Deprecated: Hook custom_css_loaded is deprecated since version jetpack-13.5! Use WordPress Custom CSS instead. Jetpack no longer supports Custom CSS. Read the documentation to learn how to apply custom styles to your site: in /var/www/wordpress/wp-includes/functions.php on line 6078
Day 12: Communities and Survivorship Bias – Walking Contradiction

Day 12: Communities and Survivorship Bias

Communities, especially Open Source communities, tend to form some form of governance once they grow beyond a certain size. The actual size isn’t as important as the relationship among the members: when everyone knows everyone else, there’s really no need for governance. But when individuals come from different companies, or who otherwise may have different interests than the others, there needs to be some ground rules for making decisions on what does or does not get done. Without governance, projects will inevitably fork when these differences get large enough.

Typically governance is established by what most people involved like to think is a meritocracy: the hardest-working, most knowledgeable people are the ones who make the important decisions. At first glance this seems perfectly fair, and it usually is—initially, at least. Over time, though, this system is prone to the problems of Survivorship Bias. Let me illustrate how that happens.

Imagine a group of people who are on a long hike through the wilderness. There will be some people who have more skill reading a map, or operating a compass, or who know the terrain better. When the group starts out, it is only natural that these people lead the group, and they are given the title of Navigator. The group creates rules that while anyone can provide ideas as to what direction they should head in, only Navigators can make that choice. It works well for a while.

As time passes, though, and people in the group learn more about map reading and terrain features, their knowledge begins to approach the level of the existing Navigators. At that point it would seem fair to also designate these people as Navigators, since they now have enough knowledge to make directional decisions. But the rule is that the only way an existing group member can be designated a Navigator is if all of the existing Navigators agree. In other words, the process is largely subjective, as there is no objective test for competency. It also calls for a good deal of trust.

After a while, some people realize that while the Navigators have generally been doing their job well, they have made some errors that have taken the group off of the ideal path. Some group members point that out, and want to adjust course to get back to where they should have been. The Navigators, though, prefer to keep moving forward, even if it makes for a longer and more difficult hike in the long term; they prefer the feeling of moving ahead. Those who disagree go off on their own in frustration. Others within the group get the clear message that if they ever want to become a Navigator, they should curry favor with the existing Navigators. And when they do make it into that core group, they feel that they have worked hard to earn it, and anyone else who wants to reach that level has to play by the same rules that they did.

This is classic survivorship bias. The only people who can change the system are the ones who agreed with it in the first place, and thus don’t really see a problem with it. The voices of disagreement fade away until they can no longer be heard, so everyone thinks it’s all good. The system self-perpetuates.

I’ve seen this in action in several communities, but none so strikingly as in the OpenStack community, both on the Nova team as well as the Technical Committee that is supposed to provide technical leadership. I originally wrote a draft of this a year ago when I was working in that community, and became increasingly frustrated at how decisions were made. I happened to “run into” (electronically, of course) a few former Nova developers who had moved on, and when I expressed my frustration, they both said that similar feelings were why they looked to move to a different project. That’s when the role of survivorship bias became clear to me.

As I’m no longer in the OpenStack community, I don’t need to vent about particular issues or personalities. That’s history to me. I do hope that people realize that survivorship bias can shape how a community views itself, because if you are coming up short in some areas, you won’t know about it, because the people affected usually leave rather than deal with that BS. If you care about growing a healthy community, you need to make it easy and welcoming for people to share their ideas. And you should also take the time when someone who was active decides to leave to do a sort of exit interview. You might learn something important.

Leave a Reply