2026-07-10 –, Birch Salon
The Ethical Technology Assessment (ETA) framework, Inspired by the Grassroots innovation assembly for Agroecology (GIAA) and developed through OpenTEAM and the Agricultural Knowledge Concordance, is a community-owned scoring practice for technology that works more like Consumer Reports or Human Rights Watch field documentation than a compliance badge — with three confidence tiers, claim-level evidence threading, structured dispute resolution, and automatic confidence decay that forces re-review. This five-minute lightning talk introduces the framework, demonstrates the live single-page assessment tool covering Hardware, Software, Datasets, Indigenous Data Sovereignty, and Seeds & Germplasm, and leaves the audience with one clear next action: score a tool and contribute to a score in their own community before the camp ends.
The framing: most technology-ethics frameworks deployed at scale produce badges, not accountability. A vendor pays a certifier; the certifier issues a passed-or-failed mark; the mark is valid for a fixed period regardless of how the tool changes; disputes go back to the certifier who has a financial relationship with the assessed tool. The resulting framework is legible in the sense that procurement officers can check a box, but it is not trustworthy in any meaningful sense. The DWeb community knows this pattern from the supply-chain audit industry, from the carbon-credit industry, from the privacy-policy industry. Each of these produces compliance theater — the appearance of governance without the substance.
The Ethical Technology Assessment (ETA) framework was built to displace this pattern in the agricultural-knowledge commons specifically, and the architecture transfers cleanly to any commons that needs an honest-and-durable way to ask should we use this tool? The framework is documented as ETA Framework v1.0 (February 2026, OpenTEAM / Ag Knowledge Concordance, CC BY 4.0) and runs on a single-page assessment tool that opens in any browser, with no install, no account, and no cloud dependency. The lightning talk introduces the framework's three sharp design choices, shows the tool running live for approximately ninety seconds on a real tool the audience names from the floor, and closes with three concrete next actions.
Design choice one — three confidence tiers with structural decay. Tier 1 (Provisional / steward self-assessment) has its score penalized by a 0.92 confidence modifier. Tier 2 (Peer-reviewed / at least one certified peer reviewer has verified or challenged claims) is at face value. Tier 3 (Audited / a designated ecosystem steward — OpenTEAM, AKC, or Farm Hack — has completed a structured audit) gets a 1.04 boost capped at 100. A self-assessed score of 88 displays as 81. After peer review, it shows as 88. After full audit, it shows as 92. The confidence tier is as important as the score — a Tier 3 score of 78 is more trustworthy than a Tier 1 score of 90. Confidence decays: any card not re-reviewed within 18 months (12 for datasets) automatically drops one tier and receives a Needs review public flag. The framework creates its own renewal cycle, without manual administration.
Design choice two — claim-level evidence threading and structured dispute resolution. Rather than treating the assessment as a single monolithic score, each individual claim is separately reviewable and attributable. Every gate result, score bar, and flag has an attached evidence thread: who claimed it, when, on what evidence, and what its current status is (Confirmed / Contested / Under review / Superseded). When a reviewer disagrees with a claim, they open a dispute on the specific claim. The steward has 30 days to respond. If unresolved, an ecosystem steward arbitrates. The card is publicly flagged Contested claim throughout. Contested claims are not hidden — a card with two open disputes and Confidence 2/3 is more trustworthy than an unchallenged Confidence 1/3 card, because it demonstrates the community is actively engaging with the evidence.
Design choice three — one framework, five card types. The same four-gate entry structure (G1 Transparency, G2 Rights & Control, G3 No Abuse, G4 Safety & Stewardship) anchors Hardware ETA, Software ETA, DTAP — Data Trust Assessment Protocol (six trust dimensions: Consent rigour, Scientific validity, Interoperability, Discoverability, Longitudinal integrity, Governance), Guardian Connector (Indigenous data sovereignty, four pillars aligned with CARE Principles), and Seeds & Germplasm (four pillars derived from the Wheat Workers' Code of Ethics). The cards differ in the specific questions asked within each category, but the architecture is shared, the scoring is unified on a 0–100 scale, and the assessment lives on the same Pi 5-hostable infrastructure.
The live tool demo. Approximately 90 seconds in the middle of the talk. The unified ETA assessment tool — Hardware / Software / DTAP / Guardian Connector / Seeds — opens in any browser. The speaker asks the audience to name a tool. The speaker scores it on the four binary gates in real time, then walks two category scores, then commits the result with a confidence-tier badge to a shared Git repository on the projector. The audience sees the framework run in operation in under two minutes.
The three next actions, at the close.
Score a tool you actually use. The unified ETA tool URL is in the printed handout and on the QR code in the lower-right of the final slide. Take three minutes, score one tool, commit it to your community's Git repository or to the OpenTEAM / AKC public ETA registry.
Become a certified peer reviewer. The one-hour async orientation is on the OpenTEAM and AKC platforms. The certification requires familiarity with the card type, vouching by an existing reviewer, and agreement to the 30-day dispute-response standard.
Fork the framework for your domain. The framework is CC BY 4.0 and the question sets are open. Communities outside agriculture have already begun forking the structure for open-source software more broadly, for citizen-science datasets, and for community-owned hardware lending libraries. The structure works.
The deeper sessions for participants who want more. GIAA Sessions (the full 60-minute workshop), the Farm Hack Box Technical Workshop (Submission in the longer GIAA submission set), and the GIAA / Farm Hack Box Federated Tool Library session (Submission). The lightning talk is the on-ramp; the workshops are the working sessions.
This is a Joint proposal - but happy to provide my Bio in addition to those of the other participants. - Dorn Cox is a Farmer, Author and Researcher and the editor of the Talk to the... Handbook and has been a steward of Farm Hack and the OpenTEAM federated infrastructure for nearly a decade. He convenes the GIAA Infrastructure Working Group and is the primary practitioner refining the question-design discipline that the tool implements.
Anamika Dey (Honey Bee Network / SRISTI / GIAN) has spent more than a decade documenting and stewarding the Indian grassroots innovation archive — over a million records across three federated databases, with a working multilingual AI agent in production. Caroline (Schola Campesina) carries the European agroecology and CSIPM coordination thread. David Otieno (Kenyan Peasant League) provides the ground-truth demonstration of how a sovereign database is actually used at the mobile-phone ceiling of accessibility.
SJ Klein leads the Agricultural Knowledge Concordance at Code for Science & Society (CS&S), and active in the Harvard Berkman Klein Center for Internet & Society, with thought-partner networks across Wikimedia, Internet Archive, Code for Science & Society (CS&S), and the AI Commons House. SJ's background in mass-scale community-edited knowledge infrastructure — Wikipedia, Wikidata, and the broader open-knowledge tradition — is the discipline the AKC adapts from millions-of-contributors editing patterns to deep-interview stewardship patterns. Wikipedia is what happens when a community owns its own editorial discipline at scale; the AKC and the tool that supports it are what happens when that ownership extends from the textual content to the knowledge-mapping practice itself.