Iris Bohnet

Iris Bohnet applies ideas from behavioral design to gender bias in work, school, and politics. The book has a lot of good ideas and gave me a lot to think about. Overall though, I give it an average rating because the book's structure is not so great. If I were to want to try to use this in a practical setting, I feel like I would have to flip around for scattered insights that I vaguely remember.

The promise of behavioral design is that most people are not intentionally biased against women. However, we all use heuristics constantly to get through life efficiently. When those heuristics are applied in areas where gender biased behavior has been the norm, then they can lead to biased outcomes even without biased intentions.

As an aside: this book mostly discusses gender. Although behavioral design can also be applied to other types of bias, it is likely that the details vary enough that the case studies and research discussed in this book will not apply exactly. And even applying these ideas to gender, Bohnet cautions that behavioral design is always situational. Organizations should use their own data, perform their own experiments, and then encode their findings in processes and rules suited to their environment.

Unconscious bias is unavoidable. It applies any time a person is in a counterstereotypical role. How depends on the stereotypes that apply. For example, women commonly suffer from being perceived as either competent or likable. Men don't suffer from this particular bias because they are not stereotyped as likable the way that women are. In the face of unconscious bias, we cannot merely ask women to act against the bias. That can make some progress but often at a cost. For example, women who negotiate can get more, but the cost is that they will be perceived as less pleasant to work with.

We can't just ask people to stop being biased. It doesn't work. Indeed, sometimes awareness focused diversity trainings can make things worse by making people feel like they are making better decisions without actually changing anything. Instead, Bohnet recommends we follow a three step process: unfreeze-change-freeze. First, use awareness to "unfreeze" people -- open them to change. Next, perform experiments to figure out what changes work in a particular environment. Finally, "freeze" the insights from the experiments by changing environments and processes to make it easy for people to do the right thing.

One fruitful area for investment is the hiring process. Hiring is already a big investment for companies, and it's an area where relying on intuition can easily lead to bias. Bohnet's most important recommendation is that companies should use structured interview processes, both for the interview itself and while evaluating the interview feedback. Companies should also be aware of the gendered signals they may unintentionally be sending in their job descriptions and environment.

Once people are hired, it's important to make sure that the playing field is level for everyone. Women tend to have less tolerance for risk than men (although, as pointed out earlier, this may be due to the fact that risk distribution is not always gender neutral). Thus, it's important to make sure that biased outcomes are not being driven by unnecessary risk. For example, removing the penalty for incorrect SAT answers encouraged women to guess more. This is important because a willingness to guess was implicated in a portion of the SAT gender score gap. Another way to level the playing field is to design interventions that take into account how the intervention impacts the targeted group, the untargeted group, and overall. An intervention that helps one group at the expense of another is generally not worth it.

Role models are important. Ideally, these are role models within the organization. Lacking those, other role models from the field or even just role models who have overcome barriers in the past can help.

The way we craft groups is also important. Diverse groups have been found to perform better. Except when they don't. How do we interpret all of the nuanced and seemingly contradictory data? And how do we actually form groups that will perform well? Diversity improves performance by bringing multiple perspectives and problem solving approaches to a problem. A team with average skill and diversity can outperform a team of stars if those stars all bring the same strengths to the table. However, homogeneity decreases coordination costs. Combining these things, diversity can improve performance if the particular types of diversity are relevant to the problem domain and adds enough value to offset the increased coordination cost.

Building on this, having the right amount of diversity is important. A token diverse group member can be the worst of both worlds because that lone person can cause coordination costs to increase but because they are seen as a representative of their group, they are not necessarily valued for their individual skills. This is especially harmful for the token member. To get around this, groups need a critical mass of members of any group. There are no hard and fast rules but a good rule of thumb is that it will require at least three people and one-third representation is a good target to aim for. Because of this, Bohnet makes the provocative recommendation that in situations where there are only limited numbers of diverse group members, it is better to have a mix of groups that either achieve critical mass or are homogenous. This can achieve better outcomes than spreading out diverse group members evenly, resulting in more token representatives. Interestingly, the critical mass requirement is not always necessary. In groups that are so diverse that there is no one dominant group, tokenism has less of a negative effect. This provides a good reason, beyond the moral argument, for members of underrepresented groups to support each other.

Behaviors are shaped by the norms that a group or organization is subject too. Changing norms can be a powerful way to change outcomes. Setting rules can help change norms, but mainly in areas where there are not strong existing norms. In areas where there are already established norms, setting rules that try to shift those too dramatically can be worse than having no rules at all. Instead, it can be more effective to try to change norms via people's desire to imitate, compete, and gain social approval. For example, a government can emphasize that most corporate boards have some diversity rather than emphasizing that almost none have parity in gender representation. Transparency and soft defaults can also be effective tools. If people know where they stand, that can drive them to do better.

Overall, this book provides a large suite of pointers and tools. Bohnet does an excellent job of providing caveats so that her suggestions are not taken too literally. In that sense, it's very useful. However, beyond the very high level model of gathering data, running experiments, and then changing processes and the environments to point people toward the right behavior, the content is disconnected enough that it's hard to internalize the ideas.… (mais)