Can Funding Methods Be Ranked by Fairness?

Getting your Trinity Audio player ready...

Yes, funding methods can be compared and provisionally ranked by fairness—but only after defining what “fair” means. A system may be procedurally impartial while distributing money inefficiently. Another may reward scientific merit accurately but exclude independent researchers who lack institutional credentials.

There is therefore no universal fairness ranking independent of values. A useful comparison must examine several dimensions:

• Access fairness: Who is allowed to compete?
• Procedural fairness: Are decisions made consistently and without irrelevant bias?
• Epistemic fairness: Can the system recognize technically valuable work?
• Distributive fairness: Does funding roughly correspond to contribution, need, or expected benefit?
• Corrective fairness: Can mistaken decisions be challenged and revised?
• Cost fairness: How much researcher time and public money does the selection process consume?
• Temporal fairness: How long must contributors wait before receiving support?

Once these dimensions are separated, research funding systems can be ranked more meaningfully.

A Provisional Fairness Ranking of Research Funding Methods

The following ranking compares idealized versions of common funding methods. Actual implementations may perform much better or worse.

Approximate rank	Funding method	Principal fairness advantage	Principal fairness weakness
1	Transparent, contribution-based continuous funding	Rewards observable work across institutions and career stages	Evaluation models may contain errors or manipulation risks
2	Participatory funding with expert safeguards	Distributes decision-making power beyond a small committee	Popular groups may dominate participation
3	Qualified or partial lottery	Treats similarly qualified proposals equally	Does not distinguish merit within the eligible pool
4	Blinded expert peer review	Can evaluate technical quality while reducing prestige effects	True anonymity is difficult, especially in small fields
5	Conventional grant peer review	Uses relevant disciplinary expertise	Favors established institutions, grant-writing skill, and familiar topics
6	Public crowdfunding	Open to outsiders and responsive to public interest	Rewards communication, networks, and emotional appeal
7	Institutional block funding	Provides stability and supports infrastructure	Allocation may depend heavily on hierarchy and institutional status
8	Political or administrative discretion	Can pursue democratically selected public priorities	Exposed to ideology, lobbying, nationalism, and short political cycles
9	Patronage and private discretion	Fast and flexible	Highly dependent on personal access and donor preference

This table should not be interpreted as a mathematically proved total ordering. Some methods are incomparable because they optimize different forms of fairness.

Why Fairness Is Multidimensional

Consider a research lottery. Once proposals have crossed a minimum quality threshold, each may receive the same probability of funding. This is highly fair in the narrow procedural sense: reviewers cannot secretly favor one eligible applicant over another.

However, a lottery may be less fair in the distributive sense. A proposal that is substantially more useful, urgent, or technically important can receive exactly the same chance as a merely adequate proposal.

Now consider expert peer review. Reviewers may distinguish sophisticated work from weak work, producing greater epistemic discrimination. But the process may also favor:

• famous researchers;
• prestigious universities;
• fashionable disciplines;
• conventional methodologies;
• projects that can be explained in a short proposal;
• applicants who have professional grant-writing support.

A process can therefore become more merit-sensitive but less access-fair.

The central mistake is to ask whether a method is simply “fair” or “unfair.” The better question is:

Fair to whom, with respect to which resources, under what definition of merit, and over what period?

Conventional Peer Review: Expert but Structurally Unequal

Conventional grant peer review remains defensible because scientific proposals often require specialized evaluation. A random voter cannot reliably assess a difficult mathematical proof, an experimental protocol, or a new compiler architecture.

The problem is not expertise itself. The problem is the concentration of funding decisions within temporary committees operating under severe informational and time constraints.

Peer reviewers usually evaluate promises about future work rather than completed contributions. They must predict which projects will succeed, even though scientific impact is uncertain and highly skewed. They may also use institutional affiliation, publication venue, previous grants, and professional reputation as proxies for quality.

This creates cumulative advantage. Researchers who have already obtained funding are better positioned to produce preliminary results, employ assistants, submit polished applications, and apply again. A large cross-national analysis reported evidence of a Matthew effect across research funders, although it found that repeated application by previously funded researchers—not necessarily direct reviewer favoritism—was an important mechanism.

Peer review is therefore not arbitrary, but neither is it neutral.

Fairness profile of conventional peer review

• Access fairness: Low to moderate
• Procedural fairness: Moderate
• Epistemic fairness: Moderate to high in established fields
• Corrective fairness: Usually low
• Temporal fairness: Low
• Administrative efficiency: Low to moderate

Peer review performs best when proposals are readily comparable and reviewers can genuinely understand the work. It performs worse for obscure, interdisciplinary, highly original, or exceptionally long-term research.

Blinded Peer Review: Fairer, but Only Partly Blind

Removing applicants’ names and affiliations can reduce prestige-based judgment. Blinding is therefore usually fairer than openly identity-based review when identities are irrelevant to the decision.

Yet complete anonymity is difficult. Reviewers may infer authorship from:

• citations to previous work;
• specialized terminology;
• datasets or research facilities;
• geographic context;
• the narrowness of a research community.

Blinding also does not eliminate conformity bias. An anonymous unconventional proposal can still be rejected because reviewers prefer familiar theories or methods.

Blinded review should therefore be treated as an improvement to peer review, not a complete solution.

Partial Lotteries: Fair Among Roughly Equal Proposals

A partial lottery normally has two stages:

1. experts remove proposals that fail a quality or eligibility threshold;
2. funding is randomly allocated among the remaining proposals.

This method acknowledges an uncomfortable reality: review panels often cannot reliably produce an exact ranking among several strong applications.

Modified lotteries have been proposed as a way to reduce bias, lower evaluation costs, and improve diversity while preserving a minimum scientific standard. The Volkswagen Foundation has also experimented with partially randomized selection in research funding.

A lottery is especially fair when the differences between proposals are smaller than the uncertainty of the evaluation process. Pretending that reviewers can identify the 17th-best proposal but not the 18th may produce an appearance of precision without real knowledge.

However, lottery funding has clear limitations:

• it treats all eligible proposals as equal;
• it does not automatically reward previous useful contributions;
• it may fund fewer highly valuable projects by chance;
• it can be difficult to explain to taxpayers or donors;
• fairness depends heavily on how the eligibility threshold is established.

A lottery is therefore a strong tie-breaking mechanism, but a weak complete theory of scientific merit.

Crowdfunding: Open Access, Unequal Attention

Scientific crowdfunding removes several institutional barriers. Independent researchers, small laboratories, students, and open-source developers can present their work directly to the public.

That makes crowdfunding relatively strong in access fairness. It can support projects that grant committees consider too small, unconventional, or difficult to classify.

But crowdfunding allocates money according to a mixture of scientific value and marketability. Success can depend on:

• an existing audience;
• professional video production;
• English-language communication;
• emotionally attractive subjects;
• social-media reach;
• the ability to simplify a project;
• personal charisma.

A visually compelling wildlife project may be easier to fund than an abstract mathematical definition on which future theories could depend. Public attention is scarce, unevenly distributed, and not equivalent to scientific merit.

Crowdfunding is therefore fair as a right to ask, but not necessarily fair as a mechanism for deciding who receives money.

For a broader discussion, see where a donation to scientific research actually goes.

Participatory Funding: Fairer Power Distribution

Participatory grantmaking transfers some decision-making authority from professional funders to affected communities, applicants, or other stakeholders.

Its central fairness claim is political rather than purely epistemic:

People affected by funding decisions should have meaningful influence over those decisions.

Participatory systems can reveal needs that centralized institutions overlook. They can also reduce the distance between funders and recipients. Evidence reviews describe promising benefits, including stronger relationships and shifts in decision-making power, while warning that the evidence is not yet sufficient to conclude that participatory grantmaking is universally more effective.

Participation nevertheless creates its own inequalities. People with more time, confidence, language ability, or organizational support may dominate deliberation. Majority voting can also disadvantage niche fields whose importance is not widely understood.

Participatory funding becomes more epistemically credible when community input is combined with:

• technical review;
• conflict-of-interest rules;
• transparent reasoning;
• minority protections;
• appeal procedures;
• limits on coordinated manipulation.

It is fairer than closed patronage in its distribution of authority, but it is not automatically fair in its evaluation of technical merit.

Political Allocation: Democratic Legitimacy Is Not Scientific Fairness

Governments have a legitimate role in selecting broad priorities. Citizens may reasonably decide that public funds should support health, climate resilience, energy security, agriculture, or fundamental research.

But democratic control over the total budget does not imply that politicians should select individual scientific winners.

Political allocation can favor:

• projects with visible short-term outcomes;
• research concentrated in electorally important regions;
• ideologically acceptable conclusions;
• national prestige;
• established organizations with lobbying capacity;
• subjects that fit a government’s current narrative.

Scientific timescales are often longer than electoral cycles. Fundamental work may have no immediate constituency and no predictable application.

The fairest division of responsibilities may therefore be:

• democratic institutions choose broad budgets and social priorities;
• transparent expert or computational systems evaluate technical contributions;
• independent oversight detects abuse;
• the public can inspect and challenge the rules.

Replacing academic gatekeepers with political gatekeepers does not remove gatekeeping. It changes who controls it. This distinction is examined further in why replacing academic gatekeepers with political gatekeepers solves nothing.

Institutional Funding: Stable but Credential-Dependent

Universities and research institutes need stable funding. Laboratories, libraries, archives, computing systems, datasets, and long-term personnel cannot operate entirely through project competitions.

Institutional block funding is therefore often fair to scientific infrastructure. It prevents every useful activity from having to justify itself as an isolated project.

However, it is less fair to people outside recognized institutions. An independent mathematician and a university professor may produce work of similar value, yet only one has access to salary, affiliation, databases, administrative support, and institutional grant eligibility.

This is a form of credential-based distributive inequality. The system funds a person’s organizational position before it evaluates each contribution.

Institutional funding should not disappear, but it should be complemented by mechanisms that can support:

• independent researchers;
• open-source developers;
• maintainers of scientific software;
• reviewers and replicators;
• creators of datasets;
• researchers in underfunded countries;
• contributors working across disciplinary boundaries.

Continuous Contribution-Based Funding

Continuous contribution-based funding evaluates observable scientific or technical work and distributes money repeatedly rather than through occasional winner-take-all grants.

In principle, this method has several fairness advantages:

1. Researchers can be rewarded after producing useful work.
   The system depends less on persuasive predictions.
2. Small contributions can receive small rewards.
   Recognition does not have to be concentrated in a few grant winners.
3. Independent contributors can participate.
   Institutional affiliation need not be a prerequisite.
4. Funding can follow dependencies.
   Foundational software, proofs, definitions, datasets, replications, and reviews can receive credit when later work depends on them.
5. Decisions can be continuously revised.
   New evidence about a contribution’s importance can change its evaluation.
6. Administrative costs may be reduced.
   Researchers need not repeatedly suspend their work to prepare large applications.

This is the objective of AI Internet-Meritocracy, or AIIM, a proposed system for distributing donations among scientists and free and open-source software developers according to evaluated contributions. The current application is designed to gather donations and allocate them according to an AI-assisted assessment.

Continuous allocation can be fairer than episodic grant competitions because it replaces a binary distinction—funded or rejected—with divisible recognition.

Yet it introduces serious risks.

Risks of AI-assisted merit allocation

An evaluation system may:

• reproduce biases in its data;
• misunderstand obscure research;
• reward measurable activity rather than genuine importance;
• be manipulated through coordinated promotion;
• confuse citations or online visibility with dependency;
• disadvantage work that cannot be published openly;
• make incorrect judgments with an appearance of mathematical authority.

Transparency is therefore essential. Users should be able to see why an assessment was made, identify relevant evidence, dispute errors, and propose corrections.

AI should not be treated as an infallible scientific judge. Its legitimate role is closer to that of a scalable analytical instrument operating within an auditable governance process.

Can AIIM Be Called the Fairest Method?

AIIM can plausibly rank highly on several dimensions:

Fairness dimension	Potential AIIM performance
Access across institutions	High
Recognition of small contributions	High
Speed of funding	High
Support for open-source work	High
Transparency	Potentially high, depending on implementation
Evaluation of obscure technical work	Potentially high but not guaranteed
Resistance to manipulation	Unproven
Protection from model bias	Unproven
Democratic legitimacy	Depends on governance and sources of funds
Reliability at scale	Requires testing

It would be premature to declare AIIM categorically fairer than every alternative before empirical testing. Its advantages concern architecture and incentives; its actual fairness depends on implementation.

A responsible claim is narrower:

AIIM is designed to improve access, divisibility, speed, and contribution-sensitive distribution—four dimensions on which conventional grant systems perform poorly.

The project’s proposed adversarial testing is therefore important. A system that claims to distribute money fairly should be tested by people actively attempting to expose manipulation, evaluation errors, governance failures, and unequal treatment.

Why No Single Method Should Control All Research Funding

Different funding methods solve different problems.

• Institutional funding supports durable infrastructure.
• Peer review evaluates technically detailed proposals.
• Lotteries resolve uncertainty among similarly qualified applications.
• Participatory funding gives affected communities influence.
• Crowdfunding lets the public support neglected projects directly.
• Continuous contribution-based funding rewards work after it becomes observable.
• Political budgeting establishes broad public priorities.

The fairest research ecosystem may therefore be pluralistic rather than uniform.

A diversified system reduces the danger that one evaluator, metric, institution, ideology, or technical failure can exclude an entire class of valuable work. Researchers rejected by one mechanism may remain visible to another.

This resembles diversification in risk management: multiple imperfect selection systems may collectively be fairer than one supposedly optimal system with absolute authority.

A Better Way to Measure Funding Fairness

Funding organizations should publish a fairness scorecard rather than relying on general claims about excellence.

Useful indicators include:

• the percentage of funds reaching first-time recipients;
• the percentage reaching independent researchers;
• geographic and institutional concentration;
• the median time between application and payment;
• application and review costs;
• the proportion of decisions accompanied by explanations;
• successful appeals and corrected evaluations;
• funding allocated to replication, maintenance, data, and software;
• funding concentration among the top 1%, 5%, and 10% of recipients;
• applicant outcomes grouped by career stage and institutional status;
• the predictive reliability of evaluation scores;
• recipient and applicant perceptions of procedural fairness.

These measurements do not settle every philosophical dispute, but they make funding systems comparable.

They also expose trade-offs. A system may reduce administrative expense while increasing concentration. Another may broaden participation while weakening technical discrimination.

Conclusion: Funding Methods Can Be Ranked, but Not on One Axis

Funding methods can be ranked by fairness only under an explicit, multidimensional standard. There is no neutral ordering that applies to every purpose.

Conventional peer review has valuable expertise but weak access and temporal fairness. Lotteries are procedurally impartial but intentionally insensitive to differences within the eligible pool. Crowdfunding democratizes entry but rewards visibility. Participatory funding distributes power more broadly but can reproduce majoritarian and participation inequalities. Institutional funding provides stability while excluding many outsiders.

Continuous contribution-based systems such as AIIM may offer a stronger combination of open access, divisible rewards, rapid distribution, and recognition of nontraditional scientific work. Their fairness, however, must be demonstrated through transparent operation and adversarial testing rather than assumed from their design.

The correct objective is not to discover a perfectly fair allocator. It is to construct a funding ecosystem in which:

• relevant differences in contribution can matter;
• irrelevant differences in status do not;
• no single gatekeeper has final authority;
• errors can be detected and corrected;
• useful work can receive proportionate support.

That is a more defensible definition of fair scientific funding than either pure expert discretion or pure equality of chance.