A New Incentive Model For Content Trust

Eric Tze Kuan Lim; Lucas Barbosa; Rob Kopel; Sam Kirshner; Tom Pagram

REVIEW 5 major objections 5 minor 53 references

A New Incentive Model For Content Trust

T0 review · 5 major / 5 minor · reviewed 2026-08-06 · deepseek-v4-flash

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗

Pith's one-line read The paper proposes that staking bonds on claims and letting paid juries adjudicate makes content verification self-funding, with rewards for accurate assessments funded entirely by penalties for inaccurate content.

desk verdict A candid design proposal for bond-based content verification; the truth-alignment claim is an unproven assumption about jury accuracy, and the capacity theorem has a proof gap. read the letter →

arxiv 2507.09972 v1 pith:YMXZFNY2 submitted 2025-07-14 cs.GT cs.CYecon.GNq-fin.EC

Lucas Barbosa , Sam Kirshner , Rob Kopel , Eric Tze Kuan Lim , Tom Pagram This is my paper · ORCID

classification cs.GTcs.CYecon.GNq-fin.EC MSC 91A8091B4460C05

keywords misinformationveracitybondcounter-veracitycrowdsourcedfact-checkingjuryadjudicationsmartcontractsdigitalidentityincentivedesign

verification ladder T0 review T1 audit T2 compute T3 formal

The pith

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The reading

The paper proposes a protocol in which creators stake money on the truth of their claims, challengers stake an equal amount to dispute them, and a randomly selected jury decides who wins; the loser's stake pays the winner and the jury. Its central assertion is that this closes the loop: rewards for accurate assessment come solely from penalties for inaccurate content, so a platform running the protocol needs no external subsidy. The authors further claim that the chance of a small colluding bloc overturning a verdict falls exponentially as the jury grows, and that volunteer juror pools of practical size can handle the dispute rates of large platforms. If the design works, it gives content creators and fact-checkers a financial stake in accuracy and an economic reason to prefer verified content.

What carries the argument

The central mechanism is the veracity-bond contest: a creator posts collateral $\beta$ on a claim, a challenger posts an equal counter-bond, an odd-sized jury of verified users decides the dispute, and the losing side's $\beta$ is distributed according to the identity $\pi_c + \sum_j \pi_j + \pi_p = \beta$, with a fixed share $\pi_p$ reserved for the framework. A second piece of machinery is an exponential tail bound for hypergeometric sampling showing that the probability a fixed bloc of $k$ colluding jurors flips a verdict is at most $\exp[-2n(\tfrac12 - p)^2]$ with $p = k/N$, which decays exponentially as the jury size $n$ grows. Together, these are meant to make accuracy self-funding and jury manipulation statistically untenable.

What would settle it

Pilot the protocol on claims with known ground truth: if jury verdicts are not significantly more accurate than a coin flip, or if false claims attract too few challengers to activate the payout loop, then the central self-sustainability claim fails.

Watch

Extended reading notes

Core claim

The discovery the paper argues for is the payout identity $\pi_c + \sum_{j\in J} \pi_j + \pi_p = \beta$, which governs every contest: the forfeited bond of the losing creator or challenger is split wholly among the winning party, the jurors, and the framework itself. Because every payout is funded by a forfeiture, the system is self-sustaining rather than subsidy-dependent, and each role faces real risk: creators with false claims lose bonds, challengers who dispute true claims lose counter-bonds, and jurors who fail to vote or who receive poor evaluations forfeit their bonds. The paper argues that this structure aligns incentives so that truthful content is rewarded, false content is penalized, and participation in fact-checking becomes economically rational.

Load-bearing premise

The framework rests on the premise that a randomly chosen jury of verified users will reach verdicts that track the truth more often than chance across the full range of contested claims, a property the paper assumes rather than proves.

Editorial extensions

If this is right

A creator who publishes a true claim and faces no successful challenge keeps the bond, while a challenger who cannot prove a claim false loses the counter-bond, so both sides bear financial risk tied to accuracy.
Equal bonds for challengers and creators discourage frivolous or malicious disputes and prevent the jury's monetary incentives from being skewed toward one side.
A jury of about 21 verified, rated jurors keeps the chance of a coordinated 10% colluding bloc overturning a verdict below 0.2%, and larger panels drive it far below that.
On large platforms, required juror pools stay below a tenth of a percent of daily active users under conservative dispute-rate assumptions, so jury capacity is a tunable constraint rather than a fundamental bottleneck.
Linking content visibility to bond size rewards higher-stake claims with more scrutiny rather than simply more reach, because the bond is forfeitable if the claim is proven false.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

If the protocol were deployed, the most informative early test would be whether jury verdicts on claims whose ground truth is known outperform chance; that is the premise the paper does not prove.
The model converts misinformation from an engagement externality into a pricing problem, so one could imagine markets where the bond size itself reveals a creator's private confidence in a claim.
The same closed-loop payout could be grafted onto academic peer review or insurance claim assessment, with authors or applicants staking bonds and reviewers rewarded; the paper lists these as open questions rather than developed extensions.
A stable equilibrium requires that uninformed but honest jurors do not systematically outvote informed ones; absent that, the collusion bound alone does not guarantee that verdicts track truth.

Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Desk Editor's Note

read the letter

Quick take: this is a clearly written, honest design paper for a decentralized content-verification system (veracity bonds + challengers + paid juries), and it extends the authors' earlier idea with some genuinely new pieces: serial challenges, refundable juror bonds, randomized anonymous juror evaluation, a collusion bound, and a capacity rule. But the central claim that the model aligns incentives with accuracy is not supported. Eq. 1 is an accounting identity: the losing bond is redistributed to the winner and jurors. Whether that reward tracks truth depends entirely on whether a paid, randomly selected jury majority is more likely to be correct than incorrect. The paper never shows this. Theorem A.1 only bounds the probability that a fixed block of k colluders, with k/N < 1/2, sways the verdict, and it assumes the other jurors vote independently and (implicitly) correctly. It says nothing about systematic honest error, correlated false beliefs, dominant wrong majorities, or collusion between jurors and parties—things the authors themselves list as open risks in Section 6.1. So the abstract's 'self-propelling paradigm shift' is a hope, not a result.

Where credit is due: the protocol is well-specified and the authors are unusually candid about limitations. Theorem A.1 is a correct application of Hoeffding's inequality to a hypergeometric distribution, and the numerical Table 1 matches the bound. The capacity formula in Eq. 3 is a sensible rule of thumb. But the proof of Theorem B.1 has a real gap: the M/G/c model takes servers as panels of n jurors with service rate 1/h, which gives stability only if N > λnh; the stated Nmin = ⌈λnh/a⌉ is weaker when a > 1. The per-juror time cap a never appears in the queueing model. The formula may hold under a different model, but the theorem as written doesn't follow.

The other soft spot is evidential: the only empirical support cited is the authors' own earlier paper [29], which measured perceived credibility, not accuracy of verdicts. A simple simulation or a plan to test jury accuracy would strengthen this significantly.

Who should read it: people working on mechanism design for misinformation, prediction-market-like verification, or platform governance. It's a serious working paper, not a finished result. I'd send it to peer review, with the expectation that the authors fix the capacity proof and either soften or conditionalize the incentive claim. A desk reject would be too hasty.

Referee Report

5 major / 5 minor

Summary. The paper proposes a smart-contract-based protocol for content verification: creators stake a veracity bond β on their claims, challengers stake matching counter-veracity bonds, a randomly selected jury adjudicates, and the losing party's bond is split among the winning party, jurors, and the protocol according to Eq. (1). It also sketches reputation scoring, bond-based visibility, digital identity, and content provenance, and it includes two formal-looking appendix results: a hypergeometric tail bound on juror collusion (Theorem A.1) and a juror-capacity threshold (Theorem B.1). The authors explicitly frame the paper as exploratory and list open questions in Section 6.

Significance. If the central incentive-alignment claim were established, the protocol would be a useful design contribution: a self-funding mechanism that rewards truthful content and penalizes false claims without centralized fact-checking. The paper's strengths are its transparent payout identity, clear role definitions, and the correct application of Hoeffding's inequality to the hypergeometric distribution in Theorem A.1, together with concrete numerical tables and a simple capacity formula. However, the significance claimed in the conclusion—that the framework lets creators, challengers, and jurors 'collectively approximate truth at scale'—is not supported by the analysis as it stands. The paper never models or bounds the probability that a jury majority is correct, so the incentive story rests on an unexamined epistemic assumption. The contribution at this stage is a design proposal with a partial collusion analysis, not an established result about incentive alignment with truth.

major comments (5)

[§3.1, Eq. (1)] The paper's central claim that 'the rewards for accurate assessments are funded exclusively by penalties for inaccurate content' is not entailed by Eq. (1). Equation (1) is a conservation identity that holds for any verdict, whether true or false; if the jury majority is wrong, the same identity rewards the false side. The self-sustaining accuracy claim therefore requires an explicit model (or at least a clearly stated assumption) of jury truth-tracking, such as per-juror accuracy q > 1/2 with conditionally independent votes, together with a bound on P(verdict = truth). No such model appears in Sections 2.2.3–2.2.5, and Section 6.1 concedes that dominant narratives and biased consensus remain unresolved risks.
[§2.2.5 and §A.1, Theorem A.1] Theorem A.1 bounds P(X ≥ m+1) for a fixed bloc of k colluding jurors, not P(verdict incorrect). It is silent on systematic honest error, correlated false beliefs, dominant wrong majorities, identity forgery, and collusion between jurors and disputing parties. Since the incentive story depends on verdicts tracking truth, the security guarantee should be restated as a conditional result: if all non-colluding jurors vote according to an independent signal with accuracy q > 1/2, then the probability of a wrong verdict is at most ... . Without such a statement, the phrase 'collusion-resistance guarantee' overstates what is proved.
[§B.1, Theorem B.1] The 'if and only if' stability claim is not established by the argument given. The condition N ≥ Nmin gives ρ ≤ 1 in the M/G/c approximation, but positive recurrence of the backlog requires a strict capacity margin, and the mapping from a pool of N jurors to c = ⌊N/n⌋ parallel servers glosses over juror availability constraints and dependence between cases. In addition, Eq. (3) appears dimensionally inconsistent: λ is disputes per hour and h is hours per case, while a is described as hours per day, so Nmin = ⌈λnh/a⌉ requires a to be expressed in compatible time units. The qualitative scaling conclusion may survive, but the formal theorem needs repair.
[§3.2, Eq. (2)] The reputation mechanism is described as essential for filtering accurate jurors, but Eq. (2) is not a well-posed model: the quantities γaE(va|a,y) and γyE(vy|a,y) are never given substantive definitions, no update rule is specified, and no argument shows that a higher R predicts accuracy. The Bénabou–Tirole citation supplies background, not a derivation. Either supply a concrete reputation process with an accuracy guarantee or explicitly mark reputation as a design suggestion that lies outside the paper's formal claims.
[Overall, Sections 2–3] The manuscript is submitted under cs.GT and invokes game theory, but no game is formally defined and no equilibrium or participation result is proved. The incentive-alignment statements in Section 3.1 are informal consequences of the payout identity rather than strategic analysis. To make the central claim defensible, the authors need at least a stylized game with utility functions for creators, challengers, and jurors—including effort costs and risk attitudes—and a statement of what equilibrium behavior the protocol induces. Alternatively, the claim should be explicitly labeled as a conjecture rather than a demonstrated property of the model.

minor comments (5)

[Abstract and §7] The abstract's 'paradigm shift' and the conclusion's 'collectively approximate truth at scale' are stronger than what the analysis supports and should be tempered, especially given the paper's own exploratory framing.
[§A.2 and Table 1] The text says the table lists 'exact' collusion probabilities, but values below 10^-15 are clamped to 10^-15 and several entries are displayed as '<10^-10'; the caption should distinguish exact values from clamped lower bounds.
[Glossary] The glossary entry for veracity bond is missing a space ('V eracity bondFinancial collateral'), and the juror entry has a double period; these typos should be corrected.
[§2.2.4] The three-point scale is described as rating juror quality and thoroughness, but the labels 'no/neutral/yes' are the same as verdict labels; this potential conflation should be clarified.
[§2.2.3] The notation 'n, m, N∈ N' should be typeset as 'n, m, N ∈ ℕ', and the condition n = 2m+1 ≪ N should be stated with explicit bounds rather than left as an informal ordering.

Circularity Check

1 steps flagged · score 2.0 of 10

No load-bearing circularity: Eq. 1 is an accounting identity, the collusion bound is a conventional hypergeometric tail bound, and the only self-citation ([29]) supports a peripheral credibility premise, not the core derivation.

other [Section 1.3 (Introduction, Overview of the Proposed Framework); echoed in §2.2.1]
"It has been demonstrated that readers can successfully discern high-quality and reputable content outlets from misleading ones [28] and that veracity bonds increase perceived credibility [29]."

The cited prior work is by the same first author (Barbosa et al. 2025, reference [29]) and is invoked to justify treating bond size as a credibility signal. This is a self-citation, but it is not the load-bearing proof of the model: Eq. (1), the jury-majority mechanism, and Theorem A.1 are stated and proved without relying on [29]. The paper explicitly labels the overall claim a hypothesis and lists jury bias among open limitations, so the self-citation is a minor supporting premise rather than a forced derivation.

full rationale

The paper is an exploratory protocol design, not a fitted predictive model. Eq. (1), πc + Σπj + πp = β, is an accounting identity describing how the forfeited bond is split; it is true by construction and the paper does not pretend to derive an external accuracy result from it. The claim that the system rewards accurate content does depend on the unproven assumption that jury majorities track truth, and the paper itself flags bias and dominant narratives as open risks in §6.1, but this is a correctness/evidence gap, not a circular step, since no parameter is fitted and no conclusion is defined in terms of its own output. Appendix A.1 is a standard Hoeffding bound on a hypergeometric tail and is self-contained and checkable; Appendix B is a routine queueing stability calculation. The only circularity-adjacent element is the self-citation [29] for the premise that veracity bonds increase perceived credibility, used to motivate visibility weighting; because the core incentive identity and collusion guarantees do not depend on that premise, this is a minor self-citation rather than load-bearing circularity. Accordingly, score 2.

Assumptions & free parameters 6 free parameters · 6 assumptions · 4 invented entities

The paper contributes a mechanism design rather than a measurement or a fitted model. Its free parameters are design constants left unspecified, and its key axioms are domain assumptions about jury accuracy, identity integrity, and the signaling value of bonds. The only standard-math assumptions are the hypergeometric Hoeffding bound and queueing results, one of which is applied incorrectly in the sufficiency direction of Theorem B.1.

free parameters (6)

Veracity bond principal beta = unspecified
No default value is set; all payouts scale with beta.
Jury size n = examples: 21, 31, 35
Authors suggest 21 for 5% corruption and 31 for 20%; the deployed value is left open.
Juror bond fraction gamma = unspecified (0 < gamma < 1)
Called an adjustable constant; its value affects juror participation and no default is given.
Payout shares pi_c, pi_j, pi_p = unspecified
Only constrained by Eq. 1; no concrete percentages are proposed.
Capacity inputs lambda, h, a for Table 2 = Reddit: lambda=108 disputes/h, h=0.5-2 h, a=2-8 h/day
Chosen for illustration, not measured; the resulting Nmin values are sensitive to these choices.
Challenge and deliberation period lengths = unspecified
Listed as open questions in Section 6.2; the model requires them but does not set them.

assumptions (6)

standard math Hoeffding's inequality for hypergeometric distributions bounds the tail of the colluding-juror count (Appendix A.1).
Unproved background result imported from references [42,43] and applied correctly to X ~ Hypergeometric(N,k,n).
standard math Little's law and positive recurrence of an M/G/c queue ensure finite expected backlog and latency (Appendix B.1).
Standard queueing results are invoked; however, the model treats jurors as continuously available and does not incorporate the per-juror time cap a, so the sufficiency direction is not established as written.
domain assumption Challenge arrivals follow a Poisson process and service times are independent with mean h (Appendix B.1).
Needed for the M/G/c queue model; not justified with platform data.
domain assumption There exists an externally checkable ground truth for contested claims that a lay jury can approximate with better-than-random accuracy (Sections 2.2.3 and 2.2.5).
The whole mechanism's usefulness depends on juries reaching verdicts correlated with truth; the collusion proof only bounds coordinated blocs, not honest error or dominant wrong beliefs.
domain assumption Digital identity (W3C DIDs plus verification) can prevent Sybil attacks well enough that the juror pool is not systematically flooded (Section 4).
The authors list this as a limitation in Section 6.1; the framework assumes the problem is solvable.
domain assumption Larger veracity bonds increase perceived credibility and scrutiny, so bond size can be a valid signal (Sections 3.3 and reference [29]).
Relies on the authors' own earlier empirical work; not independently established here.

invented entities (4)

Veracity bond (VB)
purpose: Financial collateral staked by creators to signal claim accuracy and to fund rewards for challengers and jurors.
A new financial instrument in this framework; the only supporting evidence is the same authors' earlier paper [29], which is not independent.
Counter-veracity bond (CVB)
purpose: Equal stake required from challengers to deter frivolous disputes and to preserve symmetrical incentives.
Introduced as part of the proposed protocol; no independent empirical validation.
Refundable juror bond (gamma*beta)
purpose: Collateral posted by jurors to ensure attendance, written assessments, and acceptable evaluation ratings.
A mechanism parameter; its design is described but never tested.
Reputation score R from Eq. 2
purpose: Used to filter and rank jurors by perceived prosocial motivation rather than purely financial motives.
Adapted from Benabou and Tirole [44], but the specific functional form in Eq. 2 is a proposal, not a measured quantity.

how reviews work

0 comments

Cite this review

Pith. "Pith review of A New Incentive Model For Content Trust." pith.science (2026). https://pith.science/paper/YMXZFNY2

@misc{pith2026250709972,
  author       = {Pith},
  title        = {Pith review of: A New Incentive Model For Content Trust},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/YMXZFNY2}},
  note         = {Machine review of arXiv:2507.09972}
}

read the original abstract

This paper outlines an incentive-driven and decentralized approach to verifying the veracity of digital content at scale. Widespread misinformation, an explosion in AI-generated content and reduced reliance on traditional news sources demands a new approach for content authenticity and truth-seeking that is fit for a modern, digital world. By using smart contracts and digital identity to incorporate 'trust' into the reward function for published content, not just engagement, we believe that it could be possible to foster a self-propelling paradigm shift to combat misinformation through a community-based governance model. The approach described in this paper requires that content creators stake financial collateral on factual claims for an impartial jury to vet with a financial reward for contribution. We hypothesize that with the right financial and social incentive model users will be motivated to participate in crowdsourced fact-checking and content creators will place more care in their attestations. This is an exploratory paper and there are a number of open issues and questions that warrant further analysis and exploration.

Figures

Figures reproduced from arXiv: 2507.09972 by the authors.

**Figure 1.** Creators can deposit a veracity bond immediately after writing the content. If the content is found to be misleading, the bond is forfeited and redistributed to the challenger who contested and the jurors who deliberated. This forfeiture occurs only if the jury verdict rules against the creator, which treats the bonded content as a formal legal contract. Recipients should receive payment through a blockchain-based s… view at source ↗

**Figure 2.** Challengers can deposit a counter-veracity bond immediately after writing the challenge. The requirement for a counter-veracity bond also discourages frivolous or malicious challenges. Those initiating a challenge must be sufficiently confident in their contest to risk losing the bond. This mechanism promotes thorough fact-checking and reduces exploitation attempts. In addition, each challenger can submit only one … view at source ↗

**Figure 3.** A fixed percentage of the losing bond (VB or CVB) is distributed to the jury and the winning party. Jurors would ideally participate by reviewing evidence, providing detailed assessments, and casting informed votes within the deliberation period. Should a juror fail to fulfill these responsibilities, they are deemed inactive and must be promptly replaced by the next best available candidate from the pre-established… view at source ↗

Figures from the paper (3 more)

**Figure 4.** Figure 4: Juror evaluations are randomly assigned to a selected number of viewers who do not have a conflict of interest with that contest. Jurors are required to compose detailed assessments explaining their verdicts, outlining the evidence considered, the relative weight assig…

**Figure 5.** Figure 5: Collusion probability (on a log-scale) as a function of jury selection ratio (n/N) for varying population sizes and corruption levels. The three panels show results for populations of N = 100, 1 000, and 10 000 individuals, each with different percentages of unfaithful…

**Figure 6.** Figure 6: Adapted from [PITH_FULL_IMAGE:figures/full_fig_p010_6.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

53 extracted references · 45 canonical work pages

[29]

Toward trustworthy content: the role of challengers, juries and veracity bonds in digital media platforms

Lucas Barbosa et al. “Toward trustworthy content: the role of challengers, juries and veracity bonds in digital media platforms”. In: Industrial Management & Data Systems (2025)

work page 2025
[1]

Who can you trust?: how technology brought us together–and why it could drive us apart

Rachel Botsman. Who can you trust?: how technology brought us together–and why it could drive us apart . Penguin UK, 2017

work page 2017
[2]

Can Language Models Recognize Convincing Arguments? 2024

Paula Rescala et al. Can Language Models Recognize Convincing Arguments? 2024. arXiv: 2404.00750 [cs.CL]

arXiv 2024
[3]

On the conversational persuasiveness of GPT-4

Francesco Salvi et al. “On the conversational persuasiveness of GPT-4”. In: Nature Human Behaviour (May 19, 2025). issn: 2397-3374. doi: 10.1038/s41562-025-02194-6

doi:10.1038/s41562-025-02194-6 2025
[4]

Growth, degrowth, and the challenge of artificial superintelligence

Salvador Pueyo. “Growth, degrowth, and the challenge of artificial superintelligence”. In: Journal of Cleaner Production 197 (2018), pp. 1731–1736

work page 2018
[5]

Merging with the machines: Information technology, artificial intelligence, and the law of exponential growth

Ray Kurzweil. “Merging with the machines: Information technology, artificial intelligence, and the law of exponential growth”. In: World Future Review 2.1 (2010), pp. 61–66

work page 2010
[6]

Deep fakes and the infocalypse: What you urgently need to know

Nina Schick. Deep fakes and the infocalypse: What you urgently need to know . Hachette UK, 2020

work page 2020
[7]

Adapting Fake News Detection to the Era of Large Language Models

Jinyan Su, Claire Cardie, and Preslav Nakov. “Adapting Fake News Detection to the Era of Large Language Models”. In: (2023). arXiv: 2311.04917 [cs.CL]

arXiv 2023

Show all 53 references

[8]

Understanding the Capabilities, Limitations, and Societal Impact of Large Language Models

Alex Tamkin et al. “Understanding the Capabilities, Limitations, and Societal Impact of Large Language Models”. In: (2021). arXiv: 2102.02503 [cs.CL]

2021 arXiv
[9]

The spread of fake news by social bots

Chengcheng Shao et al. “The spread of fake news by social bots”. In: arXiv preprint arXiv:1707.07592 96 (2017), p. 104

2017 arXiv
[10]

Social media and fake news in the 2016 election

Hunt Allcott and Matthew Gentzkow. “Social media and fake news in the 2016 election”. In: Journal of economic perspectives 31.2 (2017), pp. 211–236

2017
[11]

Lack of trust in the news media, institutional weakness, and relational journalism as a potential way forward

Seth C Lewis. “Lack of trust in the news media, institutional weakness, and relational journalism as a potential way forward”. In: Journalism 20.1 (2019), pp. 44–47

2019
[12]

Memory for fact, fiction, and misinformation: The Iraq War 2003

Stephan Lewandowsky et al. “Memory for fact, fiction, and misinformation: The Iraq War 2003”. In: Psychological Science 16.3 (2005), pp. 190–195

2005
[13]

The big short: Inside the doomsday machine

Michael Lewis. The big short: Inside the doomsday machine . Penguin UK, 2011

2011
[14]

Bias, Bullshit and Lies: Audience Perspectives on Low Trust in the Media

Nic Newman and Richard Fletcher. “Bias, Bullshit and Lies: Audience Perspectives on Low Trust in the Media”. In: SSRN Electronic Journal (Jan. 2017), pp. 10–26

2017
[15]

2023 Edelman Trust Barometer Australian Report

Edelman. 2023 Edelman Trust Barometer Australian Report. Edelman, 2023. url: https://www.edelman. com.au/trust/2023/trust-barometer

2023
[16]

The spread of true and false news online

Soroush Vosoughi, Deb Roy, and Sinan Aral. “The spread of true and false news online”. In: Science 359.6380 (2018), pp. 1146–1151. doi: 10.1126/science.aap9559

2018 doi
[17]

Going to extremes: How like minds unite and divide

Cass R Sunstein. Going to extremes: How like minds unite and divide . Oxford University Press, 2009

2009
[18]

The filter bubble: How the new personalized web is changing what we read and how we think

Eli Pariser. The filter bubble: How the new personalized web is changing what we read and how we think . Penguin, 2011

2011
[19]

The disaster of misinformation: a review of research in social media

Sadiq Muhammed T and Saji K Mathew. “The disaster of misinformation: a review of research in social media”. In: International journal of data science and analytics 13.4 (2022), pp. 271–285

2022
[20]

C2PA.org

Coalition for Content Provenance and Authenticity (C2PA). C2PA.org. 2023. url: https://c2pa.org

2023
[21]

Cardano Founder Charles Hoskinson Renews Interest in Purchasing CoinDesk

Binance. Cardano Founder Charles Hoskinson Renews Interest in Purchasing CoinDesk . 2024. url: https: //www.binance.com/en/square/post/17875447050538

2024
[22]

About Community Notes

Twitter. About Community Notes . 2023. url: https://communitynotes.twitter.com/guide/en/about/ introduction

2023
[23]

Birdwatch: Crowd Wisdom and Bridging Algorithms can Inform Understanding and Reduce the Spread of Misinformation

Stefan Wojcik et al. Birdwatch: Crowd Wisdom and Bridging Algorithms can Inform Understanding and Reduce the Spread of Misinformation . 2022. arXiv: 2210.15723 [cs.SI]

2022 arXiv
[24]

Do explanations increase the effectiveness of AI-crowd generated fake news warnings?

Ziv Epstein et al. “Do explanations increase the effectiveness of AI-crowd generated fake news warnings?” In: 2021. arXiv: 2112.03450 [cs.HC]

2021 arXiv
[25]

Motivated skepticism in the evaluation of political beliefs

Charles S Taber and Milton Lodge. “Motivated skepticism in the evaluation of political beliefs”. In: Amer- ican journal of political science 50.3 (2006), pp. 755–769

2006
[26]

X’s community notes is spreading false information about Taylor Swift’s bodyguard

Bellingcat. X’s community notes is spreading false information about Taylor Swift’s bodyguard . 2023. url: https : / / www . bellingcat . com / news / 2023 / 10 / 20 / xs - community - notes - is - spreading - false - information-about-taylor-swifts-bodyguard/

2023
[27]

Americans’ perspectives on online media warning labels

Jeremy Straub and Matthew Spradling. “Americans’ perspectives on online media warning labels”. In: Behavioral Sciences 12.3 (2022), p. 59. 14

2022
[28]

Fighting misinformation on social media using crowdsourced judgments of news source quality

Gordon Pennycook and David G Rand. “Fighting misinformation on social media using crowdsourced judgments of news source quality”. In: Proceedings of the National Academy of Sciences 116.7 (2019), pp. 2521–2526

2019
[30]

Lateral reading and monetary incentives to spot disinformation about science

Folco Panizza et al. “Lateral reading and monetary incentives to spot disinformation about science”. In: Scientific Reports 12.1 (2022), p. 5678

2022
[31]

Incentives, Gamification, and Game Theory: An Economic Approach to Badge Design

David Easley and Arpita Ghosh. “Incentives, Gamification, and Game Theory: An Economic Approach to Badge Design”. In: ACM Trans. Econ. Comput. 4.3 (June 2016)

2016
[32]

Short - & long-term effects of monetary and non-monetary incentives to cooperate in public good games: An experiment

Mathieu Lefebvre and Anne Stenger. “Short - & long-term effects of monetary and non-monetary incentives to cooperate in public good games: An experiment”. In: 15.1 (2020), pp. 1–17

2020
[33]

Goal setting and monetary incentives: When large stakes are not enough

Brice Corgnet, Joaqu ´ ın G´ omez-Mi˜ nambres, and Roberto Hern´ an-Gonzalez. “Goal setting and monetary incentives: When large stakes are not enough”. In: Management Science 61.12 (2015), pp. 2926–2944

2015
[34]

Game-Theoretic Analysis of Cooperation Incentive Strategies in Mobile Ad Hoc Networks

Ze Li and Haiying Shen. “Game-Theoretic Analysis of Cooperation Incentive Strategies in Mobile Ad Hoc Networks”. In: IEEE Transactions on Mobile Computing 11.8 (2012), pp. 1287–1303

2012
[35]

Making it work in free agent work: The coping practices of Swedish freelance journalists

Maria Norb¨ ack and Alexander Styhre. “Making it work in free agent work: The coping practices of Swedish freelance journalists”. In: Scandinavian Journal of Management 35.4 (2019), p. 101076

2019
[36]

Bitcoin: A peer-to-peer electronic cash system

Satoshi Nakamoto. Bitcoin: A peer-to-peer electronic cash system . 2008. url: https : / / bitcoin . org / bitcoin.pdf

2008
[37]

Ethereum White Paper: A Next-Generation Smart Contract and Decentralized Application Platform

Vitalik Buterin. Ethereum White Paper: A Next-Generation Smart Contract and Decentralized Application Platform. White Paper. Ethereum, 2013. url: https://ethereum.org/en/whitepaper/

2013
[38]

Creating markets in no-trust environments: The law and economics of smart contracts

Helen Eenmaa-Dimitrieva and Maria Jos´ e Schmidt-Kessen. “Creating markets in no-trust environments: The law and economics of smart contracts”. In: Computer law & security review 35.1 (2019), pp. 69–88

2019
[39]

Freelance journalism in the 21st century: Challenges and opportunities

Tiana L Templeman. “Freelance journalism in the 21st century: Challenges and opportunities”. In: (2016)

2016
[40]

Do differences make a difference

Deborah Son Holoien. “Do differences make a difference”. In: The effects of diversity on learning, intergroup outcomes, and civic engagement. Princeton, NJ: Princeton University (2013)

2013
[41]

Avocado consumption and risk of cardiovascular disease in US adults

Lorena S Pacheco et al. “Avocado consumption and risk of cardiovascular disease in US adults”. In: Journal of the american heart association 11.7 (2022)

2022
[42]

Probability Inequalities for Sums of Bounded Random Variables

Wassily Hoeffding. “Probability Inequalities for Sums of Bounded Random Variables”. In: Journal of the American Statistical Association 58.301 (1963), pp. 13–30. issn: 01621459, 1537274X

1963
[43]

The tail of the hypergeometric distribution

V. Chv´ atal. “The tail of the hypergeometric distribution”. In: Discrete Mathematics 25.3 (1979), pp. 285–

1979
[44]

Incentives and prosocial behavior

Roland B´ enabou and Jean Tirole. “Incentives and prosocial behavior”. In: American economic review 96.5 (2006), pp. 1652–1678

2006
[45]

The sybil attack

John R Douceur. “The sybil attack”. In: International workshop on peer-to-peer systems . Springer. 2002, pp. 251–260

2002
[46]

Decentralized identity: Where did it come from and where is it going?

Oscar Avellaneda et al. “Decentralized identity: Where did it come from and where is it going?” In: IEEE Communications Standards Magazine 3.4 (2019), pp. 10–13

2019
[47]

Decentralized Identifiers (DIDs) v1.0

W3C. Decentralized Identifiers (DIDs) v1.0 . 2022. url: https://www.w3.org/TR/did-1.0/

2022
[48]

Decentralized Identity and Verifiable Credentials

Microsoft Security. Decentralized Identity and Verifiable Credentials. Whitepaper. Microsoft, 2022

2022
[49]

Verifiable Credentials Data Model 1.0

W3C. Verifiable Credentials Data Model 1.0 . 2019. url: https://www.w3.org/TR/vc-data-model-1.0/

2019
[50]

Blockchain and distributed ledgers as trusted recordkeeping systems

Victoria L Lemieux. “Blockchain and distributed ledgers as trusted recordkeeping systems”. In: Future technologies conference (FTC). Vol. 2017. 2017, pp. 3–10

2017
[51]

PROV-O: The PROV Ontology

W3C. PROV-O: The PROV Ontology . 2013. url: https://www.w3.org/TR/prov-dictionary/

2013
[52]

Fundamentals of queueing theory

John F Shortle et al. Fundamentals of queueing theory . John Wiley & Sons, 2018. 15 A Collusion-Resistance Guarantee A.1 Formal Proof Theorem A.1 (Collusion-Resistance Guarantee) . Let a pool of N jurors contain exactly k colluders, and draw uniformly at random and without rep...

2018
[287]

doi: https://doi.org/10.1016/0012-365X(79)90084-0

issn: 0012-365X. doi: https://doi.org/10.1016/0012-365X(79)90084-0

doi

Pith tools

Reviewed August 6, 2026 · model on record in the stance chip above.

[1] [29]

Toward trustworthy content: the role of challengers, juries and veracity bonds in digital media platforms

Lucas Barbosa et al. “Toward trustworthy content: the role of challengers, juries and veracity bonds in digital media platforms”. In: Industrial Management & Data Systems (2025)

work page 2025

[2] [1]

Who can you trust?: how technology brought us together–and why it could drive us apart

Rachel Botsman. Who can you trust?: how technology brought us together–and why it could drive us apart . Penguin UK, 2017

work page 2017

[3] [2]

Can Language Models Recognize Convincing Arguments? 2024

Paula Rescala et al. Can Language Models Recognize Convincing Arguments? 2024. arXiv: 2404.00750 [cs.CL]

arXiv 2024

[4] [3]

On the conversational persuasiveness of GPT-4

Francesco Salvi et al. “On the conversational persuasiveness of GPT-4”. In: Nature Human Behaviour (May 19, 2025). issn: 2397-3374. doi: 10.1038/s41562-025-02194-6

doi:10.1038/s41562-025-02194-6 2025

[5] [4]

Growth, degrowth, and the challenge of artificial superintelligence

Salvador Pueyo. “Growth, degrowth, and the challenge of artificial superintelligence”. In: Journal of Cleaner Production 197 (2018), pp. 1731–1736

work page 2018

[6] [5]

Merging with the machines: Information technology, artificial intelligence, and the law of exponential growth

Ray Kurzweil. “Merging with the machines: Information technology, artificial intelligence, and the law of exponential growth”. In: World Future Review 2.1 (2010), pp. 61–66

work page 2010

[7] [6]

Deep fakes and the infocalypse: What you urgently need to know

Nina Schick. Deep fakes and the infocalypse: What you urgently need to know . Hachette UK, 2020

work page 2020

[8] [7]

Adapting Fake News Detection to the Era of Large Language Models

Jinyan Su, Claire Cardie, and Preslav Nakov. “Adapting Fake News Detection to the Era of Large Language Models”. In: (2023). arXiv: 2311.04917 [cs.CL]

arXiv 2023