Why Funding Agencies Should Explain Every Rejection—and How AIIM Already Does This

Getting your Trinity Audio player ready...

Scientific funding agencies should provide a meaningful explanation for every rejected application. A rejection should identify the decisive reasons, the evidence or criteria behind them, and whether the problem concerns scientific quality, feasibility, eligibility, portfolio priorities, or limited funds.

A bare statement such as “your proposal was not selected” is not adequate scientific governance.

Researchers invest weeks or months preparing applications. When an agency rejects that work without explaining its reasoning, it wastes information, conceals possible errors, and prevents both applicants and the public from evaluating whether funds were allocated consistently.

Some major funders already provide reviewer comments or summary statements. The United States National Institutes of Health supplies reviewed applicants with a summary statement, while the National Science Foundation provides declined applicants with reviews and an explanation of the basis for the decision. UK Research and Innovation programs also provide various forms of assessor feedback. These practices are valuable—but feedback remains inconsistent across agencies, programs, review stages, and countries.

AI Internet-Meritocracy, or AIIM, approaches the problem differently. Rather than treating evaluation as a private committee judgment followed by a binary acceptance or rejection, AIIM makes the evaluation itself part of the output. Its purpose is to produce an inspectable assessment of scientific contribution, impact, dependencies, and demonstrated usefulness.

A Funding Rejection Is a Scientific Decision

A grant rejection is not merely an administrative event. It is a decision about:

• which research questions deserve attention;
• which methods appear credible;
• which risks society should accept;
• which researchers receive time and institutional stability;
• and which branches of knowledge may develop further.

Funding decisions therefore help shape the future structure of science.

When their reasoning is hidden, funding agencies exercise substantial epistemic power without corresponding accountability. Researchers cannot determine whether a proposal failed because of a genuine methodological weakness, a misunderstanding, a conflict with program priorities, an unusually competitive round, or simple budget scarcity.

These are not equivalent reasons.

A technically defective proposal may need correction. A strong proposal below the funding cutoff may need another source of support. A proposal rejected because it falls outside a call’s scope may be entirely valid science submitted to the wrong program. A high-risk proposal may be rejected not because it is unsound, but because a portfolio committee preferred safer projects.

A responsible decision letter should make these distinctions explicit.

What Every Rejection Explanation Should Contain

A useful explanation does not need to be a long essay. It needs to be specific enough to reconstruct the decision.

At minimum, a funding agency should disclose five elements.

1. The criteria applied

The applicant should know which formal criteria controlled the decision. These might include scientific merit, novelty, methodology, feasibility, investigator capability, expected impact, cost, ethics, or program relevance.

Publishing generic criteria on a website is not enough. The rejection should explain how those criteria were applied to this particular proposal.

2. The decisive weaknesses

Reviewer comments frequently contain numerous minor criticisms. Agencies should distinguish incidental comments from the issues that actually determined the result.

For example:

The proposal was rejected primarily because the preliminary data did not support the central feasibility assumption.

That statement is substantially more useful than:

Reviewers raised concerns regarding feasibility.

The first identifies a correctable evidentiary problem. The second merely repeats an evaluative label.

3. The strengths recognized

A rejection explanation should also record the proposal’s strengths. This helps prevent a rejected application from being interpreted as scientifically worthless.

A proposal may be original but underdeveloped, rigorous but outside the program’s scope, or valuable but less urgent than competing submissions. Preserving these distinctions helps other funders, institutions, and future reviewers interpret the result correctly.

4. The role of budget and portfolio considerations

Funding decisions are comparative. A proposal can be scientifically strong and still lose because the agency has insufficient money or seeks a balanced portfolio.

The National Science Foundation explicitly notes that program officers may consider factors such as portfolio balance, alternative approaches, transformative potential, capacity building, and special program objectives in addition to external reviews.

When such considerations determine the outcome, the applicant should be told. An agency should not disguise a portfolio decision as a defect in the science.

5. Whether revision would change the result

Applicants need to know whether resubmission is appropriate.

A rejection might mean:

• revise the methodology;
• provide stronger evidence;
• clarify the presentation;
• apply to a different program;
• wait for a more suitable call;
• or do not resubmit because the proposal conflicts with a fixed policy.

Without this information, researchers may waste additional months revising proposals that have little realistic chance of success.

Why Explanation Improves Scientific Quality

Constructive feedback can convert the cost of evaluation into reusable scientific knowledge.

Reviewers have already spent time identifying uncertainties, missing controls, unsupported assumptions, unclear definitions, or weaknesses in project design. If this information remains inside a committee, much of the intellectual value of peer review is lost.

Providing feedback allows researchers to:

• correct genuine errors;
• strengthen experimental or mathematical methods;
• present assumptions more clearly;
• identify missing literature;
• redesign unrealistic work plans;
• and redirect projects toward more suitable funding channels.

Research on grant-review feedback has found that applicants do not always view feedback as fair or useful: in one study, only about 56–60% considered the feedback appropriate, depending on the assessment measure. This indicates that merely sending reviewer comments is insufficient. Feedback must also be relevant, well-informed, consistent, and clearly connected to the final decision.

Explanation Makes Bias Easier to Detect

Opaque rejection systems can conceal prestige bias, disciplinary conservatism, conflicts of interest, or unequal standards.

An unexplained rejection cannot easily be audited. A reasoned rejection can be examined for internal consistency:

• Was the same criterion applied to all applicants?
• Did the reviewer misunderstand the proposal?
• Was the criticism supported by the application?
• Did institutional reputation influence the assessment?
• Were unconventional methods rejected merely because they were unfamiliar?
• Did the final decision contradict the reviewer scores?
• Were irrelevant personal or institutional factors considered?

Transparency does not automatically eliminate bias. It makes bias more observable.

This distinction is important. No human or AI evaluation system should be assumed to be infallible. The objective is not to declare the evaluator perfectly neutral. The objective is to make its reasoning inspectable enough that mistakes and systematic distortions can be identified.

This principle also underlies AIIM’s approach to non-discriminatory scientific funding: evaluation should focus on outputs, dependencies, citations, use, replication, and other evidence of contribution rather than relying primarily on institutional status.

Explanation Protects Unconventional Research

Transformative research frequently appears uncertain before its value becomes obvious.

A committee may reject an unconventional proposal because:

• the field lacks established terminology;
• the expected application is difficult to predict;
• few reviewers understand the method;
• the work crosses disciplinary boundaries;
• the researcher lacks conventional credentials;
• or the contribution challenges an established framework.

An explanation creates a record of why the work was rejected. If the research later proves important, the scientific community can study which assumptions caused evaluators to overlook it.

Without a record, institutional failure disappears into silence.

This matters especially for foundational mathematics, theoretical research, open-source scientific software, negative results, replication work, and other contributions whose value may emerge through downstream dependencies rather than immediate commercial or publication impact.

AIIM is designed to address this problem through continuous and retroactive assessment. Instead of requiring a committee to predict a contribution’s future importance perfectly, the system can update its evaluation as evidence accumulates. A neglected theory, dataset, method, or software library can receive greater recognition when later work begins to depend on it.

Existing Funding Agencies Already Show That Explanation Is Possible

The argument for universal explanations is not purely theoretical. Major agencies already implement parts of it.

The NIH generally gives reviewed applicants access to scores and a summary statement containing reviewer feedback.

The NSF states that when a proposal is declined, the principal investigator receives unattributed reviews, an applicable panel summary, information about how the proposal was reviewed, and an explanation of the basis for the declination.

Some UKRI programs provide assessor comments supporting their scores, and UKRI describes feedback as a way to improve transparency and applicant learning. It also acknowledges the principal limitations: added workload and inconsistent feedback quality.

These examples demonstrate that explanations are administratively feasible. The remaining task is to make them universal, structured, decision-relevant, and auditable.

Why Reviewer Comments Alone Are Not Enough

A collection of anonymous comments is not necessarily an explanation of the final decision.

Reviewers may disagree. One may consider a proposal highly original, while another considers it insufficiently grounded. A panel may disregard one review. A program officer may apply portfolio considerations that were never mentioned by the reviewers.

The final explanation must therefore distinguish three layers:

1. Reviewer observations: what individual evaluators said.
2. Decision synthesis: which observations the agency accepted and how conflicts were resolved.
3. Allocation rationale: why the proposal was not funded after scientific, strategic, and budgetary considerations were combined.

Without this synthesis, applicants receive raw review material but not the actual reasoning of the institution.

How AIIM Already Provides Explainable Evaluation

AIIM’s model differs from a conventional grant competition in one fundamental respect: the assessment is not supposed to end with an unexplained yes-or-no judgment.

AIIM evaluates published scientific and open-source contributions and estimates their relative merit or contribution. The resulting assessment can include the signals on which the evaluation depends, such as:

• the content and claimed results of the work;
• citations and references;
• scientific or software dependencies;
• documented use by other projects;
• replications and independent reviews;
• originality and scope;
• downstream contributions;
• and evidence that later work relies on the evaluated output.

AIIM therefore does not merely say, “You were rejected.” It provides an evaluation that can explain why a contribution received its current level of recognition and funding.

World Science DAO describes AIIM as an algorithmic system with open records, transparent assessment, and auditable funding flows. Its current public model asks researchers to submit their research websites, has AI evaluate their contributions, and allocates available funds proportionally rather than dividing applicants into permanent categories of winners and losers.

This is an important structural difference.

Under a traditional grant system, an applicant may receive nothing because a proposal ranked just below an administrative cutoff. Under a proportional system, a smaller or less certain contribution can receive a smaller evaluation and potentially a smaller payment rather than being treated as having zero value.

In AIIM, a Low Score Is Not a Permanent Rejection

Scientific value changes as new evidence appears.

A paper that initially seems isolated may later become influential. A software library may acquire thousands of downstream users. A mathematical result may become necessary for a new theory. An experimental finding may be replicated. Conversely, a highly praised claim may weaken if replication fails or unsupported assumptions are discovered.

AIIM can update evaluations as these relationships become visible.

The explanation is therefore dynamic:

The contribution currently receives limited weight because there is little verified downstream use, insufficient independent confirmation, or uncertainty about its claimed novelty.

Later, the evaluation may change:

The contribution’s weight increased because several independent projects now depend on its method and two groups reproduced its central result.

This is more informative than a permanent rejection issued before the work’s real influence can be observed.

For a broader comparison, see AIIM versus conventional European grant funding, including the distinction between proposal-based prediction and continuous evidence-based allocation.

Explainability Does Not Mean Revealing Everything

Funding transparency still requires legitimate safeguards.

Agencies may need to protect:

• reviewer identities;
• confidential commercial information;
• personal data;
• unpublished technical details;
• security-sensitive research;
• and information that could expose reviewers to retaliation.

However, confidentiality should protect people and legitimate secrets—not conceal the logic of the decision.

An agency can provide a detailed explanation without naming reviewers. It can describe decisive criteria without publishing confidential proposal content. It can release aggregate scoring and decision logic while redacting sensitive information.

The correct principle is maximum decision transparency compatible with legitimate confidentiality.

AI Explanations Must Be Auditable

AI-generated explanations introduce their own risks.

A language model can produce a fluent explanation that sounds plausible without accurately reflecting the computation that determined the score. This is sometimes called a post-hoc rationalization: the system generates a persuasive narrative after the decision rather than revealing the actual basis of the decision.

AIIM should therefore connect every explanation to traceable evaluation evidence.

A robust explanation system should show:

• which criteria affected the result;
• which source materials were examined;
• which claims remain uncertain;
• which dependencies or citations were detected;
• how strongly each signal affected the assessment;
• whether multiple evaluators disagreed;
• and what evidence would materially change the score.

The explanation should be reproducible enough for auditors to test it. This is why AIIM’s commitment to public scoring, inspectable funding logic, and auditable records matters more than the mere use of artificial intelligence.

Appeals Should Correct Errors, Not Create Endless Bureaucracy

A requirement to explain decisions does not imply unlimited appeals.

Funding agencies reasonably worry that detailed feedback could generate disputes, increase administrative costs, or encourage applicants to litigate every unfavorable judgment. UKRI’s review of peer-review practices identifies both the transparency benefits and the workload risks associated with expanding feedback.

A balanced system can separate three processes:

• Clarification: The applicant asks what the explanation means.
• Correction: The applicant identifies a factual, procedural, or conflict-of-interest error.
• Scientific disagreement: The applicant disagrees with an evaluator’s legitimate judgment.

Only the second category necessarily requires formal reconsideration. An applicant should not receive funding merely because they disagree with a reviewer. But they should be able to correct a decision based on the wrong paper, a false factual premise, an undisclosed conflict, or a demonstrable procedural violation.

AIIM can apply the same distinction. Users should be able to contest incorrect inputs, missing dependencies, misattributed work, unsupported claims, or inconsistent rule application without turning every difference of scientific opinion into an automatic reversal.

A Practical Standard for Funding Agencies

Every rejection notice should answer the following questions:

What was decided? Which criteria controlled the decision? What were the decisive reasons? What evidence supports those reasons? Did budget or portfolio constraints affect the result? What could the applicant change? Can factual or procedural errors be challenged?

This standard is demanding enough to create accountability but narrow enough to implement at scale.

AI can assist by summarizing reviewer reports, identifying disagreements, checking whether the stated reason is supported by the record, and generating structured decision letters. Human decision-makers should then verify that the explanation accurately represents the actual decision.

The goal is not to automate politeness. It is to preserve the causal chain between evidence, evaluation, ranking, and funding.

From Rejection Letters to Continuous Scientific Assessment

The deeper problem is not simply that rejection letters are too vague. It is that conventional funding systems compress complex scientific judgments into a binary outcome.

A proposal receives a grant or receives nothing.

AIIM offers a different model:

• evaluation can be multidimensional;
• funding can be proportional;
• explanations can identify specific strengths and weaknesses;
• evidence can accumulate over time;
• decisions can be revised;
• and the records can remain auditable.

This transforms rejection from an institutional verdict into an updateable scientific assessment.

A contributor who receives little funding should be able to see why. A contributor whose work later becomes important should not remain trapped by an outdated judgment. Donors and public agencies should also be able to inspect why money moved toward one contribution rather than another.

Conclusion: No Public Funding Without Public Reasoning

Funding agencies distribute scarce resources, shape careers, and influence which scientific questions society investigates. Such power requires explanation.

Not every rejected proposal is wrongly rejected. Not every unconventional idea is a breakthrough. Not every reviewer disagreement indicates bias. But every applicant deserves to know the actual basis of the decision.

Explanations improve research, expose errors, discourage arbitrary gatekeeping, preserve institutional knowledge, and make funding systems easier to audit.

AIIM already moves toward this standard by treating evaluation as an inspectable, evidence-linked process rather than a secret committee verdict. Its continuous, proportional model also reduces the severity of binary rejection: a contribution can receive limited recognition now and more funding later as its verified usefulness becomes clearer.

The governing principle should be simple:

No scientific funding decision should be more opaque than the science it claims to evaluate.

Funding agencies demand that researchers explain their hypotheses, evidence, methods, limitations, and conclusions. Funding agencies should be held to the same standard.