Is fraud academic? Honestly!

 

I was listening to a podcast about fraud in academia which resonated with me. I left academia behind, not because of any fraud that I had seen (in the strict sense; there were some I knew of, but did not know, whom I might describe as ‘old frauds’ in the sense that their pet theories and work supporting them seemed unworthy of their reputation; not because of dishonesty but because the results reported seemed insufficient), but partly because of the incentives I saw for dishonesty.

              As some of the interviewees noted, there is massive incentive to commit fraud in research: a relentless focus on interesting, positive results, little penalty for being caught out and a career environment heavily focussed on status and reputation as judged by publications and excitement generated.

              The presenter, Stephen Dubner, admitted that he had thought of academia as the last bastion of rational respectability: an outpost of dedication to truth. He was disappointed to be presenting the various findings about the sector and the few well-publicised stories of confirmed academic fraud.

              I think that he was right. Academia is indeed an outpost of dedication to truth. The fact that researchers and statisticians looking into academic fraud think that 2% is a lower bound for academic fraud is remarkable not because it is so high, but because it is low. Academia is a career one undertakes at least partly from love of research and a commitment to seeking truths. The incentives to commit fraud take gradual effect, as documented by a couple of confirmed fraudsters and the people looking into academic fraud. They explained that, as with so much wrongdoing in all walks of life, it starts off minor; pushing things here and there, picking the ‘best’ of three rather than taking whatever result happens, filling in a blank when one test in a series failed, trying to meet targets to get good work done, avoid blame for another experimental failure, look good in front of lab leaders and colleagues and not let people down, and ends up in falsifying whole data sets.

              The incentives to commit academic fraud are simple to understand. It’s a high-pressure world, in which only interesting results get published and publications in more respected journals determine the fate of one’s career. Negative results (‘we looked into this and found no effect’) are not valued. Academic review is a theoretical concept, but promotes sticking to established theories, and rarely looks in detail at all the experimental detail. Not that review of a manuscript could detect all forms of fraud: if I say I did something and make up the results well, it’s impossible to know I’m lying just from my own report about it. Reviewing papers is unpaid and takes a lot of time which must be on top of a day job. No-one is paid to review.

              Respect and career advancement come from lying; there are no established processes to detect lying and if someone is caught there are often (but not always) minimal penalties, as the employer wants to avoid the reputational damage; it might not be clear whether it’s truly a lie, or incompetence or error; and one instance might not seem so awful. This means that anyone doing completely honest work is at a major disadvantage.

              When I describe it in broad generalities like that, it should be obvious that academia is not special. Our entire economy works off people writing CVs with minimal references, or else references written by helpful friends who got along well with the candidate. There are established jokes about seeing through phrases used in academic papers and CVs: highfalutin language disguises a very mundane reality.

              The difference between academic and more generic careers is that academia involves a certain love of truth and desire to discover facts. Other jobs are far more simply getting money for one’s time. So we should expect the estimate of academic fraud to be lower than fraud in the wider world of jobs.

              Fraud in the rest of the economy is estimated, but these estimates refer to crimes such as diverting funds to the incorrect bank account. Fraud is 38% of all reported crime, and it is estimated that only one in seven cases are reported. Reported frauds cost individuals £1.2bn; almost 3,000,000 cases of fraud were reported in 2023.

              Another estimate puts total losses in the economy far higher than the reported figures: £219bn. This report admits that its estimate is only that, but notes that because businesses prefer not to publicise being the victims of fraud it is hard to get solid data to improve the estimate. The vast majority of this amount is an estimate of procurement fraud, at about 4.5% of expenditure.

              The detail of all this is interesting, but not the point here. 4.5% is indeed higher than the 2% given as a lower bound in academia. Yes, these are different things: academic fraud is about one’s own achievements and results, but they do blend into each other. For example, included in procurement fraud is simply under-provision of goods or services: i.e. making greater claims or promises than actually delivered. Which brings me onto a familiar subject.

              Is it likely that people are willing to defraud others in a business context, but draw the line at doing so on their own behalf? Only if there are significant penalties that they do not mind their employer experiencing but object to suffering themselves. It seems likely, given the rewards on offer (better jobs giving more pay and more experience for future jobs) that the fraud rate in business can be extrapolated directly as a lower bound to the fraud rate in employment.

              ≈5% might seem like a small number. Does that not mean that 19 out of 20 people are honest? Firstly, I think this can safely be taken as a lower bound, given the incentives and penalties. Secondly, how many applicants are there for most jobs? More than 20? If you take the most impressive 3 CVs to interview, it’s incredibly likely that you get at least one dishonest one.

              The first response might be to say “That’s what the interview is for”, but let’s stop for a moment. Are interviews really designed to detect fraud? Many people fancy that they are insightful, but are they insightful enough to detect all lies? Are interviews really carefully thought-through to provide the best chance of detecting falsehoods, or are they an onerous burden on rushed managers who are already too busy (or else they wouldn’t be recruiting staff to help!)? And even if interviewers attempt to detect fraud, rather than to establish rapport and use liking as a (truly terrible) proxy for trustworthiness, is it possible to do so with a few questions? I aver that accurately assessing skill, ability and potential in interview (or other forms of recruitment) is itself a skill that needs to be learned and practiced; and yet most people have not studied it nor had anything more than perhaps one brief session from HR (who themselves might not know much) to prepare them for it. Interviews, like academics reviewing papers, are an extra piece of work one must do on top of one’s core expertise.

              Of the hundreds of interviews I have had, only two had a chance of detecting if I was lying. When I put it like that, I am disgusted with myself for not making up more achievements. I have known people who outright lied: gave themselves job titles and roles they did not have; or claimed to have known subjects they merely knew of and expected to be able to learn quickly; or claimed experience which would have been professional malpractice given their role and seniority. In all these examples the person got a good job off the back of such statements.

              I, however, have said things such as “I can learn that really quickly” or “My experience auditing large numbers of accounts and backing files means I could easily prepare accounts”, but interviewers seem to have wanted straightforward “yes, I have that experience”.

              And if we have a few dishonest people with insufficient skills or dedication to doing a good job in positions of power, we will be even less likely to have interviews go well; they will be run by people who are in over their heads whose self-regard depends on the idea that salesman-like rapport and confidence are legitimate and important qualifications for doing the job, even beyond the standard bias that humans have that mistakes such things for ability.

              What are the results of such a problem?

              In academia, where results are measurable facts and theories, we can also see the effects of fraud. For example, one known fraud was a German anaesthetist (American: ‘anesthesiologist’) who claimed that a blood substitute for people receiving dialysis or transfusions was effective, when it turns out that it is harmful. His self-belief and trust in the underlying rationale for the product led to outright fraud and an estimatable number of deaths (ill people might have died anyway, so it is an estimate rather than precise figure).

              In another example discussed in the podcast, psychologists claimed that putting the signature at the top of a form rather than the bottom increased the likelihood of people telling the truth in disclosures, and businesses and government invested millions of dollars in schemes to use this finding in things such as tax returns or insurance disclosures. Eventually someone assessed the effectiveness of their scheme and the original research was found to have been made up.

              In business, the results will be no smaller. But because we do not assess the results of employment fraud, there is no easy reference. Many of us will have seen someone awful at the job needing massive oversight; a large amount of my time in some jobs has been making the best of mistakes and error-ridden data created by an organisation that is rolling along on estimates and hopes when real data and accurate figures ought to be available.

              The risk from underqualified people is not in obvious, discrete losses, but in hard-to-track effects. Decisions made by senior staff using rough estimates which they assume are reliable information. Risks needing detailed understanding to manage materialising and being explained away as unforeseen problems that are bad luck. Organisations don’t just fail because one person draws a salary and does nothing at all; or because an incompetent staff member spends a lot of money on a specific thing that is a complete waste. Organisations fail because each moment of incompetence is a small bit of grit in what could be a well-oiled machine. There are other sources of grit, and organisations that are not well-designed at all, and organisations that have little capacity for disruption, but the more grit, the more likely an organisation is to fail.

              We now have in top positions across all sectors people who thrived in an environment with massive incentives to lie; an environment which rewarded the ability to smarm one’s way into a stranger’s good graces over any job-related ability; an environment that allowed confidence and self-belief to be better than humility or clear-headed self-knowledge. Did they thrive because of such an environment or despite it? We cannot know without knowing each of them, and their personal history, in detail. But we can guess that across the whole population, they will tend to have those ‘beneficial’ traits more than the people who didn’t thrive.

              Is the solution to parachute more women or black people into top roles? Of course not! We have been talking about human nature and a national systemic problem; race and sex have nothing to do with it. Yet some people propose this as a solution. I think we might as well send 18 year olds to run multinational companies, since people in power also tend to be old. Young people will be less likely to have been corrupted by gradual exposure to a corrupting world; and the hard detail will have been summarised by expert staff. There is more in favour of this idea than picking on women or ethnic minorities. Yes, people in good roles learn more than people who do not get them, and so those lucky 18 year-olds will soon become genuinely more knowledgeable, but a lottery is better than a system that actively works against virtuous people.

              People who value confidence, ‘judicious’ dishonesty and charisma will recruit people like them. Because we give them a pulpit, if such people were in power, we would end up with self-help mantras such as ‘just believe in yourself’, a self-serving piece of advice which protects their ego more than helping the world… although in world where humility is a disadvantage, it could also be good advice.

              The solution is for some of these people; the slightly more sensible, balanced ones, to learn on the job as they have done throughout their careers so far, and change what they value now that they have the power they were chasing. Shareholders, journalists and the wider public should be actively suspicious of people in very senior roles rather than treating them as more respectable and insightful. Anyone who deserves the position should easily prove themselves worthy of respect through their behaviour and insight; those who have only experience and anecdote, valuable those these might be, do not deserve platforms or adulation.

              Anyone can learn experience; anyone can accumulate anecdotes; almost anyone can learn to be virtuous (although many might be so only out of conformity, rather than a deep desire to be moral). If we want to invest in people by giving them roles beyond their current, official, abilities, we should look for insight; speedy assessments of novel situations. Not coincidentally, the Civil Service entry test, which was once the gold standard for selecting able applicants, did exactly that when my mother ran it. News cuttings, reports, performance data in a pile and the task to decide how hospitals were performing, or where funding should be directed. A lie on a CV, or a charismatic character, cannot help with filtering through such things.

              No-one enjoys exams. They are stressful. Talking to people in a chat feels nicer. It feels nicer precisely because there are no hard data points generated; precisely because one can avoid exposing weakness; precisely because it can feel friendly. All the reasons it feels nicer are the reasons why it is worse. Until recruitment (except for sales positions, where being a pushy spiv is often a genuine qualification) uses tests by default, companies will be that much worse: 5%, or probably much more. And that’s a big difference in a supposedly competitive world.