Research Integrity Can't Stay a Volunteer Backlog

A research-integrity problem does not become operationally serious only when a paper is retracted. It becomes serious much earlier, when the allegation arrives and nobody is quite sure who owns the next step. An editor forwards a concern. A volunteer committee meets when calendars allow. A staff member starts a spreadsheet. A reviewer remembers a similar case from last year. The journal has a policy, but the case itself moves through inboxes, memory, and goodwill.

That model is under strain. Retraction Watch reported on July 2, 2026 that the Association for Computing Machinery is hiring a director of research integrity after a rapid rise in misconduct allegations created a backlog of cases: https://retractionwatch.com/2026/07/02/association-computing-machinery-society-acm-research-integrity-role/. The details are ACM-specific, but the lesson is not. Research integrity is becoming a staffed publishing function, not a side responsibility that can be absorbed indefinitely by editors and volunteers.

Journal leaders should treat that as a practical warning. The hard part of integrity work is no longer only writing the policy. It is building the operating system around the policy: intake, triage, evidence handling, author communication, reviewer confidentiality, institutional coordination, editorial independence, correction metadata, and a record of why each decision was made.

The Policy Is Not The Workflow

ACM already has detailed publication policies. Its policy page says credible allegations may involve plagiarism, author misrepresentation, content falsification, misconduct related to peer review, confidentiality breaches, coercion and abuse, copyright issues, and human-subjects policy breaches: https://www.acm.org/publications/policies. Its retraction policy lists grounds for retraction, explains who can make decisions, requires supporting evidence for allegations, and says retraction metadata should appear through CrossMark or similar systems: https://www.acm.org/publications/policies/retraction-policy.

That is the point. Even mature policy language does not remove the case-management burden. Someone has to decide whether a complaint is credible enough to open. Someone has to preserve files, compare versions, check reviewer history, examine images or code, ask for author responses, coordinate with an institution, decide whether an expression of concern is needed, draft notices, and update metadata. None of that happens simply because a policy exists on the website.

The operational failure mode is usually quiet. A journal does not announce that a case is aging in a shared inbox. It does not tell readers that an allegation is waiting for a specialist who has not been appointed. It does not show authors that three similar cases are being handled differently across titles. Yet those delays and inconsistencies shape the integrity of the record as surely as the final notice does.

What A Dedicated Integrity Function Actually Owns

A named research-integrity lead is not a magic fix. It is useful only if the role owns the parts of the workflow that otherwise fall between editorial, legal, production, platform, and society governance. At minimum, five queues need clear ownership.

• The intake queue: allegations, screening flags, PubPeer posts, reader emails, reviewer concerns, author disputes, and institutional notifications should enter one trackable route rather than several private inboxes.
• The evidence queue: manuscripts, reviews, author declarations, image checks, similarity reports, correspondence, data links, contributor records, and prior decisions need controlled storage and a chain of custody.
• The decision queue: editors, publishers, committees, and counsel need defined roles so advice, authority, conflicts, and escalation paths are not improvised case by case.
• The publication queue: corrections, retractions, expressions of concern, withdrawals, watermarks, article-page status, DOI metadata, Crossmark updates, and index notifications need to move together.
• The learning queue: repeated signals should feed policy changes, reviewer governance, submission checks, special-issue oversight, and staff training, not disappear once one case closes.

Without those queues, the journal may still make defensible decisions, but it will struggle to prove that the process was timely, consistent, and complete. That matters when an author appeals, an institution asks for records, a journalist reports on a backlog, or a reader wants to know whether the published literature was corrected promptly.

Batch Problems Need Batch Infrastructure

Post-publication integrity work increasingly arrives in clusters. Paper mills, compromised peer review, tortured phrases, manipulated images, recycled datasets, guest-edited special issue abuse, and undisclosed third-party involvement rarely affect only one manuscript. One case can turn into fifty. One reviewer account can implicate several journals. One paper-mill pattern can cross publishers.

COPE updated its retraction guidance in 2025 with more explicit attention to large-scale systematic manipulation, including paper mills and batch retractions: https://publicationethics.org/guidance/guideline/retraction-guidelines. That guidance is not just an ethics document. It is an operational demand. If related articles need to be investigated together, the journal needs a way to group cases, compare evidence, notify authors, preserve editorial independence, and publish consistent notices without delaying every article until the slowest file is resolved.

This is where small journals are especially exposed. A society journal may have excellent editors and a strong publication ethics committee, yet no shared case ledger, no standard evidence package, no status dashboard, and no way to see whether similar allegations are appearing across titles. The result is not negligence. It is overload.

Tools Help Only When Governance Comes First

The industry is building more shared infrastructure for this problem. STM describes the Integrity Hub as a cloud-based environment for publishers to check submissions for research-integrity concerns, including patterns associated with paper mills, while respecting data privacy and competition-law constraints: https://stm-assoc.org/what-we-do/strategic-areas/research-integrity/integrity-hub/. Its public case-study page describes publishers using shared tools to scale investigations and strengthen editorial workflows: https://stm-assoc.org/what-we-do/strategic-areas/research-integrity/integrity-hub/the-stm-integrity-hub-in-action/.

That direction is useful, but it should not tempt journals into outsourcing judgment to a flag. A screening tool can identify a risk pattern. It cannot decide whether an allegation is sufficiently supported, whether an author explanation is credible, whether a reviewer conflict tainted the decision, whether an expression of concern should be issued during a long investigation, or how a notice should be worded. Those are governance decisions.

The healthy model is tool-assisted and human-accountable. Screening, similarity checks, image tools, reviewer-identity checks, and metadata comparisons should feed a governed workflow with named owners, timestamps, permissions, and decision records. If a tool creates a signal but the journal cannot show who reviewed it, what evidence was considered, and why the decision followed, the tool has added noise rather than integrity.

Where Platform Governance Matters

For publishers managing several journals, the integrity workload is also a platform problem. The workflow has to connect submission checks, reviewer governance, role-scoped access, audit trails, correction metadata, production outputs, and portfolio reporting. The point is not to replace editorial judgment. It is to make sure the evidence, ownership, and downstream record do not fragment across inboxes and spreadsheets.

That matters most when a concern crosses boundaries. A suspicious reviewer pattern may touch multiple titles. A correction may need article-page changes, Crossmark metadata, and index notifications. A misconduct allegation may require restricted access for legal or institutional correspondence while still giving editors the context they need. Platform governance is what keeps those actions connected.

A 30-Day Integrity Operations Audit

Journal leaders do not need to hire a full department before improving the workflow. Start by mapping the current path of one allegation from arrival to closure. Use a real closed case if possible; if not, run a tabletop exercise with a plausible scenario such as duplicated images, compromised peer review, undisclosed AI-generated content, authorship sale, or a data-availability challenge.

• Name the first recipient and the backup recipient for each type of concern.
• Define what information is required before a case is opened and what can be requested later.
• Create a standard evidence folder with access rules, timestamps, and version retention.
• Set target review points: initial triage, author contact, institutional contact, interim notice decision, final decision, and metadata update.
• List who can approve corrections, expressions of concern, retractions, removals, and appeals.
• Check whether the final public action updates the article page, PDF, XML or JATS, DOI metadata, Crossmark status, feeds, and index notifications.

The audit should produce one uncomfortable metric: how many days a credible concern can sit before a named person is accountable for the next action. If the answer is unclear, the journal has a staffing and workflow risk even if its ethics policy is strong.

The Practical Takeaway

The practical takeaway for journal leaders is simple: research integrity work needs an operating model. Policies, committees, and expert volunteers remain essential, but they cannot be the whole system when allegations are rising, evidence is technical, and corrections must propagate through metadata infrastructure.

Name the owner. Build the case ledger. Define the evidence package. Time-box the decision points. Connect the public notice to the metadata update. Then review the pattern of cases every quarter. A journal that can do those things will handle difficult allegations with more consistency, less delay, and a stronger record when its decisions are challenged.