a purpose-built digital humanities research space that deploys a collaborative, learner-centred approach to knowledge production across disciplines.

The lab occupies 3,600 ft² (242 m²) within the newly renovated St. Jerome's University Library. It was built to support exactly the kind of work Evidence Jam demands: expert and student researchers working side-by-side in teams, with access to high-performance computing, digital fabrication tools, and flexible collaborative space. It also includes a Makerspace with professional-grade 3D scanning, 3D printing, and high-performance computing units — a fitting home for a hackathon pushing the edges of AI-assisted ecological knowledge.

Photos via dragenlab.ca · St. Jerome's University, University of Waterloo.

Facilities

  • 3,600 ft² collaborative lab space
  • Makerspace with HPC units & fabrication tools
  • Flexible team work areas
  • GPU workstations for local model inference
  • Strong campus WiFi throughout

Food & Coffee

Breakfast and lunch is provided on both Saturday and Sunday. Coffee is available from 9 am each morning. We are also planning an evening social event — details to follow. Participants with dietary requirements will be accommodated.

St. Jerome's University is located at 290 Westmount Road North, Waterloo, ON — a short walk from the heart of the University of Waterloo campus and accessible via the ION light rail. Learn more at sju.ca and dragenlab.ca.

Register Interest

Applications will be open to teams of 2–6. Teams are selected based on the range and quality of skills represented. Each applicant provides a ~150-word biography; one member is designated as the corresponding contact. Selected teams complete a Team Charter before the event.

You may also apply as an individual — we will do our best to match you with a team that complements your skillset.

Ready to map the evidence?

Registration is managed via Luma. Click below to register your interest —
full applications will open once the date is confirmed.

Register on Luma →

Sponsors & Partners

Evidence Jam 2026 is looking for sponsors to support prizes, food, and infrastructure. Prize categories include the Gold Raccoon, Green Raccoon, and Most Valuable Possum. We are particularly interested in contributions of API credits, hardware, cash prizes, and software subscriptions.

If your organization is interested in sponsoring, please contact Tim.A@uwaterloo.ca.

Organised by

Tim Alamenciak
Carleton University & University of Waterloo

Infrastructure

Nibi — Sustainable HPC
GPU Workstations — On-site inference
University of Waterloo
St. Jerome's University (venue)

Support

Evidence Jam 2026 is made possible through the support of the following organisations, who provide infrastructure, venue, and knowledge systems for the event.