Sway by Pragya Agarwal review – how we are all unconsciously biased

In experiments, people perceive an approaching spider to be moving much faster than it really is, and faster than a ping-pong ball or other neutral object moving at the same speed. It is reasonable to deduce, then, that humans generally have an unconscious bias against spiders. Nothing too depressing about society follows from this. But the idea that we are prey to unconscious bias in more important areas – to do with decision-making, and how we treat our fellow bipeds – has in recent decades become a hot topic. It is at the root of what is called “behavioural science” and “nudge politics”, which reports suggested were driving the British government’s laissez-faire early coronavirus strategy. But how strong is the evidence that it exists?

It was the field of behavioural economics, as described in Daniel Kahneman’s bestseller Thinking, Fast and Slow, that demonstrated that humans do not make mathematically perfect decisions about probability; they instead rely on rough rules of thumb, and often go wrong. Some of these habits are uncontroversial, such as confirmation bias (you tend to notice only the evidence that confirms what you already believe). But the inference by the field that such biases mean – as Pragya Agarwal uncritically repeats here – that “humans are not naturally rational” is extremely dubious, especially since it depends heavily on which definition of rationality you use.

For example, the finding that people care about the possible loss of some money more than they care about the possible win of the same amount is considered an irrational bias, because mathematically they are equal sides of the same bet. So trying harder to prevent the loss is called the bias of “loss aversion” – at least among comfortably tenured academics. On the other hand, if you have very little money in the first place, it is perfectly rational to care very deeply about losing an amount that might ruin you, and it is sensible not to risk that for a 50/50 chance of winning.

Agarwal, a data scientist who has worked on the dynamics of social interactions, also endorses some of the wilder claims of evolutionary psychologists in the area of rationality. My favourite among these is the crazy theory that, as she endorses it, “our brains have evolved to reason adaptively rather than rationally or truthfully”, which is the fashionable modern version of the Cretan liar paradox. If it’s true, after all, then I have no reason to believe it is true.

That’s not to say that we are not prey to any failures of rationality at all, of course – a fact already conceded by Jonathan Swift when he sardonically modified Aristotle’s definition of man as the rational animal to man as “the animal capable of rationality”. And Agarwal tells some colourfully persuasive stories about how such potential bugs in our mindware contribute to the degradation of our public sphere. The flipside of confirmation bias, she explains, is the less snappily named “disconfirmation bias”, which is when we “spend considerable energy in denigrating arguments that run counter to our existing beliefs”. Remind you of any particular social media platform? “Often when we are picking up our phones or tablets and browsing, we are doing so while waiting in a queue or having our coffee or pretending to be working, and so we skim, share and retweet quickly,” she writes, almost as though she has had a crack surveillance team following me for years, and it is this very rapidity that makes biased responses more likely. Cleverly, Agarwal also notes an asymmetry in terms for the phenomenon whereby people on social media end up hearing only from people who think like themselves. I, an intellectual, lament the fact that I am in a “filter bubble”, while you, a bigoted yahoo, live in an “echo chamber”.

Evolutionary explanations must be part of the puzzle here, but it is often tempting to exceed their proper scope. Prejudice against people not like ourselves, for example, doubtless originated in early humans for perfectly good reasons. “Humans appear to have a bias that members of out-groups or competitive tribes are less magnanimous and more dangerous than those who belong to our in-groups,” Agarwal writes, which was fair enough on the savannah. But such explanations might imply that these biases are innate, so she also denies that they are “necessarily hardwired”, and suggests they can be unlearned. This optimistic message, though, can sit in awkward tension with the simplistic neuro-determinism by which she claims that specific brain regions (such as the amygdala) or hormones (oxytocin) programme our responses.

For issues of racial and other discrimination, do we really need the idea of “unconscious” bias at all? Someone scrawling Islamophobic graffiti on a wall is obviously quite conscious about it. Agarwal has chosen throughout to use the terms “implicit bias” and “unconscious bias” interchangeably. The first is linked to a controversial psychological tool called the Implicit Association Test, the arguments over which show the difficulty of proving the existence of something that is, ex hypothesi, unconscious and so unavailable to the subject’s introspection. So the IAT uses reaction times as a proxy for the assumed bias: if a white person is slower by some number of milliseconds to associate a black face with a desirable social role, for example, then they are said to have an implicit bias against black people.

Despite doubts over particular methodologies, we can agree that there must be something to the idea that racist and sexist prejudice can operate below conscious awareness, and Sway succeeds impressively in its drive to show that justice is undermined by a rich array of social biases about age, gender, race, accent, and so forth. Unconscious bias is a plausible explanation for why so many white American policemen, even those who are not explicitly racist, end up shooting so many unarmed black men; why hurricanes given women’s names (eg, Katrina) are perceived as less frightening than those with men’s names; and why nearly a fifth of people in Britain who are stopped in the street and asked to prove their immigration status are, in fact, British. (Agarwal is appropriately scathing about the “post-racial” claim that there is no racism in Britain any more: an idea, I have noticed but she is too polite to say, that tends to be most eagerly promoted by those who are, in fact, racists.) One can disagree with her presentation of the science, but Agarwal’s diagnosis of the political harms of bias is passionate and urgent.

What can we do about our own biases? The key is to pause, breathe, and realise that the spider is not really racing towards us so fast. Research on “debiasing” has attracted scepticism from those invested in the “nudge politics” model – understandably, because for nudge politics to work, we must remain biased automata vulnerable to its paternalistic guidance. And yet many biases can be overcome if we are made aware of them, and incorporate a check for them into our slow, deliberative thinking. “We need time, intention and adequate cognitive capacity and resources to be aware of the activation of stereotypes,” Agarwal writes hopefully, “and then to significantly reduce the application of any stereotypical beliefs on others around us.” Her own book can, perhaps, help readers do the same.

• Sway is published by Bloomsbury (RRP £16.99).