Trust is the product. Here's the methodology, the confidence system, the collaboration, and everything we don't know.
Every claim on this site carries a confidence flag: ⬤ strong (peer-reviewed or government-published), ◐ moderate (single study, advocacy-reported, or dated), or ○ flagged (extrapolation, with the assumption stated). This exists because of how the site was made: an AI can misplace confidence, and a human can't check everything, so instead of implying certainty by tone, we declare it per claim. Click any chip to see the source, its type, and when we retrieved it. If you can't verify a claim in under sixty seconds, we've failed and want to know.
Fair question: AI runs on data centers, and data centers are thirsty. We won't dodge that; the footprint is real and belongs in any honest accounting. Here's the other side of the ledger: the actors organized around capital, including the industries whose plastic fills these lakes, are adopting AI without hesitation, to extend reach and produce faster. Mission-driven organizations cannot afford to abdicate the same leverage. And beyond leverage: the people who will sit in the rooms deciding how water resources are used by and for this technology need tactile knowledge of it, not received opinions. You cannot legislate what you have never touched. Abstention carries costs of its own: leverage conceded, and a seat at the table forfeited.
This project was built by a human researcher and an AI collaborator, deliberately and transparently. The AI did the wide synthesis: gathering public sources, cross-referencing, drafting, structuring data. The human did what humans do: direction, judgment, domain instinct, and, repeatedly and catch by catch, the correction of AI overreach. (Example: an early draft claimed organizations were "unaware" of each other's aligned asks. The human caught it: practitioners almost certainly know; what's missing is a public, durable record. The claim you'll find on this site is the corrected one.) Every substantive claim survived a human read for what the field calls face validity and we call the bullshit check.
This synthesis is an April–July 2026 snapshot of public artifacts, and it has edges we want visible. Some of the region's best data is public in principle but gated in practice: the Alliance for the Great Lakes' Adopt-a-Beach dataset is available by email request, and EPA's GLENDA requires a free Central Data Exchange account (cdx.epa.gov); both access routes confirmed, neither yet incorporated. No community-owned Indigenous datasets are incorporated, by design; information about tribal organizations is limited to what those organizations publish publicly about themselves. Under the CARE Principles for Indigenous Data Governance, the architecture holds space for Indigenous data under Indigenous authority, if and when nations choose to bring it, and not before. We did not attempt a Canadian provincial deep-dive, so Ontario's policy layer is thinner than the eight U.S. states'. And no economic modeling appears anywhere in this work, because none specific to Great Lakes plastic pollution could be found to cite, an absence Section 4 of the report names as a gap in the field itself. Insider knowledge always exceeds public artifacts; where we say something "does not exist," we mean no public evidence of it was found, and the corrections door below is open.
Every live source chip on this site draws from one registry, rendered here directly from the same file the chips read (data/sources.json; download it, it's yours). The report's fuller numbered source list lives in Appendix A, with citation detail and confidence grades for every reference.
When this site says a program, channel, or map "does not exist," it means we found no public evidence of one. Insider knowledge always exceeds public artifacts. If you are the counterexample, if the thing we called missing is your job, we would genuinely love to hear it, and the data updates within days: one edit, one file, everywhere at once. To send a correction or addition, open an issue on the project repository: github.com/shaycranmer/thewholeboard/issues.