# Paying The Fact Creator: Validation Is A Cost Nobody Bears, And The Question Is Not Whether A Claim Is True But Whether This Is A Correct Use Of It **version** v0.33.54 **date** 31 July 2026 **from** Human (project lead) **to** Strategy, Commercial, Product **type** Strategy brief *Twenty-second of 31 July. The memo's own worked example is verified and corrected, and the research supporting its central observation is grounded and cited. Offered to be built on and challenged.* --- ## What This Is A commercial model for keeping evidence healthy, built on a distinction most fact-checking misses: **the corpus already holds that a stakeholder is accountable for a decision, and this memo adds that the party who produced the underlying fact, evidence, opinion or conclusion should carry a continuing responsibility for its accuracy, which nobody currently does because there is no mechanism and no money in it; validation is not free, since somebody's time is consumed every time a claim is checked, and inside a company that cost is absorbed by ordinary operating expense while outside one it is borne by nobody, which is why the public evidence base decays; the sharper move is that the question worth paying for is not whether a claim is true in itself but whether it is being used correctly in this context, for this conclusion, in this chain of reasoning, because the common failure is not fabrication but a sound finding applied to something it never supported, sometimes to conclusions the original work contradicts; the memo's own example proves the point better than it realises, since the ten-thousand-hours claim originates in a 1993 study of violin students where the figure was an average rather than a threshold, where half the top group had not reached it, where the mechanism was deliberate practice rather than time served, and where the students concerned were not yet experts, and the original author spent the rest of his career trying to correct the popularisation through books, articles and an open letter, which is precisely the absence of a mechanism this memo proposes to build; the graph-depth intuition is also documented, since a study of one biomedical belief traced a network of two hundred and forty-two papers carrying over two hundred thousand supporting citation paths and identified the conversion of hypothesis into fact through citation alone, alongside the practice of citing papers that do not quite say what the citer implies; and the model is a micropayment flowing from the party benefiting from a claim back to the party who produced it, in exchange for an authoritative statement about whether this particular use is sound.** It is the twenty-second document of 31 July (cross-ref: the v0.33.44 evidence-economy brief, the v0.33.44 evidence-packs brief, the v0.33.54 registrar-of-evidence brief, the v0.33.54 paragraph-is-a-folder brief, and the v0.33.53 every-paragraph-is-a-graph brief). New contributions: **continuing accuracy responsibility attached to the fact creator, the identification that validation cost is absorbed inside firms and unfunded outside them, contextual validity proposed as the billable question rather than truth, the ten-thousand-hours case worked as the canonical illustration, and the observation that the caching principle makes per-use validation economically viable.** ## The Missing Party The corpus's model attaches accountability to whoever accepts a risk. This memo adds a party that has so far been absent. The project lead: **"the stakeholder is accountable for the decision, but we also should make the fact creator, the entity, the system, the person that created the evidence or the opinion or the statement or the conclusion, they should also have the responsibility of accuracy over time."** That is a genuine gap. A decision cites evidence; the person deciding is accountable; and the person who produced the evidence has no continuing relationship with it at all. Once published, a finding is on its own, and whatever happens to it afterwards happens without its author. ## Validation Costs Money, And Only Companies Pay It The economic observation underneath is the load-bearing one. The project lead: **"every time we have one of these actions, it takes time, it costs money, because ultimately somebody's time is being paid, and it requires somebody to do a bit of work, so there isn't a neutral cost of the person who needs to validate that finding."** And the asymmetry that follows. The project lead: **"in companies that's okay, because that's already covered by the cost of the company operating, but in the real world at the moment we don't have that."** That explains something that otherwise looks like collective negligence. Inside a firm, someone is paid to check whether a claim supports a decision, because the firm bears the consequence. Outside, the same work benefits everyone and is paid for by no one, so it is done sporadically by volunteers, aggrieved original authors, and occasional journalism. The public evidence base is a commons with no maintenance budget, and it decays accordingly. ## The Question Is Not Whether It Is True Here is the distinction that makes this a different product from fact-checking, and it is the most important idea in the memo. The project lead: **"the angle is not just, is this statement correct in itself; the question is, is this correct in this context, for this use, for this conclusion, for this kind of set of events."** Existing verification asks whether a claim is true. That question is often easy and usually not the one that matters. The failure mode that actually causes harm is a true finding applied to something it never established. The project lead: **"you have a lot of stuff that was discovered in one way that gets used in other ways, ironically sometimes even contradicting the original study."** ``` CONVENTIONAL FACT-CHECKING CONTEXTUAL VALIDATION is the claim true? is this use of it sound? binary, one answer forever depends on the context each time answerable by a third party best answered by the originator mostly already done almost never done at all ``` The consequence is stated plainly and is worth carrying into any pitch. The project lead: **"you have a lot of decisions, a lot of people in the world impacted by decisions based on top of studies, or analysis, or events that actually never occurred like that."** ## The Worked Example, Checked The memo reaches for an example and asks for it to be verified, which is itself the process being proposed. The project lead: **"the 10,000 hours, which was done by the research on the violinists."** Checking it: the underlying work is a 1993 study of violin students at a Berlin music academy, and the popularisation was a 2008 book. The memo attributes the popularisation to a name the transcription garbled; the author is Malcolm Gladwell, and the researcher is Anders Ericsson. The memo's instinct that the popular version is wrong is correct, and the specific distortions are documented by the original author himself: - **It was an average, not a threshold.** The top group had accumulated around ten thousand hours by age twenty on average, and roughly half of them had not reached that figure. The popularisation reported it as though all of them had. - **The number is arbitrary.** The original author described it as catchy rather than meaningful, noting it could as easily have been eleven thousand. - **They were not yet experts.** The students were highly skilled and still students, admitted to a strong academy rather than winning international competitions. - **The mechanism was deliberate practice, not time.** Structured, effortful, feedback-driven work, which is a much narrower thing than accumulated hours. - **Nothing in the study supported the popular inference.** Showing that anyone could reach expertise with ten thousand hours would have required a different experiment entirely, taking randomly chosen people through it. The study showed only that among students already good enough to be admitted, the better ones had practised more. Individual variation is enormous, with reports of one chess player reaching master level in around three thousand hours and another requiring more than twenty thousand. So this is not a false claim being repeated. It is a true finding, correctly reported at its origin, applied to a conclusion it does not support, by a great many people making real decisions about how to spend years of their lives. ## What The Original Author Actually Had This is the part that makes the case for the model, and it is worth stating starkly. The researcher disagreed publicly and persistently. He wrote articles, gave interviews, published a book laying out the corrections, and wrote an open letter with a title that captures the frustration exactly: on the danger of delegating education to journalists. He spent a substantial part of his remaining career on it. **And none of it attached to the claim.** Every subsequent use of the ten-thousand-hours figure carried none of his objection, because there was no mechanism by which it could. The correction lived in different documents, findable only by someone already suspicious enough to look. The person best placed in the world to say *that is not what my study showed* had no channel, no standing at the point of use, and no economic reason to keep doing it beyond his own conviction. That is the absence this memo proposes to fill: not a way to publish a correction, which existed, but a way for the correction to be attached to the claim, at the moment of use, by the person with the authority to make it, and paid for by whoever is benefiting. ## The Graph-Depth Intuition Is Documented The memo describes an effect and the research names its mechanisms. The project lead: **"you have a couple of nodes where the evidence looks sound, but once you go down a couple of nodes further, like who said this and who said this, then sometimes it's false, or at best it should not be used the way it is being used."** A 2009 study in a major medical journal constructed the complete citation network for a single biomedical belief. It found **two hundred and forty-two papers, six hundred and seventy-five citations, and over two hundred and twenty thousand citation paths** supporting the claim. Analysing it identified three distortion mechanisms: - **Citation bias**, the preferential citing of supporting evidence and the neglect of papers that refuted or weakened the belief, particularly critical primary research. - **Amplification**, the expansion of the belief system by papers presenting no data addressing it at all. - **Invention**, including what the study calls the conversion of hypothesis into fact through citation alone. Commentary on the work adds a fourth that maps exactly onto this memo's contextual point: **citation diversion**, the citing of papers that say something relevant but not precisely what the citer implies they say. The distortions were also found extending into grant applications, sometimes used to justify funding requests. So the memo's intuition is not a hunch. **Authority accumulates through the network without anybody adding evidence**, and depth in the graph correlates with distortion rather than with confidence. That is the empirical case for validating the path rather than the node. ## The Model The proposal is a micropayment flowing from the beneficiary of a claim back to whoever can authoritatively judge its use. The project lead: **"the person who made the claim should have a revenue stream, and that should be connected to a revenue stream, or at least the opportunity to."** With the alternative. The project lead: **"or somebody, maybe even not that person, but somebody trustworthy; there should be an entity whose job is to verify this, that should be paid for it, and should keep it fresh."** And the economics stated as a principle. The project lead: **"the rewards of the person that is benefiting from that extra analysis need to trickle down to the people who actually did the original analysis."** Two design notes follow from work elsewhere in the corpus. **Caching makes it viable.** Per-use validation sounds prohibitively expensive because popular claims are used constantly. But the unit is not the use, it is the *pair* of claim and context. Once somebody has established that the ten-thousand-hours finding does not support a general claim about anyone reaching expertise, that answer is reusable by everyone making the same move. This is the content-hash principle from the paragraph-folder work: validate once per distinct pair, cache, and the marginal cost of the thousandth query approaches zero while the first one pays for it. **It is the same shape as the registrar.** The brief written earlier today on a registrar of evidence made the produce-once-consume-many argument for assessments. This is the same argument applied to claim validity, and the two would naturally live in the same place. ## Journalists And Research Entities The memo extends this beyond academia. The project lead: **"we also need to have a commercialisation model for journalists, and entities who do research, they should also have a way to confirm that research is correct and has been used in that particular way."** That is arguably the larger market. Journalism produces enormous quantities of factual material that is subsequently cited, aggregated, summarised and repurposed, almost always without any return to the originator and frequently without their context. A mechanism that let a publication assert, per use, whether a claim drawn from its reporting is being applied soundly would serve two purposes at once: a revenue stream attached to accuracy rather than to attention, and a check on the distortion that occurs downstream. It is also the part most likely to attract a partner, since the constituency is organised, motivated, and already looking for revenue models that do not depend on impressions. ## What This Does Not Try To Be - **Not fact-checking.** The question is whether a use is sound, not whether a statement is true. - **Not a correction service.** Corrections already exist and do not attach to the claim; attachment at the point of use is the product. - **Not dependent on the original author.** A trusted third party can hold the role where the author is unavailable or unwilling. - **Not free.** The whole argument is that validation costs money and currently nobody pays it. - **Not per-use pricing.** The unit is the claim-and-context pair, cached and reused. ## Honest Tensions | Tension | Note | |---------|------| | The buyer wants a yes | The party paying for validation is usually the party who wants the claim to hold, which creates pressure toward affirmation and is the central design problem | | Paying for validation versus paying for approval | The mitigation is to pay for the answer rather than the affirmative and to publish refusals, which is the audit model and carries the same well-known weakness | | The author as judge | Originators are not neutral about their own work and have an incentive to bless uses that spread it, which argues for a third party in at least some cases | | Authors die, retire, and decline | A model resting on the originator needs a succession answer, and the most contested claims are often the oldest | | Volume versus economics | Most claims are used far too often for individual adjudication, so the caching answer is doing a great deal of work and has not been tested | | A validation layer as a chokepoint | Any body that adjudicates which uses of evidence are sound acquires real power over what can be argued, which is a serious governance question rather than a footnote | | Contextual validity is a judgement | Whether a use is sound is frequently arguable, so the output is an authoritative opinion rather than a fact, and it must be presented as such | ## Open Questions | Question | Notes | |----------|-------| | Who pays, and at what moment? | The beneficiary is the natural payer and the least motivated to hear no, which may mean the payment has to be structural rather than voluntary | | How is a claim-and-context pair identified? | The technical crux, since caching depends on recognising that two uses are the same use | | What does a validation record look like? | Sound, unsound, or out of scope, with reasoning, and a date after which it should be revisited | | Who validates when the author cannot? | The succession model, and how a third party earns the standing the originator had | | How are refusals made visible? | A model that only publishes affirmations is worse than none, so negative results need a route to the point of use | | Does this attach to the corpus's own graph? | Every risk citing a source is an instance of the same problem, so the mechanism may be internal before it is a market | ## Relationship To Previous Briefs | Date | Document | Relationship | |---|---|---| | 5 Jul | `v0.33.44__strategy-brief__sg-send-evidence-economy-force-of-proof-fact-certification-two-prices-evidence-based-revenue-models.md` | Fact certification and evidence-based revenue; this supplies the contextual-validity refinement and a worked case | | 5 Jul | `v0.33.44__strategy-brief__sg-send-evidence-packs-as-a-service-agentic-api-sg-vaults-skills-model-on-demand-micropayments.md` | On-demand micropayments for small units of evidence work, which is the billing shape this needs | | 31 Jul | `v0.33.54__strategy-brief__sg-send-registrar-of-evidence-we-should-not-build-act-coverage-as-ai-washing-test-participant-cannot-own-register.md` | Produce once and consume many, applied there to assessments and here to claim validity | | 31 Jul | `v0.33.54__arch-brief__sg-send-paragraph-is-a-folder-filesystem-as-data-source-hashes-as-change-detection-stopping-rule-for-word-annotation.md` | The caching principle that makes repeated validation economically viable | | 28 Jul | `v0.33.53__arch-brief__sg-send-every-paragraph-is-a-graph-eu-ai-act-definitions-as-nodes-twins-as-hooks-with-concepts-appendix.md` | Meaning carried by connection, and the used-but-undefined finding, which is the same distortion inside a legal instrument | --- ## Key Claims | # | Claim | |---|-------| | 1 | The fact creator should carry a continuing responsibility for accuracy, which currently nobody does | | 2 | Validation consumes somebody's paid time, so it is never free | | 3 | Firms absorb that cost through operating expense; outside a firm nobody pays it, which is why the public evidence base decays | | 4 | The billable question is not whether a claim is true but whether this use of it is sound | | 5 | The common failure is a true finding applied to a conclusion it never supported, sometimes one the original work contradicts | | 6 | The ten-thousand-hours case is the canonical example: an average reported as a threshold, half the group below it, deliberate practice reduced to time served | | 7 | The original researcher spent years correcting it through books, articles and an open letter, and none of it attached to the claim | | 8 | Citation research documents the graph-depth effect, including the conversion of hypothesis into fact through citation alone | | 9 | It also names citation diversion, the citing of work that does not quite say what the citer implies, which is the contextual failure exactly | | 10 | The model is a micropayment from the beneficiary of a claim to whoever can authoritatively judge its use | | 11 | Caching by claim-and-context pair is what makes repeated validation economically viable | | 12 | Journalism is plausibly the larger market, and the constituency is already seeking revenue models not based on attention | --- ## Sources - The 1993 study of violin students, the 2008 popularisation, and the original researcher's own corrections, namely that the figure was an average rather than a threshold, that roughly half the top group had not reached it, that the number was arbitrary, that the students were not yet experts, that the mechanism was deliberate practice rather than accumulated time, and that nothing in the study supported the general inference drawn from it: https://www.salon.com/2016/04/10/malcolm_gladwell_got_us_wrong_our_research_was_key_to_the_10000_hour_rule_but_heres_what_got_oversimplified/ and https://www.6seconds.org/2022/06/20/10000-hour-rule/ and https://www.leapaheadapp.com/en/blog/10000-hour-rule-debunked - Reports of the enormous individual variation in time to expertise, and of the open letter written by the original researcher on the danger of delegating education to journalists: https://unhustle.substack.com/p/the-10000-hour-rule-myth-and-the and https://www.scotthyoung.com/blog/2024/01/23/10000-hr-rule-myth/ - The 2009 citation network analysis identifying citation bias, amplification, and invention including the conversion of hypothesis into fact through citation alone, across a network of two hundred and forty-two papers and over two hundred and twenty thousand supporting citation paths, with the distortions extending into grant applications: https://pubmed.ncbi.nlm.nih.gov/19622839/ and https://www.semanticscholar.org/paper/How-citation-distortions-create-unfounded-analysis-Greenberg/d860c6d3e941e1cb28fd9f899035fb926fba7747 - Commentary identifying citation diversion, the citing of papers that say something relevant but not precisely what the citer implies: https://scatter.wordpress.com/2009/07/28/its-nine-oclock-do-you-know-what-your-citations-say/ --- This document is released under the Creative Commons Attribution 4.0 International licence (CC BY 4.0).