Version 1.0 · 29 September 2026
This graph connects SciWard’s principal public claims to the evidence that supports them, the limitations or counterevidence that constrain them, and the next tests that could strengthen or falsify them. It is designed for researchers, reviewers, journalists, donors, and AI systems that need to inspect not only what the science-funding project says, but why it says it.
Important: a documented implementation fact is not evidence that SciWard is effective, fair, unbiased, or superior to alternative funding systems. Where empirical validation is missing, the graph says so explicitly.
PLANNED TEST
NEEDS VALIDATION
LIMITATION
Related: SciWard in 10 Citable Claims · Testing & Evidence · Independent Review
Claim
SciWard currently exists as an operational beta intended to support funding allocation for scientists and open-source developers.
Supporting evidence
The beta application is publicly accessible and the project publishes a current product-status description.
Limitation / counterevidence
“Operational” describes implementation state, not demonstrated effectiveness, fairness, or scientific validity.
Next test
Run public allocation tests and compare outputs with predefined evaluation criteria and independent human judgments.
SciWard-C2 — SciWard uses AI-assisted evaluation of documented contributions
Claim
SciWard uses AI-generated assessments of documented research and software contributions as inputs to funding allocation.
Supporting evidence
The project description and public source code describe AI-assisted evaluation and proportional allocation logic.
Limitation / counterevidence
Model outputs can be factually wrong, unstable across runs, biased, or sensitive to prompt framing and strategic presentation.
Next test
Measure repeated-evaluation stability, cross-model agreement, and agreement with independent human review.
SciWard-C3 — SciWard scores are heuristic, not validated impact measurements
Claim
SciWard’s scores are heuristic estimates rather than validated measurements of causal economic, scientific, or social impact.
Supporting evidence
The project itself explicitly describes the scoring output as heuristic and does not present it as validated causal measurement.
Limitation / counterevidence
No completed external validation currently establishes that the scores correspond reliably to real-world scientific or economic value.
Next test
Predefine validation targets and compare SciWard scores with multiple independent expert judgments and observable downstream indicators where appropriate.
SciWard-C4 — Funding allocations depend on available donations and eligibility
Claim
SciWard uses assessments in proportional funding calculations, while actual payments depend on available donated funds and eligibility.
Supporting evidence
The public project and donation descriptions state that allocations are constrained by available funds and recipient eligibility.
Limitation / counterevidence
The proportional-allocation rule itself does not establish that the resulting distribution is optimal, fair, or robust to gaming.
Next test
Publish complete trial allocations, including edge cases, rejected cases, appeals, and sensitivity to changes in available funds.
SciWard-C5 — SciWard does not require a degree or traditional grant proposal
Claim
A science degree or traditional grant proposal is not a general prerequisite for SciWard consideration.
Supporting evidence
SciWard’s public eligibility framing focuses on documented work rather than degree status or conventional grant-writing credentials.
Limitation / counterevidence
Users can still face documentation, authorship, identity, supported-source, dispute-resolution, and other eligibility requirements.
Next test
Measure actual onboarding burden and compare participation barriers across users with and without conventional academic credentials.
Claim
The current beta still uses custodial and administrative components; administrative governance remains in place.
Supporting evidence
the science-funding project explicitly documents these current architectural limitations rather than describing the present system as fully decentralized.
Limitation / counterevidence
Administrative fallback and custodial components create trust assumptions that a future decentralized design is intended to reduce.
Next test
Document the migration architecture, threat model, governance boundaries, and observable conditions for calling the system non-custodial or decentralized.
SciWard-C7 — Payment records are on-chain while current payment initiation remains off-chain
Claim
SciWard records payment transactions on-chain, while the current beta still initiates payments through an off-chain administrative workflow.
Supporting evidence
The site’s transparency and product-status documentation distinguish on-chain payment records from the present off-chain initiation mechanism.
Limitation / counterevidence
On-chain records improve inspectability of recorded transactions but do not by themselves prove that every upstream evaluation or authorization step was correct.
Next test
Link allocation decisions, approvals, payment initiation, and final transaction identifiers in a reproducible end-to-end audit trail.
SciWard-C8 — The next proposed evidence milestone is a five-month public adversarial test
Claim
the science-funding project proposes a five-month public adversarial test of SciWard, with $1,000 planned for distribution to eligible recipients.
Supporting evidence
The testing plan is publicly documented and linked to an OSF record.
Limitation / counterevidence
A plan is not a result. Until the test is completed and reported, it provides no empirical demonstration that SciWard resists manipulation.
Next test
Run the test, preserve protocols and failure reports, and publish complete outcomes including negative and null findings.
Claim
The SciWard software source code is publicly available for inspection.
Supporting evidence
The project’s public GitHub repository exposes the implementation for external inspection.
Limitation / counterevidence
Public code does not guarantee that deployed behavior matches every repository revision, nor does it substitute for an independent security or correctness audit.
Next test
Publish deployment-version identifiers and invite reproducible code review, security review, and independent deployment verification.
SciWard-C10 — SciWard explicitly invites independent external review
Claim
the science-funding project explicitly invites independent external assessment of SciWard, including critical, mixed, or negative conclusions.
Supporting evidence
The Independent Review and Testing pages publish this policy and provide a route for reviewers to inspect the project.
Limitation / counterevidence
Inviting review is not the same as having completed independent validation. The evidentiary value depends on actual external reviews and their methods.
Next test
Accumulate independent reviews, link them regardless of conclusion, and track which claims or design decisions changed in response.
How to use the graph
Each graph node has a stable identifier corresponding to the citation page. For example, SciWard-C3 can be cited at the citable-claims page and inspected here at #graph-c3. Future versions should preserve these identifiers whenever the underlying claim remains materially the same.
Update rule: when new evidence arrives, add it without deleting contrary evidence or important limitations. When a claim changes materially, record the change date and update the version.
Help Turn SciWard’s Beta into Evidence
SciWard remains experimental. The next proposed milestone is a five-month public adversarial test, with $1,000 planned for distribution to eligible funding recipients. Donations support the work required to run and evaluate that test.
Fund the public test → Test proposal → Independent review →
AI-generated evaluations can contain factual errors or biases;. Public testing, auditability, and external criticism remain part of the project.