When we should really challenge our own, and other people’s omission bias
Biases (cognitive and behavioural tendencies) and heuristics (mental shortcuts) are often associated with bad decisions, but it is worth bearing in mind their evolutionary origins before we label them as unconditionally problematic. Whether we are thinking of risk aversion, confirmation bias, the representativeness heuristic, or hyperbolic discounting, there are good reasons why we have been carrying these tendencies for thousands of generations.
For our evolutionary success is in part the fortunate result of these biases. Imagine how well our ancestors would have fared if none of them had been risk averse, and they’d encounter a sabre tooth tiger as they went for a stroll in the woods after lunch. Before they’d have been able to utter “Hmm, interesting!” they’d have been lunch themselves. Without confirmation bias, they’d have constantly got lost on the way to the stream, incapable of establishing there was enough evidence (“I didn’t see that leaf yesterday, did I”) that they were (literally) on the right track, and died of thirst. Without the innate predisposition to group things (predators, streams) together based on stereotypical similarity, they’d have naively walked straight into the path of any wild animal they had not encountered before, or failed to recognize sources of water. And if they had not had a tendency to favour immediate and certain rewards over uncertain, distant ones, they’d not have survived long enough to benefit from the latter.
Of course, we don’t always act according to these tendencies – context matters – and we don’t all have these tendencies to the same degree. But biases can be helpful when we don’t have the time or the resources to consider a decision in a reasoned manner.
One such bias is the so-called omission bias – the preference for inaction (even if it is harmful) over action (that is equally harmful). It can be linked with the precautionary principle, which recommends caution if an action has the potential to cause grave harm, and there is insufficient scientific evidence that it is safe. Specifically, in such circumstances, no action would be taken, or any action already happening would be paused.
The great suspension
While omission bias is not inherently maladaptive, if the consequences of doing nothing are far reaching, it can be problematic. This was arguably the case over the last week, when we have been able to observe a mass outbreak of omission bias in Europe. On 12 March, Denmark, Norway and Iceland stopped administering the AstraZeneca (AZ) vaccine following reports of blood clots in patients after they were given the jab. Over the next few days, more countries followed suit, and at the time of writing 18 countries have suspended the use of the AZ vaccine, all waiting for more information from the European Medicines Agency (which has since reiterated its recommendation to continue using the AZ product).
Uncertainty in itself is not sufficient to invoke the precautionary principle not to do something or to stop what is already being done. The future is inherently uncertain, and it is practically impossible to guarantee nothing bad will happen if we take a particular action. A few weeks ago, parts of a Boeing 777 engine came down on a residential area of Denver. Is it reasonable therefore to decide not to go out, in order to avoid being hit by falling fragments of a plane overhead? Few people would judge this to be the case – it just seems too unlikely to frighten all but the most nervous person in the world.
Could the precautionary principle rightly be invoked to justify the suspension of the administration of the AZ vaccine? Is there a sufficiently high likelihood of blood clots occurring following a jab? AZ’s own phase 3 trial data shows that four thromboembolic events occurred in individuals who received the vaccine, while there were eight in the control group. Arguably, the trial involved a relatively small number of people (24,000), of whom – remember the earlier controversy about the AZ vaccine regarding its suitability for older individuals – a small proportion was over 65 years old. However, the number of reported cases of blood clots across the EU and the UK, where 17 million doses have been given, is about 30.
This is proportionately a lot smaller than in the trial (1.8 per million vs 167 per million), but this in itself does not mean blood clots are not being caused by the vaccine. We need to compare the rate at which these events occur in people who received the vaccine with the base rate across the unvaccinated population. That base rate is around 1:1000 per year – so in a group of 17 million people we would expect 17000 cases per year, equivalent to well over 300 per week. Bearing in mind that the risk of blood clots rises rapidly with age to around 1% for people over 85, and that most of the vaccinated people are over 60, we’d expect even more. If anything is striking about the number of cases of blood clots among those vaccinated, it is that there are far, far fewer than we’d expect.
Irrational motives?
The numbers don’t remotely suggest that the AZ vaccine is the cause of the blood clots. So what might nevertheless be inspiring the policymakers in 18 countries to hit the pause button? A clear candidate is base rate neglect, the failure to take into account the background prevalence of a phenomenon, and only look at it in the presence of a presumed cause – ignoring statistics and being led by specific events. A possible contributing factor is the salience bias: undue focus on what stands out (like a case of deep vein thrombosis shortly after the administration of a substance), and little or no attention to the vastly larger number of cases where nothing happens. Another likely influence is the post hoc ergo propter hoc fallacy (Latin for “following this, hence caused by this”). It is indeed often tempting to see (and indeed seek) a causal connection between two successive events. But two events merely succeeding one another is not remotely sufficient to imply a causal link. I am sure you don’t believe the sun went up this morning because you went to bed last night.
We have probably also seen herd behaviour, a form of conformity bias, describing a tendency to take cues from others, rather than develop and follow one’s own judgement. This is sometimes done to avoid the effort of hard reasoning, but more frequently it is to avoid embarrassment or blame (sometimes referred to as CYA). And perhaps it is the fear of being blamed that ties all these irrational motives together into one possible fundamental reason for omission bias: when things go wrong, you’re less likely to be blamed for being excessively cautious, than for taking a risk. Risk aversion can have a strong self-preserving element.
Omission bias is a blunt defence against acting unwisely. To verify whether we are (or someone else is) exhibiting this bias inappropriately, it is a good idea to check for the potential presence of any or indeed all of the above influences.
Don’t forget the trade-offs
But let’s imagine that there was sufficient evidence for a serious side effect, for example, because thromboembolic events occur more frequently than the base rate. Would that in itself be enough reason to halt immunization with the AZ vaccine? Doing so would then prevent possible side-effects. But to justify it, this benefit would also need to be weighed up against the disadvantages. Delaying vaccination will lengthen the time before the pandemic comes to an end, and hence lead to more economic cost, and more lives lost to COVID-19. Furthermore, the suspension will stoke up vaccination hesitancy. It is not clear whether this trade-off was made in the 18 countries.
Just like in the notorious Trolley Problem, we have a tendency not to interfere, in the belief that doing nothing leaves things in some kind of natural state, and therefore we cannot be blamed. But pausing the use of the AZ vaccine is a deliberate decision too, which will not have been without consequences. They will be measured in millions of euros, and in hundreds of lives lost.
In contrast with what the Italian saying, dolce far niente, the sweetness of doing nothing, suggests, there can be a distinct bitterness to doing nothing.
Koenfucius 