jml's notebook

Notes: Rebel Ideas

A friend gave me Rebel Ideas by Matthew Syed for Christmas, saying that he liked the pictures. That's actually a pretty good review.

Rebel Ideas is about diversity, but not in the way that I'm used to reading about diversity. The book leads with the thesis that diverse teams will perform better, and that the dichotomy between hiring the best and hiring for diversity is a false one. So far so boring. All of this has been accepted wisdom in my circles for years now.

However, the book takes an angle that I hadn't thought of, one that's incredibly obvious in hindsight and doesn't need a whole book to develop.

You see, I thought that the false dichotomy between talent and diversity came down to biased tests. If you have a take home assignment in your interview process, you bias against people with family commitments (statistically, women). If you have algorithms questions in your interview, you bias toward recent graduates and folk who've had cushy enough jobs that they can practice toy problems instead of building things for users. All of this is true, but in Rebel Ideas it is very much an aside.

No, the real reason talent & diversity are not odds is because the problems most of us face are so complex that even the most talent or insightful person can only see a tiny part of the problem space. If we want to see more of the problem space, we need people with different perspectives. Much of the time, these different perspectives can be at least partly a function of different demographic diversity, but they don't have to be.

All of this chimes well with my existing prejudices. For example, the book Hard Facts points out that the evidence suggests individual-based performance compensation can harm team-level outcomes. When I raise this with people in my network, I'm regularly surprised by how negatively and strongly they react to the idea.

The book then elaborates more on how this diversity can help, how to stop groups from becoming less diverse via group think, and so forth.

Two ideas are worth highlighting, and then a little personal reflection.

The first is the difference between dominance and prestige hierarchies. Stereotypically, your boss is above you in the dominance hierarchy, but the engineer two desks away who everyone asks for code review is above you in the prestige hierarchy. People behave differently when the different hierarchy modes predominate. Syed suggests that a prestige hierarchy is better for figuring out what to do and a dominance hierarchy is better for making sure the thing gets done. It's an interesting idea, and one that speaks to the tension that runs through engineering management: when is it helpful to tell smart, capable people what to do?

The second is that diverse groups feel worse. Syed reports on an experiment where subjects were asked to solve a complicated murder mystery in teams. The first group of teams, A, were made up of four friends. The second group of teams, B, were made up of three friends and a stranger.

The teams in A had a much more pleasant experience than the teams in B, and they were much more confident of their answers. The people in the B teams reported having difficult discussions, a challenging experience, and were much less confident of their answers. However, the B teams were also right way more often than the A teams.

If you think about this, it isn't much of a surprise. It's satisfying for people to understand your idioms and pleasing to be told you how right you are. Contrariwise, the more foreign ideas you're exposed to, the more you become conscious of just how little you really know.

Most of us have probably had the experience of being in a difficult meeting about some contentious topic, and then leaving the meeting only to carry on discussing that topic with some of our more like-minded colleagues. If only the others would just see! But the thing they don't understand is…! It's humbling to think of this as a form of blindness or narrow-mindedness.

Reflecting personally, there are things I very highly value that aren't universally valued. Off the top of my head: intellectual rigour, technical expertise, action-oriented discussions. Conversations with people--especially those in other functions--can be difficult because they value other things more highly. If I want us as a group to be able to solve more complex problems, I will have to be more ready to accommodate differences of approach, and will probably have to exercise more empathy, while at the same time holding on to the things I value.