The perils of survivorship bias in science and academia


This is a guest post by Dave Hemprich-Bennett.

In academia and science we pride ourselves in being evidence-led. Our research stems from countless hours of painstaking work, yet when we give advice or plan our futures we fall back onto ‘common sense’, conventional wisdom and personal experience. However it is important to realise that we are not perfectly rational actors and so often fall afoul of basic logical errors, one of which is forgetting how unrepresentative we and our peers are of those we seek to help.

Survivorship bias is a common logical error where we draw conclusions based upon those who have passed some sort of selection process while ignoring those that did not, typically because they are less visible to us. We focus on the ‘survivors’, while forgetting those who went through a similar process and did not ‘survive’. A simple form of survivorship bias would be deciding that an accident was not especially dangerous, as everyone we spoke to afterwards hadn’t died. The obvious issue with this would be that the people we spoke to after the accident would by definition be alive and able to talk about their alive-ness, but it doesn’t necessarily follow that they are representative of everybody who was involved. They’re just the visible exceptions who made it through some selection process.

Weirdly this bias is entrenched in academia and science, where we often find ourselves giving advice to and setting expectations of those earlier in their careers than us. Most of us have only good intentions when we do so, but our perspectives are inevitably informed by our own previous experiences and those of our peer-group, all of whom are by definition ‘survivors’. In doing this we accidentally assume that our peer-group is representative of those earlier in their careers than us, when in fact our peers are actually representative of those who have overcome various rounds of selection. Because of this we tend to underestimate how damaging many such selection processes can be to those going through them: after all, we ‘survived’, and so did most of our peer-group. Thus we end up giving advice that may have been good for our younger selves in our past circumstances, but that is often misinformed for the recipient. In addition to this, our previous success can stop us from noticing many aspects of those selection processes, even where they prevent many other suitable candidates from progressing.

On Twitter we seem doomed to spend eternity debating the pros and cons of unpaid labour, which is either an ‘unfair hurdle’ or ‘paying our dues’ depending on the speaker’s perspective. Many feel the requirement to work unpaid is fine: after all “I did it, and it never did me any harm”. However that perspective ignores the simple truth that to be making this argument the speaker was one of a minority who would be unharmed by this hurdle, whether through family support or financial background. This advantage will have then benefitted them later on, making them more likely to succeed and enter a position of power, from where they can confidently state that unpaid labour isn’t an issue.

Hardships are often perpetuated the most by those who have suffered, but overcome them. As an example, financial limitations shaped my postgraduate education. This included taking a ‘less competitive’* masters degree than many peers because it was cheaper than internationally renowned ones. Now that I am a postdoctoral researcher it is tempting to say that the financial hardship and ‘less competitive’ degree were irrelevant: my abilities were all that counted, and so anyone with potential merely needs to work hard enough and they will succeed. But this is precisely the wrong interpretation, as basing any conclusions on my ‘success’ ignores the many in similar circumstances whose progress was stopped by them. For every person who manages to struggle through these financial barriers there are many more who did not, the issue is that these cases are typically much less visible. 

This survivorship bias is found in many other aspects of science and academia. Hazing remains widespread, with people assuming that because they went through similar circumstances and persisted, perpetuating such behaviours is fine, or even required and beneficial. More benignly, it comes into play with a lot of career advice, with the common statements “follow this course, it worked for me” or “keep at it, things will work out eventually”. We’re regularly advised to work extremely high numbers of hours, usually by somebody who is in the minority able to do exactly that. 

Again, of course the speaker thinks these things are true; for they are one of the subset for whom that was the case. But these attitudes ignore the fact that many who have attempted a similar path did not pass the selection process. In much of this, the ‘survivor’ may not be passing judgement on whether the barriers are fair or not, but it is crucial that we do so. Many of the potential hurdles which we have overcome, knowingly or not, had no reflection on our capabilities as scientists or academics. Keeping them in place, whether consciously or not, pushes excellent people out of their chosen career paths. It is no coincidence that those who are earliest in their careers seem to be most strongly opposed to the status quo: the system selects for those who are least disadvantaged by it, so those at the top are inevitably less opposed to how things are. Those who are more easily able to overcome these barriers often have social capital, privilege and resources at their disposal, which many in underrepresented groups in academia do not have. These ‘assets’ have no bearing on someone’s ability as a researcher. We must acknowledge these factors and try to reduce them, to ensure that someone’s abilities, rather than their background and circumstances, impact on who succeeds.

This checking of biases is more important now than ever: in the age of COVID-19 our students are trying to survive in a world very different from the one in which we went through. The repercussions on these cohorts of students will last a generation, touching on nearly every aspect of professional life (attending conferences, publishing papers or chapters, completing courses, or finding that next position). Falling back onto our notions of what was once fine for us is now more erroneous than ever.

*Any notion of the ‘competitiveness’ of a university or degree programme is of course nonsense. But that’s an argument for many other blog posts.

4 thoughts on “The perils of survivorship bias in science and academia

Leave a Reply