If there's one thing that COVID-19 has shown me, it's that implementing behavioural science, or fully relying on behavioural insights can do just as much damage as it can do good.
Let me start of this article by saying that I am a behavioural scientist, that I love my field and that I think it's has shown a lot of promise, has brought forth great research with some great results and effects and that its existence is an addition to both academia, industry and the world in general. But not always!
The idea that a small change can lead to a difference, even a big difference, is great. But the issue here is that you need to be very clear on what constitutes an actual difference, let alone a big difference. Now most behavioural science, when it comes to testing for differences and whether an intervention has had an effect, is grounded in significance testing: where an alpha level of 0.05 and lower is seen as significant and significantly different. To have an alpha of 0.05 or lower, means that there was a 5% (or lower) chance of the measured outcome occurring, if we were to accept the null hypothesis is true. The null hypothesis is the assumption that there is no difference. So if we think there isn't actually a difference between having an intervention and not having the intervention, the outcome that we now have only had a likelihood of 5% (or lower) to occur. Given that this is very low, we reject the null hypothesis, and call the difference significant. Now what determines the significance? Well, sample size is one of the things. The amount of data available will more accurately represent the mean and the standard deviation. These are two measures feeding into indicating what is significantly different. So if you have a very small sample, you can misrepresent the actual distribution of data, which can lead to a difference becoming significant when it actually falls right within the biggest chunk of the normal distribution (bell curve), or the other way around.
Why is this relevant? Well, with small sample sizes a non-significant difference can constitute 40% (depending on what you're testing), because you can always fall on an outlier that isn't balanced out. Whilst a significant difference when having a large sample size might actually only constitute a 2% difference. For example: there are researcher who have tested several interventions when it comes to paying taxes. They found that some interventions can lead to a 2% (significant) increase in people who pay their taxes. This effect does make a big difference in the amount of money accumulated in tax revenue, but 2% still seems like a very small impact.
Now you're probably wondering why I thought writing a really basic statistics segment was a good idea. Well, I didn't think I would write one, but it is important to know that a big difference and s significant difference are very different. Within science, it's often the latter that people are focussing on. Now why does that matter? Well, if you need a big difference, rather than a statistically signifcant difference, this is not the intervention for you. I might be an odd one out as a behavioural scientist when I say that if you want big and immediate results, behavioural science is often not what you're looking for. You know what is? Policy and law. If you want to stop people from doing a certain behaviour, you can educate them, you can provide them with better options and you can nudge them towards those better options. But this still leaves all options open. Some people perceive this type of governance as paternalistic, others just think it's weak. However the case, it doesn't send that clear a message as to what the repercussions are for chosing the "non-preferred choice." And in this case, non-preferred refers to society as a whole, not to the preferences of the individual. When it comes to nudging and behavioural science-type interventions, often the repercussions of choosing the non-preferred option don't actually fall on the individual. When it comes to annihilating undesirable behaviours and helping society, you're going to need a bigger effect than simply a few percentiles. The message needs to be clear and the repercussions need to be clearer. If you make something illegal the message has been made pretty clear: you can still choose all options, but chosing "wrong" will have immediate and clear repercussions for the individual who made the choice. The level of accountability is high in this case. Let's actually examplify this: An example I often think of is flying. Flight prices, especially when short-distance, are dirt cheap. Sure, these companies try to overcharge you for luggage that never seems to be "the right size" and all that type of shenanigans, but still, the very low ticket price remains. This ticket price is often so low that I cannot imagine them actually paying for the entire cost of the flight. I'm not talking flight staff, plane maintenance and fuel, I'm talking about the detrimental effects getting a plan into the air has on the environment. There is no way that ticket prices are accounting for that cost. To do something about this behaviour (the blatant negligence of the environment for profit), who are you going to nudge? The consumers? They still have to get to their destination. A lot of flights that stay on the continent are taken by business people. Are you going to nudge the companies? How do you nudge someone to give a damn about the environment? Well, I've got a behavioural measure for you: incentivisation. What incentive structure am I proposing? Well, you can go two ways: help pay for the flights and their effect on nature (sounds a lot like an agricultural subsidy, another sector which is a massive polluter, and OVERproducer...) or punishing the behaviour. You can implement policies and laws that require that the prices charged take into account the detrimental environmental effects. But which government has stuck out their neck to do that? No one. You wouldn't want the graces of multinationals now, would you? When it comes to sustainability there is a lot that behavioural science has been able to do on a consumer level, but it's not enough. Especially not if every success on an individual level is set off against the output of massive multinations. You're trying to herd the mice into one corner whilst you have an army of elephants stomping through the place, that makes no sense. I'm not saying these things don't add up, because they do. But as soon as a lot of unsustainable behaviours would have been made illegal, for both consumer and company, the effects would have been ten, if not hundred-fold.
It seems some countries and governments have gone a bit too crazy about nudging all their problems away. Behavioural science is almost used as a way of not having to make unpopular decisions and implement harsh measures. Well, as a behavioural scientist I have this to say: grow a pair.
Let's not forget that sometimes we have to make unpopular decisions to even make it to the next election!
