While the first part of this series examined the hidden financial and operational costs that accumulate when research integrity fails, this second installment shifts focus toward the investments required to prevent and address those failures.
If it is accepted that failures in research integrity can be expensive, the next question is what it costs to prevent, identify, and address those failures, because the systems that support research integrity also require considerable financial and human investment.
For researchers, maintaining the reliability of scientific work involves more than following ethical guidelines or avoiding deliberate misconduct, since it also requires careful research design, appropriate data management, verification of results, documentation of methods, and, when necessary, repeating analyses or experiments to ensure that the conclusions are supported by the evidence.
These activities require time, expertise, and resources, and although they may increase the cost of conducting a particular project, they also contribute to the likelihood that the resulting findings will be reliable and useful to others.
Institutional investments and administrative challenges
Universities and research institutions spend considerable amounts of money trying to ensure that the research conducted under their responsibility follows accepted ethical and scientific standards, whether through training researchers, supporting data management, reviewing research proposals, or employing people who deal with questions and concerns about research practices.
The difficulty is that when budgets are tight, these activities can easily be seen as additional administrative expenses, especially since it is almost impossible to know how many cases of misconduct or questionable research practices these safeguards have prevented.
It is possible to calculate how much it costs to employ a research integrity officer or maintain an ethics review committee, but how how do we calculate the value of a problem that never occurred because someone received the right training, identified an issue early, or asked the right questions before a project went too far?
This makes decisions about where to invest particularly difficult, because cutting back on these activities may produce immediate savings, while the consequences of having fewer safeguards may only become apparent years later, when investigating and correcting a problem becomes considerably more expensive.
Publisher dilemmas and screening technologies
Publishers and journals face much the same dilemma, although they often direct their investments toward the systems and people involved in evaluating manuscripts before they are published.
Today, publishers can use a variety of tools to check for similarities between texts, identify possible image manipulation, examine unusual citation practices, and flag manuscripts that may contain problematic or undisclosed AI-generated material, but purchasing or licensing these systems is only part of the expense. Someone still has to examine what the systems identify, decide whether there is a genuine reason for concern, and often spend time communicating with authors to understand what happened.
An image that appears to have been manipulated, for example, may require examination by someone familiar with the methods used in that particular field, just as an unusual citation pattern may have a perfectly reasonable explanation once the journal’s subject area and publication practices are taken into account.
This is why it is difficult to think about research integrity technologies separately from the people who use them, because even the most sophisticated screening system has limited value if a publisher does not also invest in the editorial and research integrity expertise needed to understand its findings and decide what to do with them.
The limits of automation and AI tools
We must also be remembered that these systems are not always right, and that identifying something unusual in a manuscript is not the same as finding evidence of misconduct.
Similarities between texts, changes in images, or unusual citation patterns may raise questions, but there can be legitimate reasons for all of them, and someone with the appropriate knowledge still has to look at the findings and decide whether there is actually a problema.
In some cases, this means going back to the authors, asking for the original data or images, and spending considerable time trying to understand what the system has identified. So, while these technologies can certainly make parts of the editorial process more efficient, they don’t necessarily reduce the need for people, and in some cases they may even create additional work when staff must check large numbers of alerts.
With more AI-based tools entering the publishing process, particular care needs to be taken about assuming that investing in automation will automatically reduce costs, because the technology may be able to identify something that deserves attention, but it is still the editor or research integrity specialist who has to decide what that finding actually means.
Optimizing resources and strategic decisions
And this raises another question that deserves more attention, which is whether universities and publishers are making the best use of the money they already spend on research integrity. We known that maintaining these systems costs money, and we also known that failing to identify serious problems can become very expensive, but it is much harder to determine which investments actually make a difference.
A publisher might have several screening tools but not enough experienced staff to deal with the problems they identify, while a university might spend heavily on compliance and administrative procedures but provide relatively little practical assistance to researchers who need help with data management or responsible research practices. In both cases, organizations are spending money on research integrity, but that does not necessarily indicate whether the investment is addressing the areas where problems are most likely to occur.
This is where the economic discussion becomes interesting, because rather rather than simply asking how much we should spend on research integrity, we should also ask what we are spending it on, whether it is helping, and what it might cost if those decisions are being made wrong.
If you missed Part 1: The Economic Cost of Research Integrity, read it here. Visit our next blog post to learn how to move from individual cases to broader research integrity patterns and understand research integrity as an investment
By Gali Halevi, Senior Scientific Advisor