Monthly Archives: July 2017

Fighting bias with quota and lottery

Before you cry out “quota? – no way!” imagine you have the power to decide which of the wonderful girls in the Genius school can enter the charmed circle. Your first impulse might be to select students who are most like the existing members of the circle, the prototypical candidate. On second thoughts, you might wonder which talents and skills the circle is currently missing out on. It is obvious that a diversity of skills would be most advantageous for its future projects and adventures. Selection is a stressful task, because, if you selected the wrong student, the circle will weaken. At the same time your reputation as a shrewd selector would suffer.

Diversity is a necessity, given the changing demographics in our societies, but it also brings benefits. There is a business case, and there is evidence that science thrives in diverse teams. Both the number of hypotheses and the number of solutions are increased if groups working on a problem have different perspectives and contribute different points of view. A diversity of viewpoints helps us to avoid being stuck in a rut. Clearly, charmed circles, such as academy members, grant holders and award winners, need to be diverse if we have any hope of solving the pressing problems of our times. Continue reading Fighting bias with quota and lottery

Increasing diversity is not easy and the previous post mentions some of the hidden biases that make it an uphill struggle. But we should not feel defensive about this. Remember, our social brain is processing vast amounts of information in split seconds and therefore has short cuts and preference settings. These lead us straight back into the same old choices. If we want greater diversity, we need to interrupt this processing flow. There would never be any change if we didn’t. To become aware of hidden biases is one way to achieve this, but setting quota could be a better way.

Lets talk about Quota

With the charmed circle at Genius school it would make sense to have a quota for missing skills. You could also insist on a quota for boys, in case the school decides to become co-ed in the future. With modern selection panels the same applies. Under-represented minorities have special skills and preferences that go to waste because they are not in the circle. What’s sad is that the existing members may not even realise that this is their loss.

Quota, in theory, can easily be applied, as long as there is a pool of excellent candidates. But there is a snag. In selection for excellence, who would like to think that the reason they were selected was because of quota? This is not exactly bolstering their self-esteem! It is also a reason for feeling denigrated if you are not selected. Some will feel that they should have been selected by rights, but did not have the required minority status. Everyone will feel demeaned.

The situation changes when we take into account the size of the charmed circle. If there is only one winner, quota setting will probably not work. But it can work well, when more candidates have to be attracted. For instance, setting a quota for boys when the school becomes co-ed. New people with different ideas and different skills can be selected as soon as a quota is adopted. Quota here is strictly for the purpose of building capacity. It says ‘we are open for you – even though we weren’t open to you before – please apply’.

It follows that my personal view on quota is twofold: At this point I would probably not endorse quota when the selection is about giving rare awards. In contrast, I would endorse quota when the selection is about capacity building. The most likely scenario here is that there are many candidates and the number of awardees is somewhat flexible. This is different from the typical scenario of rare awards. When there are many excellent candidates, but, for whatever reason their names did not come to mind and they were not put forward, then a strong case for quota can be made. One practical move adopted by some learned societies is to add placeholders for individuals from a desired group (e.g. underrepresented minorities) without touching the historical annual quota for new members. This is capacity building because it means that additional members can be elected as long as they are from an underrepresented group. Clever. Nobody loses, and the institution wins praise for its willingness to change.

Even if there are no set quota, beware that as a selector, you will get eggs thrown at you if you systematically exclude minority members. There has been a “softly, softly” approach that avoids setting quota but strives for gender balance, and this has been quite successful in many areas, where charmed circles have been broken into. Speaker lists at conferences that do not have a reasonable proportion of women are rightly condemned. Academies whose membership remains stubbornly ‘senior white male’ have come in for criticism and even derision. Reputation is tied in with the willingness to modernise, not with continuing old traditions. Fortunately, the desire for a good reputation is such a strong motivation that changes in selection practice follow of their own accord even without the formal need to set quota.

Lets talk about using a lottery element

Instead of making impossibly difficult decisions, why not choose by lottery? This can make sense when the candidates are of equal merit, and in this case, a lottery is the one tool we have to definitely rule out any hidden bias. Juries in civil and criminal courts are selected randomly as a first step in the process. City states and universities have employed lottery procedures (also known as sortition) with the effect of minimising corruption and avoiding conflict. Famous examples include Athens, Venice and Basel. In these historical cases governing elites opened up to individual newcomers, but avoided being taken over chaotically. In addition, if an individual was rejected by lottery their honour was protected.

As a candidate, how would you feel? If you are selected you will feel lucky, but you should not feel big-headed; if your are not selected, you will feel unlucky, but should not feel disparaged. As a selector you would probably feel confused! Your job after all is to discriminate between candidates – the very opposite of leaving selection to chance. UsIng chance seems like abandoning reasoned argument. But sometimes comparisons between candidates are impossibly difficult. Here, bias is likely to gain the upper hand. The trigger-happy unconscious mind will always nudge the dithering conscious mind. Post-hoc justification of intuitive choices is not a triumph of reason.

Here again there is a practical move, already adopted experimentally by some funders of grant applications: a lottery element is used as part of the usual procedure of discussion among panel members. Panels often agree on their top candidates who clearly stand out from the rest. They also usually agree on candidates who are not competitive. This leaves a middle field where there is less certainty. Through discussion consensus can sometimes be reached which allows separating the stronger from the weaker cases in this middle field. But there is often not enough funding for all the strong candidates, and those just below the top tier will be the subject of much discussion which leaves selectors often uneasy about their final choices. This is where a lottery could be the tool we have been waiting for. It is the one tool we have to avoid bias and guarantee fairness.



The trouble with unconscious bias

Fine-tuned balancing scales are a beautiful thing. We can find out with great accuracy which of two objects is heavier. We’d like our mind to have just such a device when we have to evaluate the merit of different individuals. Then we could make fair decisions about which grant proposals should be funded, and where prizes should be awarded. But, alas, we have no objective scales to measure differences in merit. Instead we have clunky subjective scales that operate with distinct biases.

I have written about unconscious bias before, e.g. here linked to a short animation, and here.  In a more lighthearted vein I also developed a cartoon dialogue with the BBC 100Women programme. I still stand behind what I said in 2015, but I also feel that reminders and refreshers are important. Gender bias is still grimly hanging on. I have been trying to keep track of the continuing empirical work on bias and I am convinced that there is more work to be done. Hence the present post, which is paired with another post on my current ideas on quota and the possibility of using a lottery element to achieve truly unbiased selection.  Continue reading The trouble with unconscious bias

Are we really unable to select for excellence on meritocratic grounds – pure achievement, as opposed to wealth or position? We like to believe we can, and we have set up thousands of selection panels and prize juries on the basis of this belief. But doubts are creeping in. There seems to be a way for race and gender prejudice to tip the scales behind our backs. Prejudice seems ubiquitous and it seems to start very early in childhood. We like to think this could be an unwanted effect of our environment, something that can be reversed by training. In fact, it is something located deep down in our mind/brain and not easily erased.

Why are our mental scales so prone to bias? Perhaps as far as our ancient social brain is concerned, accuracy isn’t everything. When we interact with other people there are more pressing concerns than discriminating subtle differences in achievement. No one knows how many bits of information have to be sifted to make quick decisions about other people, whether they belong to our group and whether they will cooperate or compete. All we know is that there are shortcuts that make it possible to constantly align with other social agents, if only to be able to fly in the flock.

Getting along with other people is vitally important to us. We automatically assess them with brute self-interest, mostly hidden from our conscious Self. This is why the mental scales are biased and why they let down selection panels. This makes us uneasy because our considered assessment of individuals is often completely at odds with our crude automatic assessment. The unconscious part of our social brain is trigger happy in recognising in-groups and out-groups. Yet our feelings of belonging are fluid and we can belong to many different in-groups. It makes no sense to avoid strangers because their appearance evokes a threat response in our ancient amygdala, when in fact they might become useful allies. We have learned that we benefit more from cooperation than from conflict, and that oddballs can make outstanding discoveries. We have learned that men can be nurturing and that women can be competitive.

All this learning does not erase the ancient short-cuts that the social brain has acquired over eons of evolution. However, we can agree that work of selection panels improves if we process information about candidates in a slow and deliberate fashion, using the conscious part of our brain. When we need to make accurate judgements about the merits of different individuals when we allocate grants or awards, and this means we need to becalm our highly inflammable social brain.

Here I want to mention just three shortcuts that guide information processing in the unconscious social mind/brain. First, the so-called ‘availability heuristic’. This results in a preference for the familiar, for example for a candidate who is similar to previous successful candidates. Second, there is a desire for affiliation, which means we implicitly favour those who are like us and/or those who belong to a more socially dominant group. Third, we all believe that we are less biased than other people and have better arguments. We also believe that we are less subject to conflict of interest than others. For example, 61% of doctors thought pharmaceutical industry promotions did not affect their prescribing; only 16% believed this to be true for other doctors.

Knowing about biases does not neutralise them. They operate below consciousness and it is very unlikely that we can delete our unconscious short cuts. It might even be dangerous to do so, with unwanted side effects. But there is one way to make to ‘put them in their place’, in a manner of speaking. We can monitor and challenge each other because we see bias more easily in others than in ourselves. This is a bit uncomfortable – because we cannot see the beam in our own eye while we can see ‘the mote in our brother’s eye’. Of course we should not blame each other. It’s human to have unconscious biases.

What can we do to counteract bias? Not through intensive training programmes (see Footnote). However, it is useful to be aware of unconscious biases and selection panels need to be reminded of them as they make their difficult decisions. Slowing down the decision making process allows the conscious part of our brain to reflect and to query the reasons for our rash intuitive judgements. This is best done in groups when we can discuss different reasons. We can never ever be unbiased because this is how the brain works, where strong prior beliefs are affecting our perception and experience. Once we admit that subjective factors play into our judgement we can be more sceptical of our feelings. We can’t help it that our feelings are subtly biased against minority candidates. It’s precisely because there are so few of them. It means they fall outside the norm, always an awkward place to be.

Diversity helps us make better decisions (more on this in the next post). We now know that a diversity of viewpoints helps us to avoid being stuck in a rut. By listening to others’ point of view we can counteract the fact that we tend to be more critical of others’ theories, and uncritical or our own.

Footnote Unintended consequences of conscious efforts to counteract bias.  Brown et al. 2011 argued that people feel licensed to act on bad motives if they feel they have the moral high ground. Affirming one’s egalitarian or pro-social values and virtues subsequently facilitates prejudiced or self-serving behaviour, an effect referred to as “moral credentialing.” In a study people who had ‘credentialed themselves’ were more likely to cheat in a maths test, especially if they could easily rationalise this behaviour. In another study Monin & Miller, 2001 showed that people are more willing to express prejudice when their past behaviour has established their credentials as morally superior non-prejudiced individuals. Here, people were first given the opportunity to disagree with a blatantly sexist statement. Later they were more willing to favour a man over a woman for a stereotypically male job. Other studies confirmed this rebound effect.