The Credit Problem: Why Authorship Lists Are a Poor Record of Who Did the Work

The authorship list is the oldest surviving instrument in the research credit system, and it was designed for a medium with a hard physical constraint: a printed page had room for a heading, and the heading had room for names. Everything a scientific project consists of — years of method development, argument, repair, negotiation and failure — is compressed into an ordered sequence of surnames. The compression is lossy, and not randomly so: it discards particular kinds of contribution, belonging to particular kinds of people.

This would be a formatting complaint if the byline were merely descriptive. It is not. It is the primary input to almost every consequential decision made about a researcher: appointment, tenure, promotion, fellowship, grant renewal. And because publication cycles run twelve to thirty-six months, it reports on work that concluded years earlier. A researcher’s standing is assembled from a delayed, compressed summary of a fraction of what they have done.

What the byline was built to do

The authorship list does one job well: it assigns responsibility for a published claim. If a result is challenged, it identifies who stands behind it. Its conventions — first author, last author, corresponding author — encode a rough sense of who drove the work and who supervised it.

What it was never asked to do is inventory labour. And it carries a restriction that no ordering convention can soften: to appear on the list at all, the project must have produced a paper. Credit sits downstream of the publication decision and inherits every bias in it — towards positive results, towards novelty, towards work that concluded rather than work that ruled something out.

The contributions the list cannot see

Consider what reliably falls outside it. These are generic cases, and every laboratory director will recognise all of them.

• The technician who built and validated the assay everything downstream depends on, and who appears in the acknowledgements if at all, because infrastructure is not a finding.
• The early-career researcher who spent two years running a competent, well-designed trial that returned a negative result. It was never written up, and the contribution leaves no trace of any kind.
• The external specialist brought in to solve a blocking problem — a statistical pathology, a synthesis route, an instrument fault — who solves it in a fortnight for a fee and no formal credit anywhere.
• The wet laboratory providing physical validation for a computational group’s prediction. Without it the prediction is a hypothesis; with it, the paper exists. Credit is split by negotiation between unequal positions.
• Everyone attached to a programme discontinued after a portfolio review, whose work was neither wrong nor finished.

Some 8.8 million researchers work across more than 21,000 institutions, largely in isolation, and this is the shared record of what any of them has contributed. The pattern in what it omits is consistent: it under-records the technical work that makes results possible, the negative findings that spare others a dead end, and any contribution by someone without the standing to insist on a place in the list.

An instrument that registers only work which ended in a publication does not measure contribution. It measures the subset of contribution that survived a publication decision, and presents that subset as the whole.

System	What it records
ORCID	A persistent identifier for the researcher
CRediT	The role a named author played on a published work
Journals	Output that survived a publication decision
Contribution ledger	Every project event, attributed as it happens — including work that was never published

What ORCID and CRediT already solve

Two systems have made real progress here, and an honest account must credit them. ORCID provides a persistent identifier that distinguishes a researcher from everyone with a similar name and follows them across institutions and countries. Its most important contribution is conceptual: it established the principle that a researcher’s identity belongs to the researcher, not to whichever organisation currently employs them.

The CRediT taxonomy addresses the compression problem directly. Instead of an ordered list of names, it describes contributions through fourteen defined roles — conceptualisation, methodology, investigation, validation, software, supervision, writing and others — so a paper can state what each named person did. That is a substantial improvement on position in a queue.

Both share a boundary, which is a matter of scope rather than a flaw. Both attach to outputs, and in practice the output is a publication. Role assignments are asserted by the authors themselves at submission and are not independently verifiable. Neither addresses the years between the work and the paper, projects that produced no output, or people who never made the author list. Neither was built to. The territory is simply uncovered.

A contribution-level record, in practice

The alternative is not a better list but a different unit of record: the individual contribution rather than the finished paper, logged when it happens rather than when it is published. On DimenChain, every project creation, commit, milestone and governance vote is permanently recorded on-chain and attributed to the participant who made it — a signed, timestamped transaction, publicly auditable on Arbiscan without asking the host institution to confirm anything. New collaborators are admitted by vote of existing stakeholders through weighted smart-contract governance tied to equity stake, so joining a project is itself a dated, attributed event, as is the decision to admit you. Open collaborator roles can carry defined equity shares, for example 15%, locked in a smart contract with royalties routed automatically. Credit and economics become one record: the wet laboratory supplying validation co-owns the resulting intellectual property rather than being thanked in a footnote.

The granularity changes what can be said about a career. Instead of “fourth author of eleven”, the record supports a specific claim: on this date, this person committed this milestone to this project, it was validated by these parties, and these stakeholders admitted them. Confidentiality holds by construction — raw data stays off-chain, and only encrypted metadata, hashed proofs and contract events reach the ledger.

What a Soulbound token is actually for

Contributions are recognised with Soulbound Tokens, and the term deserves an unglamorous explanation. A Soulbound token is a credential that cannot be transferred: not sold, bought, lent or delegated. That single restriction is the design intent, because a credential that can be traded creates a market, and a market in scientific reputation is precisely the outcome to avoid.

What it provides is portability. The record attaches to the researcher rather than to an institution’s internal systems, so it survives leaving, a laboratory closing, a supervisor retiring, a grant ending. And it is verifiable by a third party without a reference letter and without the cooperation of a former employer — which matters most where that cooperation is least likely.

The objections that deserve answers

Goodhart’s law

When a measure becomes a target, it ceases to be a good measure. Research has decades of evidence: the impact factor, the h-index and raw publication counts have all been optimised against, sometimes to the point of absurdity. Any new record of contribution will be gamed too, and that should be the starting assumption rather than a scenario to be argued away.

Two things reduce the return on gaming, and neither is a cure. First, the ledger records events; it computes no score and produces no ranking. Anyone can build a league table on top of any dataset — but the entries stay specific and inspectable, so a reader who wants to know what a milestone consisted of can look rather than trust a summary. Second, manufacturing entries is not free: admission runs through a vote by stakeholders whose own equity is diluted by admitting a passenger, and the vote is attributable. That makes collusion expensive and traceable. It does not make it impossible, and any claim that it does should be treated sceptically.

Reducing scholarship to something countable

The deeper objection is that the qualities which make someone a good scientist are largely not events. Judgement, taste in problems, generosity, the willingness to tell a doctoral student their idea is wrong and then help them find a better one — none of that produces a transaction. Any structured record privileges the legible over the important.

That is correct, which is why such a record belongs in evidence supporting a judgement, not in place of one. The failure mode is not the record; it is a committee that stops reading because a number is available. The same objection applies with more force to the instrument in use today: a ranked list of surnames is also a compression of scholarship into something countable, and a considerably coarser one.

Permanence and the right to erasure

This objection cannot be fully answered, and should be stated plainly. European data protection law gives individuals a right to erasure. A ledger designed so that no party can amend it cannot honour a deletion request in the ordinary sense. The two are in genuine tension, and anyone claiming the tension has been dissolved is overselling.

What can be done is to keep the tension away from the material that matters. Personal and raw data stay off-chain, where ordinary deletion applies. What reaches the ledger is hashes, encrypted metadata and contract events attached to pseudonymous addresses, so erasure operates on the layer holding the identifying information and leaves entries that no longer resolve to an identified person. This is a mitigation, not an exemption: whether a pseudonymous identifier counts as personal data depends on what else is linkable to it, a question for a data protection officer at the start of a deployment.

Permanence also has a side that is easy to overlook when careers are the subject. Most real disputes about credit are not about whether a record should be deleted. They are about whether it can be quietly changed by whoever controls the server, after a falling-out, a departure or a change of principal investigator. A record no institution can revise retrospectively protects the least powerful person in the room, and that is usually the person whose contribution is being edited.

A complement, not a replacement

None of this evaluates science. A timestamped, attributed milestone establishes that something happened, when, and who did it — not whether it was any good. Treating cryptographic attribution as a substitute for expert judgement would be a serious misreading. That remains the work of peer review.

• ORCID — who this researcher is, persistently and independently of their employer.
• CRediT — what role each named person played in a published output.
• Peer review — whether the claim is warranted by the evidence.
• A contribution ledger — what happened, when and by whom, including the years before publication and the work that never published at all.

Where this stands today

DimenChain operates this layer as live infrastructure for research institutions, laboratories and R&D groups. Contributions, milestones and governance decisions are recorded on Arbitrum, an Ethereum Layer-2 network whose transaction costs keep per-record attribution within ordinary academic budgets, and the activity is publicly auditable on Arbiscan. The protocol is patent-protected under WIPO PCT/IB2024/058791. It operates from the Netherlands, where 31 globally ranked universities, the Leiden Bio Science Park cluster, the European Medicines Agency and a research culture committed to FAIR data principles make the case for machine-readable provenance an unusually easy one to put.

The authorship list is not going to disappear, and it should not. It does its narrow job — assigning responsibility for a published claim — better than any proposed replacement. The argument is only that it should stop being the sole record, because a great deal of scientific work is done by people it was never built to see.