SOURCES & EVIDENCE
The audit trail behind The Altruist Party's claims.
Open Governance is one of our three foundations. A movement that asks institutions to make their reasoning auditable has to do the same with its own. This page exists so that every factual claim we make can be traced to its source, checked, and challenged. Where a number is contested, we say so. Where the evidence cuts against us, we say that too. We would rather lose an argument honestly than win one on a figure that does not hold up.
This is an initial audit. It currently covers the headline claims on our own front page and the comparative-precedent base we draw on. A full claim-by-claim registry covering every figure on every page is in progress, not finished, and we say so rather than imply completeness.
Every figure below is dated. Figures go stale. When a source updates, this page updates, and the old number is corrected rather than quietly left in place.
Open Governance is one of our three foundations. A movement that asks institutions to make their reasoning auditable has to do the same with its own. This page exists so that every factual claim we make can be traced to its source, checked, and challenged. Where a number is contested, we say so. Where the evidence cuts against us, we say that too. We would rather lose an argument honestly than win one on a figure that does not hold up.
This is an initial audit. It currently covers the headline claims on our own front page and the comparative-precedent base we draw on. A full claim-by-claim registry covering every figure on every page is in progress, not finished, and we say so rather than imply completeness.
Every figure below is dated. Figures go stale. When a source updates, this page updates, and the old number is corrected rather than quietly left in place.
Part A — Our own headline claims, audited first
Before we ask anyone to trust our reading of the wider world, here is our reading of the three numbers on our own front page. Two of them needed correcting against the most current primary sources. We are showing that work rather than hiding it, because that is the entire point of this page.
Claim 1 — Public trust in government
What the strongest current source says. As of September 2025, 17% of Americans said they trust the federal government in Washington to do what is right just about always or most of the time, down from 22% a year earlier. This is one of the lowest readings since the question was first asked in 1958, when it stood at 73%.
Source: Pew Research Center, Public Trust in Government: 1958–2025 (Dec. 2025). https://www.pewresearch.org/politics/2025/12/04/public-trust-in-government-1958-2025/
Claim 2 — Nonvoters in the last presidential election
What the strongest current sources say. About 245 million Americans were eligible to vote in 2024 (voting-eligible population). Roughly 156 million ballots were cast. That leaves an estimated 89 million eligible nonvoters by the University of Florida Election Lab's measure (about 36% of the eligible population). Other credible estimates put the figure near 86 million (Environmental Voter Project) or in an 80–90 million band depending on whether the denominator is voting-eligible population, voting-age population, or registered voters.
Sources: U.S. Census Bureau, 2024 Presidential Election Voting and Registration Tables (Apr. 2025), https://www.census.gov/newsroom/press-releases/2025/2024-presidential-election-voting-registration-tables.html ; University of Florida Election Lab figures as reported by U.S. News (Nov. 2024), https://www.usnews.com/news/national-news/articles/2024-11-15/how-many-people-didnt-vote-in-the-2024-election
Claim 3 — Government waste available to fund civic infrastructure
What the strongest current source says. Federal agencies estimated about $186 billion in improper payments in fiscal year 2025 (up from $162 billion in FY2024), and cumulative improper payments since 2003 total roughly $2.8–3 trillion. Separately, GAO estimates direct annual financial losses to the government from fraud at between $233 billion and $521 billion. These are figures from the Government Accountability Office, an independent, nonpartisan body, not from us.
Sources: GAO, Payment Integrity: Agencies' Estimated Improper Payments Increased to $186 Billion in Fiscal Year 2025 (Apr. 2026), https://www.gao.gov/products/gao-26-108694 ; GAO, FY2024 improper payments ($162B), https://www.gao.gov/press-release/gao-reports-estimated-162-billion-improper-payments-across-federal-government-fiscal-year-2024 ; GAO fraud estimate ($233B–$521B), https://www.gao.gov/products/gao-25-108172
Note on the Home page. Our front page says "tens of billions of dollars in waste, improper payments, and untracked expenditures." This is understated, not wrong; the actual figure is in the hundreds of billions per year. We could strengthen it to GAO's current numbers. But one honesty caveat is binding, and it governs our funding model: money identified as improper is not the same as money recovered. Recovered amounts are a small fraction of identified amounts. Our governance and funding model depends on recovered waste, so any use of these figures must say plainly that identification is the ceiling, not the deposit. Citing $186 billion implies nothing about what a recovery program would actually return.
Before we ask anyone to trust our reading of the wider world, here is our reading of the three numbers on our own front page. Two of them needed correcting against the most current primary sources. We are showing that work rather than hiding it, because that is the entire point of this page.
Claim 1 — Public trust in government
What the strongest current source says. As of September 2025, 17% of Americans said they trust the federal government in Washington to do what is right just about always or most of the time, down from 22% a year earlier. This is one of the lowest readings since the question was first asked in 1958, when it stood at 73%.
Source: Pew Research Center, Public Trust in Government: 1958–2025 (Dec. 2025). https://www.pewresearch.org/politics/2025/12/04/public-trust-in-government-1958-2025/
Claim 2 — Nonvoters in the last presidential election
What the strongest current sources say. About 245 million Americans were eligible to vote in 2024 (voting-eligible population). Roughly 156 million ballots were cast. That leaves an estimated 89 million eligible nonvoters by the University of Florida Election Lab's measure (about 36% of the eligible population). Other credible estimates put the figure near 86 million (Environmental Voter Project) or in an 80–90 million band depending on whether the denominator is voting-eligible population, voting-age population, or registered voters.
Sources: U.S. Census Bureau, 2024 Presidential Election Voting and Registration Tables (Apr. 2025), https://www.census.gov/newsroom/press-releases/2025/2024-presidential-election-voting-registration-tables.html ; University of Florida Election Lab figures as reported by U.S. News (Nov. 2024), https://www.usnews.com/news/national-news/articles/2024-11-15/how-many-people-didnt-vote-in-the-2024-election
Claim 3 — Government waste available to fund civic infrastructure
What the strongest current source says. Federal agencies estimated about $186 billion in improper payments in fiscal year 2025 (up from $162 billion in FY2024), and cumulative improper payments since 2003 total roughly $2.8–3 trillion. Separately, GAO estimates direct annual financial losses to the government from fraud at between $233 billion and $521 billion. These are figures from the Government Accountability Office, an independent, nonpartisan body, not from us.
Sources: GAO, Payment Integrity: Agencies' Estimated Improper Payments Increased to $186 Billion in Fiscal Year 2025 (Apr. 2026), https://www.gao.gov/products/gao-26-108694 ; GAO, FY2024 improper payments ($162B), https://www.gao.gov/press-release/gao-reports-estimated-162-billion-improper-payments-across-federal-government-fiscal-year-2024 ; GAO fraud estimate ($233B–$521B), https://www.gao.gov/products/gao-25-108172
Note on the Home page. Our front page says "tens of billions of dollars in waste, improper payments, and untracked expenditures." This is understated, not wrong; the actual figure is in the hundreds of billions per year. We could strengthen it to GAO's current numbers. But one honesty caveat is binding, and it governs our funding model: money identified as improper is not the same as money recovered. Recovered amounts are a small fraction of identified amounts. Our governance and funding model depends on recovered waste, so any use of these figures must say plainly that identification is the ceiling, not the deposit. Citing $186 billion implies nothing about what a recovery program would actually return.
Part B — International and comparative evidence
We are often asked a fair question: if continuous, verified citizen participation is such a good idea, why has no one done it? The honest answer is that pieces of it exist around the world, each with instructive successes and instructive failures. We are not the first to try. We are trying to learn from what worked, what broke, and why.
Each entry below states what the case demonstrates, where it falls short, and how it maps to our own framework. The failures matter as much as the successes; a precedent base that only listed wins would not be trustworthy.
The expert consensus on internet voting — the frame for everything below
What it establishes. Before any individual case, this is the authoritative baseline. In 2018, the National Academies of Sciences, Engineering, and Medicine published Securing the Vote: Protecting American Democracy, the most authoritative single statement of expert consensus on U.S. election security. Its core findings: the technology to support secure internet voting does not exist at present, though it may in the future; all elections should use human-readable paper ballots; and marked ballots should not be returned over the internet, because no current technology can guarantee their secrecy, security, and verifiability at once.
Why it is the frame, not just an entry. This is the consensus our entire voting posture is built to respect rather than dodge. When we say "the implementation is earned, not assumed," we are restating the National Academies' own "not now, possibly later, only with robust guarantees." The report also validates two specifics: our paper-trail requirement (its central recommendation) and a nuance worth keeping: the internet is an acceptable way to deliver blank ballots to voters even where returning marked ballots is not safe. Every case below is read against this baseline.
Source: National Academies of Sciences, Engineering, and Medicine, Securing the Vote: Protecting American Democracy (2018), https://nap.nationalacademies.org/catalog/25120/securing-the-vote-protecting-american-democracy
The plain-language version of this standard, all six safeguards and who governs them, lives on The Honest Case on Voting Security.
Estonia — internet voting at national scale
What it demonstrates. Estonia is the only country that uses internet voting at meaningful scale for legally binding national elections. It ran the first binding online municipal vote in 2005 and the first online parliamentary vote in 2007. In the 2023 parliamentary election, for the first time, more than half of all votes were cast online. The system runs on a national digital-identity infrastructure used daily for banking and government, not a special-purpose election tool.
Where it falls short, and why this matters to us. Independent security researchers led by J. Alex Halderman (University of Michigan) reported in 2014 that the system had serious procedural and operational security gaps and an architecture vulnerable to state-level attackers. Estonia responded to sustained criticism by adding universal tally verifiability in 2017. Even so, the OSCE/ODIHR final report on the 2023 elections still recommends introducing end-to-end verifiability, its own authoritative confirmation that the system is not there yet. Estonia proves online voting can run at scale; it does not prove the security question is closed.
How it maps to our framework. This is the precedent that justifies our exact posture: pilot-based, paper-backed, "the implementation is earned, not assumed." Estonia is what earning it looks like in practice, including the unresolved parts. We cite it for feasibility, never as proof the problem is solved.
Sources: Estonian State Electoral Office, official internet-voting statistics, https://www.valimised.ee/en/archive/statistics-about-internet-voting-estonia ; OSCE/ODIHR, Estonia, Parliamentary Elections, 5 March 2023: Final Report (incl. the end-to-end-verifiability recommendation), https://www.osce.org/odihr/elections/estonia/551179 ; peer-reviewed security analysis (CCS'14), https://jhalderm.com/pub/papers/ivoting-ccs14.pdf
Switzerland — direct democracy, and "security before speed"
What it demonstrates. Switzerland is the strongest single precedent for our posture, on both halves of what we propose. It runs frequent binding national referenda, the clearest example of routine direct citizen decision-making in a stable democracy. And since 2004, fifteen cantons have run roughly 300 internet-voting trials under an explicit federal principle: security before speed. Internet voting there begins with exactly the populations the existing system serves worst: citizens abroad whom the postal system fails, and voters with disabilities.
Where it falls short, and why this is the model, not a caution. In 2019, both main e-voting providers withdrew their systems and trials stopped, after a public intrusion test and source-code publication revealed critical flaws. Switzerland did not push through; it paused, rebuilt the legal framework so that only fully verifiable systems are permitted, published the source code, ran public intrusion tests and a bug-bounty program, and only then restarted cautiously in 2023 in a handful of cantons. Participation is capped by law: no more than 30 percent of a canton's electorate and no more than 10 percent of the national electorate. After twenty years, it remains trial-status and capped. That restraint is the point, not a weakness.
How it maps to our framework. This is our entire method, already operating in a real democracy: pilot-based, start-with-the-underserved, open-source, independently audited, expanded only on evidence, and willing to stop. The willingness to pause when transparency exposes a flaw is precisely what separates "earned" from "assumed." When critics say our approach is naive, Switzerland is the answer: this is what disciplined, slow, honest digital-voting development actually looks like.
Sources: Swiss Federal Chancellery, E-voting (official overview, incl. caps and "security before speed"), https://www.bk.admin.ch/bk/en/home/politische-rechte/e-voting.html ; Federal Council, resumption of trials (Mar. 2023), https://www.admin.ch/gov/en/start/documentation/media-releases/media-releases-federal-council.msg-id-93455.html ; Library of Congress, New Legislation on E-Voting Trials (Aug. 2022, incl. 2019 suspension and the 30%/10% caps), https://www.loc.gov/item/global-legal-monitor/2022-08-04/switzerland-new-legislation-on-e-voting-trials-enters-into-force/
Voatz / West Virginia — the cautionary tale
What it demonstrates. This is the failure we cite on purpose. Voatz was the first internet-voting app used in U.S. federal elections, piloted in West Virginia in 2018 for overseas and military voters. In 2020, MIT researchers (Specter, Koppel, Weitzner) published a security analysis finding that a network adversary could detect how someone voted and an attacker controlling the device could alter the vote. They also found that Voatz's reliance on a third-party vendor for identity verification created privacy exposure. West Virginia and a Washington county dropped Voatz from their 2020 primaries as a direct result.
Why we cite it. It validates three of our own standing disciplines at once. The MIT team's plain statement of expert consensus, that running a secure election over the internet is "not possible today," is exactly why we refuse the word "unhackable" and frame digital voting as research to be earned. Their finding on the third-party identity vendor is exactly why we say production identity infrastructure must be nonprofit, open-source, and publicly governed, not contracted to a for-profit vendor. And the whole episode is why we insist on a paper record behind every digital vote.
Sources: MIT News (Feb. 2020), https://news.mit.edu/2020/voting-voatz-app-hack-issues-0213 ; peer-reviewed paper, USENIX Security 2020, The Ballot is Busted Before the Blockchain, https://www.usenix.org/conference/usenixsecurity20/presentation/specter
India (Aadhaar) — identity at billion-person scale, and every failure mode our rules exist to prevent
What it demonstrates. Aadhaar is the world's largest biometric identity system: roughly 1.2 billion people, more than 90 percent of India's population, each issued a number tied to fingerprints, iris scans, and a photo. It proves that identity verification can be built at population scale, and it has been credited with streamlining welfare delivery and giving many rural residents a verifiable identity for the first time.
Where it falls short, and why it validates our disciplines by negative example. Aadhaar is, in effect, a live demonstration of every vulnerability our standing rules are written to prevent. Single-mode biometric authentication produced real exclusion: elderly people whose irises would not scan, manual laborers with worn fingerprints, and others were denied benefits when authentication failed, with documented harm to the most vulnerable. Government sites exposed millions of records, and access was reportedly sold for a few dollars. The administering authority claimed its data could not be breached, the exact overconfidence we forbid ourselves. In 2018, India's Supreme Court upheld the system but limited it: private companies cannot require it, and it cannot be mandatory for banking, phone service, or school admission.
How it maps to our framework. Aadhaar is why our biometric-equity rule (multi-modal, never single-mode, demographic-equity testing before deployment) is non-negotiable; why we say identity infrastructure must be publicly governed, not handed to private vendors; why we refuse "unhackable"; and why judicial review of rights-implicating systems matters. It shows that scale without these safeguards does not empower the marginalized. It can exclude them further.
Sources: TIME, India's Supreme Court Upholds Biometric ID System (Sep. 2018, incl. enrollment-exclusion examples), https://time.com/5388257/india-aadhaar-biometric-identification/ ; Human Rights Watch, India: Top Court OK's Biometric ID Program (Sep. 2018, incl. breach and exclusion harms), https://www.hrw.org/news/2018/09/27/india-top-court-oks-biometric-id-program
Brazil — electronic voting at massive scale, and why a paper trail still matters
What it demonstrates. Brazil has run fully electronic national elections since the first nationwide electronic-only vote in 2000, with results from more than 150 million voters reported within hours of polls closing. It is one of the most heavily audited election systems in the world: in 2022, the electoral court opened the system to multiple oversight bodies, including the armed forces and federal police, and added biometric integrity tests. International observers, including the Carter Center, found the 2022 elections well-administered and transparent, and no credible evidence of significant fraud was presented in the audited elections.
Where it falls short, stated carefully, because this one carries a political trap. Brazil's machines use a direct-recording electronic design that, as the Carter Center observed, places the integrity of the result on the voting software rather than on a voter-verified paper record of individual votes. Independent security experts argue, on standard security grounds, that a physical record an ordinary auditor can check is fundamental to making any electronic count fully verifiable. That critique is correct and it matches our own paper-trail rule. It must be kept strictly separate from the false fraud narrative that political actors built around Brazil's elections: no credible evidence of fraud was found, and we do not endorse or echo that narrative. Our support for a voter-verified paper trail rests on the same security consensus as the National Academies and MIT, not on any claim that Brazil's elections were rigged.
How it maps to our framework. Brazil proves electronic vote capture works at national scale, fast, and observed as transparent, and simultaneously demonstrates why "paper backs every vote" is a rule we will not drop. It is feasibility and missing-safeguard in a single case.
Sources: The Carter Center, electoral expert mission final findings on Brazil's 2022 elections (incl. the direct-recording design observation and audit procedures), https://www.cartercenter.org/news/pr/2022/brazil-110422.html ; OAS Electoral Observation Mission, Brazil 2022 Preliminary Report, https://www.oas.org/fpdb/press/EOM-Brazil-Preliminary-Report.pdf
NIST — the evidence behind our biometric-equity rule
What it demonstrates. Our requirement of multi-modal authentication and mandatory demographic-equity testing is not a slogan; it rests on the largest study of its kind. In 2019, the National Institute of Standards and Technology tested 189 face-recognition algorithms from 99 developers across 18.27 million images. It found empirical demographic differentials in most algorithms. For one-to-many matching, false-positive rates were highest for African American women, with differentials that often ranged from 10 to 100 times depending on the algorithm.
The nuance that makes our rule the right one. NIST also found that the most equitable algorithms were among the most accurate, and that different algorithms perform very differently. In other words, the bias is real but not uniform or inevitable, which is precisely why the rule has to be test before deploy, never trust by default, and never facial-recognition-only. The evidence supports a binding equity standard, not a ban and not blind adoption.
Sources: NIST, NIST Study Evaluates Effects of Race, Age, Sex on Face Recognition Software (Dec. 2019), https://www.nist.gov/news-events/news/2019/12/nist-study-evaluates-effects-race-age-sex-face-recognition-software ; full report, NISTIR 8280 (FRVT Part 3: Demographic Effects), https://nvlpubs.nist.gov/nistpubs/ir/2019/NIST.IR.8280.pdf
Taiwan — continuous participation, and the gap our framework is built to close
What it demonstrates. Taiwan runs two participation systems worth studying. vTaiwan is a deliberation platform that used consensus-mapping software (Pol.is) to surface where citizens actually agree rather than amplify the loudest voices; at its peak it reached around 200,000 users and informed regulation on contested issues such as ride-sharing. Join, the government-run petition platform, requires a formal government response to any proposal that passes a signature threshold. These are working, at-scale examples of structured public sentiment feeding into policy.
Where it falls short, and why this is the most important entry on the page for us. vTaiwan has not driven a major decision since 2018 and now runs as a volunteer project. One of its co-creators, legislator Jason Hsu, called it "a tiger without teeth," because the government was never required to act on the consensus it produced. Scholars describe even Taiwan's admired platforms as delivering a "thin" form of participation: citizens were consulted, but binding power never left officials' hands. Participation also skewed younger and more digitally literate, the same digital-divide problem our "who this is for" framing names.
How it maps to our framework. Taiwan is feasibility proof and a diagnosis of the exact weakness our Right to Acknowledged Sentiment (Amendment III) and Binding Acknowledgment are designed to fix. Join's "cross the threshold, get a mandatory response" is essentially our acknowledgment mechanism in the wild; vTaiwan's "tiger without teeth" problem is what happens when that response is optional. We are not copying Taiwan. We are trying to add the teeth, a required, recorded response, while keeping acknowledgment short of enactment, so the mechanism strengthens representation without overriding it.
Sources: European Democracy Hub, Taiwan case study (incl. Join thresholds and figures), https://europeandemocracyhub.epd.eu/exploring-worldwide-democratic-innovations-taiwan/ ; RadicalxChange, Taiwan: Grassroots Digital Democracy That Works, https://www.radicalxchange.org/updates/papers/Taiwan_Grassroots_Digital_Democracy_That_Works_V1_DIGITAL_.pdf
Germany — structured worker voice (evidence for the Employment page, not the voting model)
What it demonstrates. Germany's codetermination model (Mitbestimmung) is long-running evidence that giving people a formal, structured voice inside the institutions that govern their working lives is workable, not utopian. Under the 1976 Codetermination Act, companies with more than 2,000 employees must reserve half their supervisory-board seats for employee representatives; firms with 500 to 2,000 employees reserve one-third. Separately, works councils (Betriebsrat) in establishments with five or more employees hold consultation and co-decision rights over working conditions, hours, and dismissals.
Where it falls short. The evidence on outcomes is mixed, not glowing: most recent studies find neutral to small positive effects on productivity, innovation, and investment, and works councils appear to function better as they mature. It is structured participation that works steadily, not a cure-all, and that honest, modest read is the right one.
How it maps to our framework, and where it belongs. This is routed deliberately. It is evidence for the Employment page's claim that structured voice and dignity in contribution are compatible with a functioning market economy, not evidence for the voting model. We log it here so the claim is sourced, but it argues for the Employment page, not the Framework.
Sources: German Federal Government, Rules on staff representation (official, incl. board-representation thresholds), https://verwaltung.bund.de/leistungsverzeichnis/en/rechte-und-pflichten/102837998 ; Harvard Law School Forum on Corporate Governance, Worker Participation: Employee Ownership and Representation (incl. mixed-evidence summary), https://corpgov.law.harvard.edu/2020/01/23/worker-participation-employee-ownership-and-representation/
We are often asked a fair question: if continuous, verified citizen participation is such a good idea, why has no one done it? The honest answer is that pieces of it exist around the world, each with instructive successes and instructive failures. We are not the first to try. We are trying to learn from what worked, what broke, and why.
Each entry below states what the case demonstrates, where it falls short, and how it maps to our own framework. The failures matter as much as the successes; a precedent base that only listed wins would not be trustworthy.
The expert consensus on internet voting — the frame for everything below
What it establishes. Before any individual case, this is the authoritative baseline. In 2018, the National Academies of Sciences, Engineering, and Medicine published Securing the Vote: Protecting American Democracy, the most authoritative single statement of expert consensus on U.S. election security. Its core findings: the technology to support secure internet voting does not exist at present, though it may in the future; all elections should use human-readable paper ballots; and marked ballots should not be returned over the internet, because no current technology can guarantee their secrecy, security, and verifiability at once.
Why it is the frame, not just an entry. This is the consensus our entire voting posture is built to respect rather than dodge. When we say "the implementation is earned, not assumed," we are restating the National Academies' own "not now, possibly later, only with robust guarantees." The report also validates two specifics: our paper-trail requirement (its central recommendation) and a nuance worth keeping: the internet is an acceptable way to deliver blank ballots to voters even where returning marked ballots is not safe. Every case below is read against this baseline.
Source: National Academies of Sciences, Engineering, and Medicine, Securing the Vote: Protecting American Democracy (2018), https://nap.nationalacademies.org/catalog/25120/securing-the-vote-protecting-american-democracy
The plain-language version of this standard, all six safeguards and who governs them, lives on The Honest Case on Voting Security.
Estonia — internet voting at national scale
What it demonstrates. Estonia is the only country that uses internet voting at meaningful scale for legally binding national elections. It ran the first binding online municipal vote in 2005 and the first online parliamentary vote in 2007. In the 2023 parliamentary election, for the first time, more than half of all votes were cast online. The system runs on a national digital-identity infrastructure used daily for banking and government, not a special-purpose election tool.
Where it falls short, and why this matters to us. Independent security researchers led by J. Alex Halderman (University of Michigan) reported in 2014 that the system had serious procedural and operational security gaps and an architecture vulnerable to state-level attackers. Estonia responded to sustained criticism by adding universal tally verifiability in 2017. Even so, the OSCE/ODIHR final report on the 2023 elections still recommends introducing end-to-end verifiability, its own authoritative confirmation that the system is not there yet. Estonia proves online voting can run at scale; it does not prove the security question is closed.
How it maps to our framework. This is the precedent that justifies our exact posture: pilot-based, paper-backed, "the implementation is earned, not assumed." Estonia is what earning it looks like in practice, including the unresolved parts. We cite it for feasibility, never as proof the problem is solved.
Sources: Estonian State Electoral Office, official internet-voting statistics, https://www.valimised.ee/en/archive/statistics-about-internet-voting-estonia ; OSCE/ODIHR, Estonia, Parliamentary Elections, 5 March 2023: Final Report (incl. the end-to-end-verifiability recommendation), https://www.osce.org/odihr/elections/estonia/551179 ; peer-reviewed security analysis (CCS'14), https://jhalderm.com/pub/papers/ivoting-ccs14.pdf
Switzerland — direct democracy, and "security before speed"
What it demonstrates. Switzerland is the strongest single precedent for our posture, on both halves of what we propose. It runs frequent binding national referenda, the clearest example of routine direct citizen decision-making in a stable democracy. And since 2004, fifteen cantons have run roughly 300 internet-voting trials under an explicit federal principle: security before speed. Internet voting there begins with exactly the populations the existing system serves worst: citizens abroad whom the postal system fails, and voters with disabilities.
Where it falls short, and why this is the model, not a caution. In 2019, both main e-voting providers withdrew their systems and trials stopped, after a public intrusion test and source-code publication revealed critical flaws. Switzerland did not push through; it paused, rebuilt the legal framework so that only fully verifiable systems are permitted, published the source code, ran public intrusion tests and a bug-bounty program, and only then restarted cautiously in 2023 in a handful of cantons. Participation is capped by law: no more than 30 percent of a canton's electorate and no more than 10 percent of the national electorate. After twenty years, it remains trial-status and capped. That restraint is the point, not a weakness.
How it maps to our framework. This is our entire method, already operating in a real democracy: pilot-based, start-with-the-underserved, open-source, independently audited, expanded only on evidence, and willing to stop. The willingness to pause when transparency exposes a flaw is precisely what separates "earned" from "assumed." When critics say our approach is naive, Switzerland is the answer: this is what disciplined, slow, honest digital-voting development actually looks like.
Sources: Swiss Federal Chancellery, E-voting (official overview, incl. caps and "security before speed"), https://www.bk.admin.ch/bk/en/home/politische-rechte/e-voting.html ; Federal Council, resumption of trials (Mar. 2023), https://www.admin.ch/gov/en/start/documentation/media-releases/media-releases-federal-council.msg-id-93455.html ; Library of Congress, New Legislation on E-Voting Trials (Aug. 2022, incl. 2019 suspension and the 30%/10% caps), https://www.loc.gov/item/global-legal-monitor/2022-08-04/switzerland-new-legislation-on-e-voting-trials-enters-into-force/
Voatz / West Virginia — the cautionary tale
What it demonstrates. This is the failure we cite on purpose. Voatz was the first internet-voting app used in U.S. federal elections, piloted in West Virginia in 2018 for overseas and military voters. In 2020, MIT researchers (Specter, Koppel, Weitzner) published a security analysis finding that a network adversary could detect how someone voted and an attacker controlling the device could alter the vote. They also found that Voatz's reliance on a third-party vendor for identity verification created privacy exposure. West Virginia and a Washington county dropped Voatz from their 2020 primaries as a direct result.
Why we cite it. It validates three of our own standing disciplines at once. The MIT team's plain statement of expert consensus, that running a secure election over the internet is "not possible today," is exactly why we refuse the word "unhackable" and frame digital voting as research to be earned. Their finding on the third-party identity vendor is exactly why we say production identity infrastructure must be nonprofit, open-source, and publicly governed, not contracted to a for-profit vendor. And the whole episode is why we insist on a paper record behind every digital vote.
Sources: MIT News (Feb. 2020), https://news.mit.edu/2020/voting-voatz-app-hack-issues-0213 ; peer-reviewed paper, USENIX Security 2020, The Ballot is Busted Before the Blockchain, https://www.usenix.org/conference/usenixsecurity20/presentation/specter
India (Aadhaar) — identity at billion-person scale, and every failure mode our rules exist to prevent
What it demonstrates. Aadhaar is the world's largest biometric identity system: roughly 1.2 billion people, more than 90 percent of India's population, each issued a number tied to fingerprints, iris scans, and a photo. It proves that identity verification can be built at population scale, and it has been credited with streamlining welfare delivery and giving many rural residents a verifiable identity for the first time.
Where it falls short, and why it validates our disciplines by negative example. Aadhaar is, in effect, a live demonstration of every vulnerability our standing rules are written to prevent. Single-mode biometric authentication produced real exclusion: elderly people whose irises would not scan, manual laborers with worn fingerprints, and others were denied benefits when authentication failed, with documented harm to the most vulnerable. Government sites exposed millions of records, and access was reportedly sold for a few dollars. The administering authority claimed its data could not be breached, the exact overconfidence we forbid ourselves. In 2018, India's Supreme Court upheld the system but limited it: private companies cannot require it, and it cannot be mandatory for banking, phone service, or school admission.
How it maps to our framework. Aadhaar is why our biometric-equity rule (multi-modal, never single-mode, demographic-equity testing before deployment) is non-negotiable; why we say identity infrastructure must be publicly governed, not handed to private vendors; why we refuse "unhackable"; and why judicial review of rights-implicating systems matters. It shows that scale without these safeguards does not empower the marginalized. It can exclude them further.
Sources: TIME, India's Supreme Court Upholds Biometric ID System (Sep. 2018, incl. enrollment-exclusion examples), https://time.com/5388257/india-aadhaar-biometric-identification/ ; Human Rights Watch, India: Top Court OK's Biometric ID Program (Sep. 2018, incl. breach and exclusion harms), https://www.hrw.org/news/2018/09/27/india-top-court-oks-biometric-id-program
Brazil — electronic voting at massive scale, and why a paper trail still matters
What it demonstrates. Brazil has run fully electronic national elections since the first nationwide electronic-only vote in 2000, with results from more than 150 million voters reported within hours of polls closing. It is one of the most heavily audited election systems in the world: in 2022, the electoral court opened the system to multiple oversight bodies, including the armed forces and federal police, and added biometric integrity tests. International observers, including the Carter Center, found the 2022 elections well-administered and transparent, and no credible evidence of significant fraud was presented in the audited elections.
Where it falls short, stated carefully, because this one carries a political trap. Brazil's machines use a direct-recording electronic design that, as the Carter Center observed, places the integrity of the result on the voting software rather than on a voter-verified paper record of individual votes. Independent security experts argue, on standard security grounds, that a physical record an ordinary auditor can check is fundamental to making any electronic count fully verifiable. That critique is correct and it matches our own paper-trail rule. It must be kept strictly separate from the false fraud narrative that political actors built around Brazil's elections: no credible evidence of fraud was found, and we do not endorse or echo that narrative. Our support for a voter-verified paper trail rests on the same security consensus as the National Academies and MIT, not on any claim that Brazil's elections were rigged.
How it maps to our framework. Brazil proves electronic vote capture works at national scale, fast, and observed as transparent, and simultaneously demonstrates why "paper backs every vote" is a rule we will not drop. It is feasibility and missing-safeguard in a single case.
Sources: The Carter Center, electoral expert mission final findings on Brazil's 2022 elections (incl. the direct-recording design observation and audit procedures), https://www.cartercenter.org/news/pr/2022/brazil-110422.html ; OAS Electoral Observation Mission, Brazil 2022 Preliminary Report, https://www.oas.org/fpdb/press/EOM-Brazil-Preliminary-Report.pdf
NIST — the evidence behind our biometric-equity rule
What it demonstrates. Our requirement of multi-modal authentication and mandatory demographic-equity testing is not a slogan; it rests on the largest study of its kind. In 2019, the National Institute of Standards and Technology tested 189 face-recognition algorithms from 99 developers across 18.27 million images. It found empirical demographic differentials in most algorithms. For one-to-many matching, false-positive rates were highest for African American women, with differentials that often ranged from 10 to 100 times depending on the algorithm.
The nuance that makes our rule the right one. NIST also found that the most equitable algorithms were among the most accurate, and that different algorithms perform very differently. In other words, the bias is real but not uniform or inevitable, which is precisely why the rule has to be test before deploy, never trust by default, and never facial-recognition-only. The evidence supports a binding equity standard, not a ban and not blind adoption.
Sources: NIST, NIST Study Evaluates Effects of Race, Age, Sex on Face Recognition Software (Dec. 2019), https://www.nist.gov/news-events/news/2019/12/nist-study-evaluates-effects-race-age-sex-face-recognition-software ; full report, NISTIR 8280 (FRVT Part 3: Demographic Effects), https://nvlpubs.nist.gov/nistpubs/ir/2019/NIST.IR.8280.pdf
Taiwan — continuous participation, and the gap our framework is built to close
What it demonstrates. Taiwan runs two participation systems worth studying. vTaiwan is a deliberation platform that used consensus-mapping software (Pol.is) to surface where citizens actually agree rather than amplify the loudest voices; at its peak it reached around 200,000 users and informed regulation on contested issues such as ride-sharing. Join, the government-run petition platform, requires a formal government response to any proposal that passes a signature threshold. These are working, at-scale examples of structured public sentiment feeding into policy.
Where it falls short, and why this is the most important entry on the page for us. vTaiwan has not driven a major decision since 2018 and now runs as a volunteer project. One of its co-creators, legislator Jason Hsu, called it "a tiger without teeth," because the government was never required to act on the consensus it produced. Scholars describe even Taiwan's admired platforms as delivering a "thin" form of participation: citizens were consulted, but binding power never left officials' hands. Participation also skewed younger and more digitally literate, the same digital-divide problem our "who this is for" framing names.
How it maps to our framework. Taiwan is feasibility proof and a diagnosis of the exact weakness our Right to Acknowledged Sentiment (Amendment III) and Binding Acknowledgment are designed to fix. Join's "cross the threshold, get a mandatory response" is essentially our acknowledgment mechanism in the wild; vTaiwan's "tiger without teeth" problem is what happens when that response is optional. We are not copying Taiwan. We are trying to add the teeth, a required, recorded response, while keeping acknowledgment short of enactment, so the mechanism strengthens representation without overriding it.
Sources: European Democracy Hub, Taiwan case study (incl. Join thresholds and figures), https://europeandemocracyhub.epd.eu/exploring-worldwide-democratic-innovations-taiwan/ ; RadicalxChange, Taiwan: Grassroots Digital Democracy That Works, https://www.radicalxchange.org/updates/papers/Taiwan_Grassroots_Digital_Democracy_That_Works_V1_DIGITAL_.pdf
Germany — structured worker voice (evidence for the Employment page, not the voting model)
What it demonstrates. Germany's codetermination model (Mitbestimmung) is long-running evidence that giving people a formal, structured voice inside the institutions that govern their working lives is workable, not utopian. Under the 1976 Codetermination Act, companies with more than 2,000 employees must reserve half their supervisory-board seats for employee representatives; firms with 500 to 2,000 employees reserve one-third. Separately, works councils (Betriebsrat) in establishments with five or more employees hold consultation and co-decision rights over working conditions, hours, and dismissals.
Where it falls short. The evidence on outcomes is mixed, not glowing: most recent studies find neutral to small positive effects on productivity, innovation, and investment, and works councils appear to function better as they mature. It is structured participation that works steadily, not a cure-all, and that honest, modest read is the right one.
How it maps to our framework, and where it belongs. This is routed deliberately. It is evidence for the Employment page's claim that structured voice and dignity in contribution are compatible with a functioning market economy, not evidence for the voting model. We log it here so the claim is sourced, but it argues for the Employment page, not the Framework.
Sources: German Federal Government, Rules on staff representation (official, incl. board-representation thresholds), https://verwaltung.bund.de/leistungsverzeichnis/en/rechte-und-pflichten/102837998 ; Harvard Law School Forum on Corporate Governance, Worker Participation: Employee Ownership and Representation (incl. mixed-evidence summary), https://corpgov.law.harvard.edu/2020/01/23/worker-participation-employee-ownership-and-representation/
Part C — What this base does and does not yet cover
The core precedents are now sourced. Honesty about an evidence base still includes naming what remains, so the gaps stay visible rather than papered over.
The core precedents are now sourced. Honesty about an evidence base still includes naming what remains, so the gaps stay visible rather than papered over.
- The per-page statistics sweep is ongoing. Each factual claim across all fourteen Catalog pages, the Framework, and the governance page still needs to be checked against a dated primary source and logged here.
- The recovered-versus-identified-waste figure needs its own sourced entry. Our funding model depends on recovered waste, and the gap between what GAO identifies and what any program would actually recover is large and under-documented. Before the funding model cites a dollar figure publicly, that recovery rate needs its own honest, sourced treatment. Flagged, not yet built.
- Switzerland's referendum frequency is asserted above from official overviews; if we make a specific numerical claim about how often Swiss citizens vote, it needs its own primary citation from the Federal Chancellery before going on a public page.
How this page is maintained
- Every statistic across the site flows here. Each factual claim on every Catalog page, the Framework, and the Home page should have a corresponding, dated entry in this base. This page is the single index; pages link to it rather than restating raw figures, so a number lives in one authoritative place.
- Primary sources first. Government data (Census, GAO, BLS, NIST), peer-reviewed research, and official election authorities outrank aggregators, advocacy organizations, and commentary. We do not cite advocacy-coded sources on either side, because a source's politics travel with it.
- Every figure is dated, and stale figures are corrected, not kept. When a primary source updates, the old number is replaced and the change is visible.
- Contested numbers are labeled contested. Where credible sources disagree (as with the nonvoter count), we give the range and the reason for the spread rather than picking the most flattering figure.
Not left. Not right. Altruist.
Long live everyone’s freedom of voice.
Long live everyone’s freedom of voice.