Certain interactive features

from this article are only

available in the full digital issue.

Read Full Issue (subscribers)

Research integrity under pressure

Science is built on a foundation of trust. Researchers, institutions, publishers, and the public all rely on the assumption that published findings represent honest, rigorous, and transparent inquiry. Yet, that foundation is under increasing strain. Across disciplines and institutions worldwide, research integrity problems are growing in scale and complexity, raising questions about the reliability of the scientific record and the credibility of the enterprise itself.

The most egregious violations of research integrity are fabrications and falsifications, which typically involve inventing or manipulating data to support a desired conclusion. While such cases have historically been considered to be quite rare, their consequences can be severe. In medicine and public health, falsified findings have led to harmful treatments, misallocated resources, and lasting damage to public trust. Andrew Wakefield’s 1998 Lancet paper, falsely linking the MMR vaccine to autism, clearly illustrates this phenomenon. Despite the paper being fully retracted in 2010, it still influences vaccine skepticism decades later.

Deliberate fraud, however, is not the only threat to the integrity of the scientific record. Even research conducted in good faith can undermine confidence in science when its findings cannot be independently reproduced. The so-called “replication crisis” has also eroded confidence in the reliability of published science. A 2015 landmark effort by the Open Science Collaboration, for example, attempted to replicate 100 published psychology studies and found that fewer than half produced results consistent with the original findings. Since then, similar concerns have emerged in other fields, such as medicine or economics.

Averting a crisis

Tackling the research integrity crisis demands coordinated action on multiple fronts. Perhaps most urgent is research assessment reform. The prevailing “publish or perish” culture remains one of the most significant structural drivers of integrity problems, rewarding quantity over rigor and leaving the open science practices that could strengthen the record still far from the norm. Significant efforts are already underway, such as CoARA and DORA, that promote more responsible research assessment by recognizing the quality and transparency of research on its own merits. Hand in hand with the move towards responsible research assessment is the widespread adoption of open science practices.

Open science represents the strongest approach to protecting the integrity of the scientific record since it allows for greater scrutiny of the data, code, and methodologies underlying published findings. In this sense, transparency is more than a vague scientific principle; it is a very practical mechanism for creating a trustworthy scholarly communications ecosystem.

When research outputs are openly available, errors and inconsistencies are more likely to be identified and flagged by the broader community. Reproducibility becomes possible in a meaningful sense because the underlying materials are accessible rather than locked away on hard drives and available only upon request. Repositories provide the infrastructure that makes this openness durable by ensuring that datasets, code, preprints, and other research outputs remain accessible and traceable over time. This creates a persistent record that allows findings to be revisited, re-analyzed, challenged, or built upon beyond the lifespan of any individual project or institution. There are some cases of open data being exploited, but these can be best addressed through more careful examination of the papers and data rather than a retreat from openness. However, while open science practices are increasing, they are still not considered to be the norm for most researchers, and there is much work to be done to advance them in the future.

The consequences of an inadequate underlying infrastructure can be profound. A recent study by Gibson and colleagues examined two widely used clinical datasets hosted on Kaggle, a popular data-sharing platform, and found that they lacked verifiable provenance and showed strong signs of being simulated or fabricated. By the time the problem was identified, the datasets had already been used to develop 125 published clinical prediction models, three of which had been applied in clinical practice and one cited in a medical device patent. The fabrication was caught after its consequences had traveled deep into the clinical literature. 

The Gibson case is instructive for what it reveals: while the datasets were openly available, there was no information about when, where, or how the data had been collected, nor was there any way to trace them to a source that could be verified. This is precisely the kind of structured provenance information that repositories are designed to capture and preserve. 

As distributed, institutionally governed infrastructures, repositories hold research outputs over time and expose the metadata that makes them traceable.

Repositories lie at the heart of open science

At the foundation of research integrity lies openness, persistence of resources, and provenance information that describes where an object came from, how it was created, and how it has changed over time. Incomplete provenance information makes it more difficult to verify whether a resource is legitimate, but provenance can be strengthened when related resources are directly linked to each other – for example, by linking publications to datasets, code, or funding sources – enabling others to verify, interpret, and reuse materials with greater confidence. Structured, detailed metadata, together with persistent access to resources, helps distinguish legitimate scholarship from manufactured or manipulated work. Persistent identifiers help ensure that citations remain accessible and can be tracked even when platforms evolve, close, or change hands. New “live” publications, such as those published in the NeuroLibre repository, a next-generation preprint platform, show how these connections can work in practice by allowing readers to review data and run underlying code while reading the manuscript.

Long-lived and well-functioning repositories are a prerequisite for building such an ecosystem. They collect and preserve the diverse outputs created through research and implement the practices required for verification. With thousands of repositories across the world containing millions of research objects, they are becoming increasingly important in safeguarding the scholarly record. An internal analysis of half the records in COAR’s International Repository Directory, for example, found close to 340 million items spanning articles, datasets, code, and reports, representing a significant portion of the research record across the world. 

The international repository network is key to building a trustworthy and verifiable information commons. The repository network is a highly distributed, federated architecture built on thousands of individual institutional, disciplinary, and national repositories, connected through shared technical standards. While there are several types of repositories (including institutional, domain, and generalist repositories), collectively they steward and provide access to millions of research outputs produced by researchers around the world. These repositories, usually managed by a university or research-performing organization, are not contingent on their commercial viability or external provider (unlike some other scholarly infrastructures), a consideration that grows more relevant as open science governance navigates an increasingly complex geopolitical landscape. They are also often nodes in national and regional networks (or aggregators such as LA Referencia in Latin America, OpenAIRE in Europe) that define the common standards required to ensure repositories provide visibility for their resources with scholarly indexes and search services. At the international level, organizations like COAR provide global coordination, develop and promote standards, and advocate for the adoption of good practices across the repository community. 

Repositories provide the practices that strengthen research integrity

As research integrity becomes a more pressing issue due to the ever-increasing volume of research outputs and the rise in paper mills and AI-generated content, COAR recognizes the urgent need to strengthen how repositories support a verifiable scholarly record in four areas: (1) richer provenance, (2) clear versioning, (3) reliable signals about retractions and corrections, and (4) connections between research outputs and open peer review. Explore each of these areas using the navigation arrows below.

Research integrity of science is not a peripheral concern. It is the condition upon which everything else depends: the translation of research into policy, the development of treatments that reach patients, and the public trust that sustains the entire enterprise. The challenges outlined in this piece, from fabrication and falsification to AI-generated papers and irreproducible methods, are not isolated failures of individual researchers acting in bad faith. They are, in large part, the predictable outputs of a system that has rewarded the wrong things for too long.

Repositories, properly resourced and interconnected, offer a persistent, institutionally governed layer of accountability beneath the scholarly record. They hold the provenance that makes verification possible, the versions that make the evolution of research traceable, the correction signals that keep the record current, and connect research outputs with open reviews and evaluations that make their assessment visible and traceable – all sustained by preservation practices that keep these records accessible over time. Each repository performs these functions for its own community, but interconnected through shared standards and coordinated practice, they form the distributed foundation that makes the scholarly record verifiable. 

The work COAR is undertaking – recommended practices for versioning, a working group on retractions and corrections, the continued growth of COAR Notify and the PRC model – is not peripheral to research integrity. It is the practical, infrastructural core of it. None of this replaces the harder work of reforming research assessment or shifting incentives away from "publish or perish." But it is equally necessary: it ensures that when the culture does shift, the scholarly record will be there to support it: verifiable, traceable, and built to last.

VISIT COAR WEBSITE

Disclosures: Kathleen Shearer is the Executive Director of the Confederation of Open Access Repositories (COAR).