By Michael Corballis 15/11/2018


It appears that there is something of a panic, especially in our universities, over the possibility of unconscious biases, especially against minority groups. This has led to workshops and publicity material designed help us recognise our biases and correct them.

Much of the impetus for this has been driven by a psychological test known as the implicit association test (IAT). When applied to racial bias, it works like this: People are shown combined pictures and words, and asked to respond quickly whether the word is “good” (such as happy) or “bad” (such as murder), and then asked to respond “white” or “black” depending on whether they see a white or a black face. If they respond more quickly to “good/white” and “bad/black” combinations than to “good/black” and “bad/white” they are deemed to be biased against black people. Responses are entered quickly on a keyboard, so that the responder doesn’t have time to apply conscious thought, implying that the bias is unconscious.

The test was devised 20 years ago by researchers at Harvard and the University of Washington, and was at first widely admired and adapted to different settings. But serious doubt has crept in. Even its inventors have had second thoughts, and in 2005 effectively conceded to the widespread criticism the IAT attracted.

As often happens, claims from psychological research are over-hyped and spread into the public arena, where they can persist long after they have become scientifically discredited. This indeed seems to be what has happened to the IAT.

For a start, it does not meet the usual criteria for psychometric testing—or even come close. It has low reliability, meaning that if the same individual is tested again, the score is likely to be different, possibly even reversed. More importantly, it has very low validity, which means it is a very poor predictor of actually discriminatory behaviour. A person may be deemed to be biased according to the IAT, but show no sign of it in everyday commerce. What this means is that the IAT should not be used as a diagnostic tool in individual cases.

That said, when large samples are tested, there does seem to be a correlation between IAT scores and measures of explicit discrimination, but the correlation is too small to be more than weakly indicative. According to meta-analyses, the extent of discrimination variance explained by the IAT range from an optimistic 5 per cent to about one per cent. Either way, it’s tiny, and we’d be much better looking at more explicit and reliable indicators. In the US, at least, you need only examine the public record of the current President.

Delving into the unconscious can be perilous. A test as erratic as the IAT can lead to accusations of racism or sexism in people whose actual behaviour is exemplary. This, in turn, can create harmful self-recrimination and unwarranted feelings of guilt. It can be used to political advantage, stirring up mass movements or even hysteria that is otherwise undocumented.

The unconscious also featured not so long ago in neo-Freudian arguments that various forms of psychological maladjustment were a result of early childhood abuse. Memory of this abuse was supposedly repressed, buried in the unconscious. The trick was to help the afflicted individual to recover those memories, and find a healthy resolution. This too raised serious problems.

For one thing, such memories can be false, and easily implanted through aggressive therapy. This means that people are sometimes falsely accused of abuse, and in some cases incarcerated. Punishing the innocent can be as damaging as failing to indict the guilty. The presumption of innocence until proven guilty is, in any case, a feature of most legal jurisdictions, and a human right under the United Nations’ Universal Declaration of Human Rights. Evidence based on the supposed unconscious should never provide adequate proof of matters to do with everyday conduct and individual safety. And it’s not just the accused who suffer. People led to believe they were abused when they were not carry a false image of themselves and their own pasts, as well as a false sense of injustice.

It is, of course, desirable to rid society of destructive human activity, such as child abuse and racial or gender bias, but this should be based on accurate evidence. In particular, it’s much better to avoid the murk of the unconscious and focus on what people actually do or say. Appealing to the unconscious is a bit like reading tea leaves or the formation of clouds, more likely to create prejudices and set up witch-hunts than to reveal truth.