Books : reviews

Iris Bohnet.
What Works: gender equality by design.
Harvard University Press. 2016

rating : 3 : worth reading
review : 24 October 2017

Gender equality is a moral and a business imperative. But unconscious bias holds us back, and debiasing people’s minds has proven to be difficult and expensive. Diversity training programs have had limited success, and individual effort alone often invites backlash. Behavioral design offers a new solution. By de-biasing organizations instead of individuals, we can make smart changes that have big impacts. Presenting research-based solutions, Iris Bohnet hands us the tools we need to move the needle in classrooms and boardrooms, in hiring and promotion, benefiting businesses, governments, and the lives of millions.

What Works is built on new insights into the human mind. It draws on data collected by companies, universities, and governments in Australia, India, Norway, the United Kingdom, the United States, Zambia, and other countries, often in randomized controlled trials. It points out dozens of evidence-based interventions that could be adopted right now and demonstrates how research is addressing gender bias, improving lives and performance. What Works shows what more can be done—often at shockingly low cost and surprisingly high speed.

We hear a lot about people trying to fix gender equality issues with little success, or even denying there is a problem. This book provides evidence that there is indeed a problem, why many existing approaches to tackling it don’t work, and lays out proven ways to ameliorate it.

I’m taking it as given that there is a problem, but if you disagree, Bohnet provides plenty of evidence of its existence. The problem is not restricted to the misogynists actively discriminating; it is also due to the unconscious biasses we all have, sabotaging our best efforts. But unconscious bias training has very little effect. And stereotypes are hard to overcome: if women aren’t appointed to certain positions because of the stereotype that they aren’t appropriate in those positions, then there will never be any evidence to overcome the stereotype. So what should we do?

The answer Bohnet advocates is behavioural design: changing not our innermost biasses, but nudging what we do in the right direction. After all, a bias that is never acted on doesn’t really matter. So Bohnet lays out a series of design changes – to our hiring and promotion processes, to our team building, to our norms, and more – to make it easier to act in a more inclusive way. These can be implemented piecemeal using an unfreeze (the old behaviours) – change (to the new behaviours) – refreeze (stick with the new behaviours) process. (This reminded me somewhat of Arnold’s microresolution approach; the changes here are “micro” relative to the business scale.)

Examples of behavioural design include: recruit staff in batches, rather than one at a time, to reduce the temptation to go for the standard option, and to allow for more diverse choices; interview one-on-one rather than in panels, and aggregate the individual interviewers’ independent scores, to avoid groupthink. Quotas can help level the playing field; to get round the perception that the “beneficiaries” of the quota are under-qualified, first choose a long-list of candidates based on quality, and only then use the quota to increase diversity; everyone eventually chosen has passed the same quality threshold.

Some of the evidence shows possibly counter-intuitive effects. For example, having a “token” minority can backfire: the way our biassed brains work means that singletons will typically be judged by their group stereotype, not by their individual qualities. Including more than one person from the particular group allows each to be seen and judged more as an individual, rather than just as a representative of their class. This is the “critical mass” effect: a minority shouldn’t be present as less than one third, or three people, in total. This is an interesting approach. It implies that if you have a class of say 40 students, 30 men and 10 women, to be partitioned into teams of four, then it is much better to have five teams with two men and two women, and five teams all men, than to have 10 teams of three men and one woman.

There are many more relatively simple ideas for change here, from wording in job adverts to de-risking applications, from negotiation processes to stereotype threats, from the importance of role models to implementing transparent processes. And Bohnet is a strong advocate for the use of data to determine the presence and shape of the problem, and the using controlled experiments to determine the effectiveness of the interventions.

I have just summarised parts of the advice: Bohnet provides the rationale and the evidence. If you are serious about improving gender equality, and equality for other under-represented groups, then this is the book for you.