St. Jerome's University, Waterloo · November 13–16, 2026

Evidence Jam 2026
Causapalooza!

A weekend hackathon using AI to turn ecological literature into structured causal knowledge — one challenge, three lanes, one shared scoreboard.

Applications are now closed

See how it works

One CAMO edge, annotated

Mowing, timed just before female flowering intervention on Ambrosia artemisiifolia

causes · direct, contributing

Adult plant height falls

seed output is size-dependent

Fewer viable seeds set

Population growth rate (r) falls below replacement level in most cases

Randomised experiment Model-projected r vs. unmanaged controls Moderate certainty

A real edge from the Causal Mosaic Schema, annotated from Lommen et al. 2018. Solid boxes are CAMO nodes; dashed boxes are the mediation pathway the edge records. This is what every lane works with.

Ecology needs answers faster than a literature review can give them

What's the best way to manage invasive buckthorn? How does species reintroduction change an ecosystem? What actually improves restoration resilience? Answering these properly can take years of literature synthesis. We're going to try it in a weekend.

Six teams get real ecological data, fully-stocked GitHub repos, and roving mentors to break logjams. We have the Nibi supercomputer and GPU desktops on site for running local models. And food — good food, all weekend.

Who it's for

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

Team size

Six teams of 2–4 people. Participants applied as groups or solo, with solo applicants matched to teams. Interdisciplinary teams are strongly encouraged and prioritized in selection.

One challenge, three lanes

Evidence Jam runs as a single competition this year. Pick the lane that fits your team's skills — every project, in every lane, is judged on the same scale for the same prizes.

Lane 1

Turn papers into a causal graph

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

You get: The corpus, the schema, a few hand-annotated examples.
Rigor is measured by: Schema-valid output, checked against papers you haven't seen.

Lane 2

Reason over the claims

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

You get: A curated set of CAMO-structured causal claims, the schema.
Rigor is measured by: Accuracy on a held-out question set, honest confidence, graceful "I don't know."

Lane 3

Help a practitioner decide

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

You get: CAMO-structured evidence, existing wireframes — use them or don't.
Rigor is measured by: A real practitioner walking through your interface with a live decision.

Every lane, one rubric

Whichever lane you pick, your project is scored on the same five criteria. The strongest overall projects win prizes.

Rigor & CAMO fidelity Technical execution Practitioner impact Innovation Presentation

What is the Causal Mosaic Schema?

The Causal Mosaic Schema (CAMO) is a data schema written in LinkML — a fill-in-the-blanks form for documenting causal claims. If a researcher says "X causes Y," CAMO gives you a way to document exactly how, and how confident the researcher is that it's causal rather than merely correlational.

It was developed together with philosophers of causality, including Dr. Phyllis Illari, who wrote the book on causality. Every team gets a half-page primer and a worked example as part of their starter repo.

Lane 1 extracts causal claims into CAMO. Lane 2 reasons over CAMO-structured claims. Lane 3 helps practitioners act on them. Same schema throughout — which is what makes it worth collaborating across lanes.

Every repo includes

The CAMO schema
One worked example
A half-page primer
Environment setup, ready to go

Awards and prizes

One grand prize by random draw, two on merit, and one voted by the room.

💻

Grand prize · random draw, every participant, equal odds

Dell Pro Max with GB10

FCM1253 · 20-core NVIDIA GB10 Grace CPU, Blackwell GPU, 128 GB LPDDR5X, 4 TB SSD. One name drawn at random from every participant — not just team leads or winning teams. Everyone who takes part has the same chance to win, regardless of how their project scores.

🦝

First & second place

Gold Raccoon

One shared competition, one set of prizes. Every lane is scored on the same rubric, and the two strongest projects overall take home Dell hardware — not the strongest project in each category.

1^st Dell Pro P24 USB-C Hub Monitor P2426HE · 23.8″ FHD, 120 Hz, USB-C hub
2^nd Dell Pro Plus Earbuds EB525 · adaptive ANC, 33 h with case

🌿

Most sustainable use of AI · funding in progress

Green Raccoon

Goes to the team combining a genuinely strong project with the smallest measured footprint — local compute and hosted model API calls both count. Only projects that clear the same quality bar as the main competition are eligible; being efficient alone isn't enough to win. We're seeking dedicated funding for this prize now — details to follow.

⭐

Best collaborator · voted by all participants

Most Valuable Possum

Goes to whoever most enriched the weekend — helping other teams, sharing their work generously, and unsticking someone else's logjam.

Schedule

The DRAGEN Lab is open 9 a.m.–6 p.m. Saturday and Sunday. Four Nibi supercomputer checkpoints and two cross-team check-ins keep momentum going across the weekend.

Friday, November 13

1:00 pmIntroductory gathering, icebreakers, overview of the weekend

2:30 pmTour of Nibi & GPU workstations, HPC training

4:00 pmPrivate brief delivered to teams

5:00 pmEnd of day

Saturday, November 14

9:00 amWork rooms open · Nibi checkpoint 1

12:00 pmLunch + cross-team check-in 1

1:00 pmWork rooms open · Nibi checkpoint 2

6:00 pmEnd of day

Sunday, November 15

9:00 amWork rooms open · Nibi checkpoint 3

12:00 pmLunch + cross-team check-in 2

1:00 pmWork rooms open · Nibi checkpoint 4

6:00 pmEnd of day

Monday, November 16

1:00 pmPresentations, judging, prizes awarded

The space

Evidence Jam takes place in the DRAGEN Lab at St. Jerome's University — 3,600 ft² of purpose-built collaborative research space inside the newly renovated library, with a Makerspace offering 3D scanning, 3D printing, and high-performance computing.

Collaborative team work areas
GPU workstations for local model inference
Makerspace with HPC units & fabrication tools
Strong campus WiFi throughout

Breakfast and lunch are provided Saturday and Sunday, coffee from 9 a.m. each morning. We're seeking funding for an evening social — if it happens, we'll announce it. Dietary requirements are accommodated.

290 Westmount Road North, Waterloo, ON — a short walk from campus, accessible via ION light rail. sju.ca · dragenlab.ca

DRAGEN Lab Makerspace, St. Jerome's University

Supported by

See you in November!