This post isn't intended to be anything stunningly original, but simply to collect a few basic, but under-utilized observations together in one place.

The question of how we can progress human knowledge and understanding is one that is of particular interest to people on this website. The most obvious reason why groups are important for these purposes is that often a lot of work is required to move a field forward and when more people are working on a problem, you might expect that more work would get done[1]. There are other reasons too, but this the main one.

This immediately leads us to the question of what factors are important for allowing a group to produce knowledge:

  • Focus area: A group's focus area can be broad or it can be narrow. As already noted, a broad focus area can dilute the group's common knowledge. If a group's focus is too narrow, then there will be too few people interested in joining or participating in the group. If the group's focus is too broad, then it'll be less likely to attract people with expertise in a particular area or to develop such talent. If you want to start a new group, picking the right focus area is crucial.
  • Common knowledge: This is knowledge that basically everyone in a group knows, apart from beginners. This is important as being able to take certain things for granted allows discussions to cover more advanced concepts or for people to explore further down the conversational tree. Unfortunately, only so much knowledge can be common knowledge and this is especially limited when groups have no way to insist that membership is conditional on completing a certain course or program. The broader a group's focus, the more shallow the group's common knowledge will be, but this can be less of a disadvantage if there are synergies among areas. Common knowledge can also backfire if it results in members becoming overly ideological or unknowingly making assumptions.
  • Norms: It's often valuable for groups to develop their own norms. I expect rationalists to have a tendency to search for a set of perfectly rational group norms, but I suspect that the ideal norms vary heavily depending on the specific group. For example, if you want to know the truth, it's often better for people to be blunter, but this can also be quite off-putting for beginners. The solution to competing access needs is often to have multiple groups with different norms[2]. One key factor when deciding norms is to remember that a group cannot coordinate on overly complex norms.
  • Self-selection: This is often a stronger force than you might expect. Groups don't just attract people with shared interests, but those who align with a group's values. Given how hard it is to change people's attributes and the unpleasantness associated with kicking someone out of a group, self-selection is often a crucial mechanism in allowing a group to be functional.
  • Origin: It's worth noting that a group often arises in reaction to another group. This can be advantageous as frustration with another group can often be quite motivating. On the other hand, the more that a group is a reaction to another group, the more likely it is to overreact and toss out the baby with the bathwater. One key challenge here is that the more a group adopts nuance, the harder it is for a group to differentiate itself from other groups. And without a clear message, it's hard for a group to attract members.
  • Subgroups: A group can form a launching pad for a subgroup to form. Subgroups typically form to focus on a specific area. They can also filter more heavily on particular attributes: for example, Bob may only invite people with consistently good takes to his discussion group. Since progress is often tail-heavy, the potential for members to form their own sub-groups is often where most of the impact will be coming from.

Anyway, I hope that this post helps you think about what kinds of groups ought to exist.

  1. ^

    However, there can be diseconomies of scale where it takes more time to follow all of the work that is being done and more time to filter out the good work from the bad.

  2. ^

    As an example, Effective Altruism tries harder to be friendlier as recruiting more members arguably allows it to increase its impact. On the other hand, the Less Wrong community is less desirous of growth, lest it lower intellectual standards.

New to LessWrong?

New Comment