Measuring contributions
Your contribution is valued by what it achieves, under rules anyone can read.
Approved by the Board for implementation, 20 November 2026. Counsel is confirming the open points, shown in [brackets]. CIX is not open yet. This Charter works under the law, our Certificate of Incorporation, our Bylaws and the CRIS AI Constitution; where they differ, those prevail. It will bind CRIS through the CIX participant terms once they are published.
In one paragraph
Part III measures a contribution by the value it creates, not by how many are made. Only verified contributions earn Value Points, under a method published before it applies and changed only on public notice. Points come from a score for quality, rarity and use, multiplied by base points for the kind of work. Verified work that is never used still earns a minimum. Every payout run is checked for fairness, gaming can be discounted under published rules, and an independent reviewer re-runs sample calculations each year.
The text
Part III is quoted in full. The formula in Article 7.3 is shown with its source kept exactly, and with a plain-words reading under it.
Part III — Measuring Contributions
Value is measured by what a Contribution achieves, not by how many are made. Part III sets what counts, how it is measured, how it is verified, and how the measure is kept fair.
Article 6 — What Counts as a Contribution
6.1 Kinds of Contribution. CRIS recognises the eight kinds of Contribution in Schedule A: observations, data, expert judgement, annotations and evaluations, corrections, code and tools, research and teaching material, and review and verification.
6.2 Effort and value. A Contribution earns Value Points for the value it creates. Where a task pays for time or effort as well, the task shall say so before the Contributor starts, and that payment is in addition to Value Points.
6.3 Unused Contributions. A Verified Contribution that is never used still earns the minimum Value Points for its kind set in Schedule A, so that Contributors are not penalised for CRIS's choices about what to use.
Article 7 — The Published Method
7.1 The Participation Value Meter. Value is measured by the Participation Value Meter, using a method published before it applies. The method shall be written in plain words, tested on historical or synthetic data before use, and changed only on public notice of at least [30] days.
7.2 Attribution methods. The Published Method shall use the attribution method suited to each product:
| Product | Method | How Value Points are earned |
|---|---|---|
| Verified datasets | Points per item | Each verified item earns points, weighted by rarity and quality |
| Evaluation panels | Equal shares | Each completed, verified evaluation earns an equal share, weighted by the expertise the task required |
| Model corrections | Measured improvement | Points in proportion to how much a correction improves the system's correspondence with reality, as measured by the Correspondence Prover |
| Large pooled studies | Shared credit | Credit shared by a published approximation of each Contributor's marginal effect |
7.3 The Contribution Value Score. Within each method, a Verified Contribution's Value Points are calculated as:
Reading the formula Your points = whether it was verified × the base points for this kind of work × a score for quality, rarity and use. A plain-words reading, not the Charter’s text.
where G is the verification gate (1 if verified, 0 if not); B is the base points the method assigns; and Q (quality), R (rarity or novelty) and U (measured use or improvement) are each scored from 0 to 1.5 under the Published Method. [The weights 0.40, 0.30 and 0.30 are proposed for adoption.]
7.4 Energy and compute. The Published Method may adjust U by the improvement achieved per Compute Thermodynamic Unit, so that rewards follow real progress rather than raw activity. The adjustment shall never reduce a Contribution's points below the minimum in Schedule A.
7.5 Rules in layers. The Board sets the principles (participants paid first, a published floor, fairness, publication). Below them sit the Published Method, then the reward rules (floor, fees, minimum payout, schedule), then each payout run. Each layer changes only through the layer above it, and a payout run is never changed afterwards; errors are corrected by a new, linked entry.
Article 8 — Verification and Provenance
8.1 Verified before valued. Only Verified Contributions earn Value Points. Verification checks a Contribution against other evidence and, where relevant, the Contributor's track record.
8.2 Provenance. Every Verified Contribution is recorded in the Knowledge Ledger with its Contributor (or pseudonym), date, task, consent record and Licence Tier.
8.3 Feedback. A Contribution that fails verification is returned with the reason, and the Contributor may correct and resubmit it.
8.4 Track record, not reputation alone. A Contributor's track record may weight verification but shall never be the only reason a Contribution is rejected.
Article 9 — Fairness and Integrity of Measurement
9.1 Fairness checks. Every payout run shall be checked for fairness by country, gender, language, disability, experience and over time. A failure shall be corrected in the next run and reported.
9.2 Language and access. Tasks shall be offered in the languages their Contributors use where practicable, and a Contribution shall not be valued lower because it was made in a language other than English or through a low-bandwidth channel.
9.3 Gaming. CRIS may detect and discount duplicate accounts, collusion, automated submissions and copied Contributions, under rules published in the Published Method.
9.4 Independent check. The Audit Council (Bylaws Section 6.8) or, until it is appointed, an independent reviewer shall sample payout runs and re-run their calculations at least once a year.
Quoted exactly from the CRIS Human Knowledge Contributor Rights Charter (version of 8 November 2026, approved by the Board for implementation on 20 November 2026). Headings, lists and tables are only laid out for reading; the figures are drawn on this site and are not part of the Charter.
In pictures
- QQualityScored 0 to 1.5× 0.40
- RRarity or noveltyScored 0 to 1.5× 0.30
- UMeasured use or improvementScored 0 to 1.5; may be adjusted per CTU (7.4)× 0.30
Article 7.2 assigns an attribution method to each kind of product. The Value Meter page explains the methods in more depth:
- Per verified item, weightedPoints per verified item, weighted by rarity and qualityGood forDatasets bought as a whole; simple to explainLimitRough for model improvements
- Marginal contributionHow much the result changes with and without a contributionGood forModel corrections and evaluationsLimitNeeds repeated tests
- Shared creditAverages each contribution’s effect across many combinationsGood forFair sharing across many contributorsLimitCostly to compute; approximations needed
The same figure appears on the Participation Value Meter page; here it illustrates Article 7.2, which assigns these methods to four kinds of product, of the proposed Contributor Rights Charter.
Article 7.5 sets the rules in layers, each changed only through the one above:
- 1Board principlesParticipants paid first; a published floor; fairness; publicationBoard, consulting the Public Benefit Office
- 2Published methodHow value is attributed and checkedContinuity Labs; counsel; Public Benefit Office
- 3Reward rulesFloor, fees, minimum payout, scheduleParticipation Platforms; counsel
- 4Each payout runCalculations under the rules in forceChecked by fairness checks and an audit sample
The same figure appears on the Participation Value Meter page; here it illustrates Article 7.5 of the proposed Contributor Rights Charter.
Article 9.1 sets the fairness checks on every payout run:
- By countryPaid in line with the value created
- By genderSimilar records, similar pay
- By languageLess common languages valued fairly
- By disabilityNo one valued lower for a disability
- By tierNewcomers can progress
- Over timeStable, explained payouts
The same figure appears on the Participation Value Meter page; here it illustrates Article 9.1 of the proposed Contributor Rights Charter.
In practice
For contributors, Part III means a well-made, rare or widely used contribution earns more than many careless ones (Article 6.2), and nothing is valued lower because it was made in a language other than English or over a slow connection (Article 9.2). If a contribution fails verification you get the reason and can resubmit it (Article 8.3).
For CRIS staff, the Participation Value Meter is where the method runs; its page describes the system in full, so this page does not repeat it. A payout run is never edited afterwards; an error is corrected by a new, linked entry (Article 7.5).
For buyers, Article 8.2 means every verified contribution in a dataset carries its provenance: contributor or pseudonym, date, task, consent record and licence tier.
