Behavioural scientists treat us all as bundles of bias. We are swayed by the first piece of information we see (anchoring bias), we do not doubt ourselves enough (overconfidence bias) and we dislike change (status quo bias). We’ll fight harder to avoid losses than to make equivalent gains (loss aversion) and we’ll overstate the chances of extending a winning streak (the hot-hand fallacy). But these ideas have recently suffered a backlash: to err is human, and neither behavioural scientists nor their critics are immune.
So many of us were seduced. Nudge, by Richard Thaler and Cass Sunstein, published in 2008, sold over 2mn copies. In the 2010s, cash-strapped governments embraced the idea that interventions as subtle (and cheap) as changing the default option on a form could have big effects. Today, the OECD lists more than 150 behavioural insights units that are part of federal governments. Boring old policymaking was out. Choice architecture was in.
Critics contended that bias merchants were overselling their wares, both when it came to the theory and its application. Some papers describing the power of priming did not withstand scrutiny, most recently one claiming that a declaration of honesty at the start of a tax return rather than the end encourages more truthful reporting. Others have questioned the core result that losses loom larger than gains. Critics of behavioural science in practice dismissed nudges as having tiddly effects, and of distracting from structural problems such as poverty or climate change. As one recent podcast snarked, “what if the government stopped doing politics and started doing marketing?”
So is behavioural science the best or worst thing ever to have hit an airport bookshop? The boring answer is that progress in social science is messy, and while it can be useful to note where our mental shortcuts lead to errors, if we’re going to save the planet we need something stronger. But boring isn’t going to hold your attention for the rest of this column. There are lessons for the field in the biases it studies.
First, beware of the tendency to discard specifics and form generalities. A forthcoming study in the Journal of Economic Literature gathers 607 estimates of loss aversion, and finds that people hate losses around twice as much as they value gains. But while loss aversion might hold on average, it is not law. Colin Camerer, one of the authors, warns that within any population a big minority of people display the opposite bias, or loss tolerance. Other recent research suggests that when the stakes are low, loss aversion goes away.
Next, avoid anchoring expectations to very snazzy results. One comparison between nudges measured by academics and by nudge units found that whereas studies produced average effects on some desired behaviour of 6.4 percentage points, practitioners saw effects only a fifth as large. (Most of the gap was because dull findings are hard to publish.) If you anchor to the effects in academic papers, everything seems small, says Elizabeth Linos, one of the authors. But a small effect might still be worth pursuing, if it is cheap enough.
Third, do not underestimate the power of the status quo. Another study explored 73 randomised control trials in America that tweaked government communications, and found that only around 30 per cent were put into practice after the experiments ended. The best predictor of whether something actually changed was whether there was already a channel of communication with the public on which to piggyback. Whatever your whizzy ideas, policymakers are people too, with competing priorities and limited attention.
Finally, watch out for a bias towards bias. That doesn’t mean you should worry about widespread fraud; it does mean being wary of isolated but cute results. And it may mean trusting your instincts. Take the hot-hand fallacy, which a 2018 paper showed was not a fallacy at all. Once the authors fixed a mathematical flaw in the original study, evidence of winning streaks in basketball emerged. The coaches and players who were convinced it was real were right.
Some practitioners are trying to improve. In March, Britain’s Behavioural Insights Team, which was spun out of the Cabinet Office in 2014, published a manifesto for applying behavioural science, billed as a response to the backlash. It included aspirations to recognise “how apparently universal cognitive processes are shaped by specific contexts”. Awkwardly, they had a lesson for me. Lists of biases may be accessible and popular, but “can create overconfident thinking that targeting a specific bias (in isolation) will achieve a certain outcome”. Everyone makes mistakes.