We can identify two different approaches to community building. The micro approach involves deeply engaging with individuals in order to change their views or to assist in their journey. When adopting this approach, programs tend to be flexible, with guidelines rather than rules and decisions sometimes being driven by the needs of individuals. This approach puts participants at the center.
The macro approach involves a level of standardization in order to ensure consistency and to keep the program manageable. While exceptions may be made to rules there is more of an effort to limit these exceptions in order to limit deliberation time if nothing else. Individual accommodations will likely need to be limited given that organizers only have so much time. This approach is typically is centered more around the needs of organizers.
Here are some examples of decisions that would make more sense from a micro approach than a macro approach
- Asking people who have expressed interest in the ideas you're sharing whether they want to grab a coffee. Or offering to have a call with anyone who wants to know more about a program.
- Moving lunch an hour earlier than originally proposed so a participant can make a lecture.
- Offering to cover a participant's plane ticket personally in order to encourage them to attend a program.
- Choosing mentors based on individual interests rather than just choosing a group of generally competent mentors.
- Individually helping participants who miss part of a program catch up on key the points they may have missed.
- In response to a participant expressing a keen interest in reinforcement learning and asking if there will be any talks on the subject, saying that you can't make any guarantees, but promising to try your best to find someone.
- Increasing the funding available to all participants in order to sway one who is on the fence, whilst also being fair to the others.
- Learning each person's individual interests and providing personalised recommendations of things to check out or people connect with.
Here are some examples of decisions that would make sense from a macro approach:
- Requiring participants to attend 6 out of 8 sessions in order to graduate from a program, strictly limiting exceptions to medical reasons.
- Fixing a formula for the amount of funding provided to each participant.
- Politely refusing to reschedule a session because rescheduling take up some amount of time/attention from other participants and some amount of organiser time even if everyone could make it.
A natural question is to ask in what circumstances would one approach make more sense than another:
- The micro approach is quite naturally better suited towards smaller groups, but it may also make sense for larger groups when there are only a few participants per organiser. However, even when there is a good organiser ratio, it will often still make sense to lean more towards macro techniques for large groups in order to standardise how people are treated across sub-groups. Further, the more participants, the more likely it is for individual judgment to result in glaring inconsistencies.
- Micro techniques are particularly useful when trying to build up a critical mass as you can use techniques that won't scale. For example, when planning the Sydney AI Safety Fellowship, I felt that we really needed at least four fellows. Below this number, I felt that the fellowship wouldn't be providing much value in terms of networking or peer learning. I was also worried that if we didn't hit this number, the fellowship would be perceived as a failure and this would discourage people from applying or volunteering to help next year.
- You need to be careful not to overdo micro as you can come off as weird or otherwise make people uncomfortable as they don't want to put anyone out.
- Micro vs. macro doesn't completely capture the divide as formality is important too. The more that decisions are made on the fly, the less a program is seen as formal/official/credible. This can be important both for attracting participants, but also if part of the purpose of the program is to raise the status of participants. However, these considerations aren't completely orthogonal in that people generally expect a greater degree of professionalism from larger programs.
- Another reason apart from size why a program might lean more towards the macro end is to encourage a higher level of commitment in participants. While not universally true, generally a program that provides limited flexibility implicitly asserts its own importance.
With the Sydney AI Safety Fellowship, I initially leaned more towards the macro end of the scale in order to establish the baseline at higher level of commitment. I then shifted further towards the micro end of the scale as I realised that this would be necessary to hit critical mass. My plan for next year is to emphasize flexibility, but if we start receiving many more credible applications than slots then I would aim to increase the expected commitment for future years.
My expectation is that community building will generally involve more focus on the micro early in the game, but over time if the community grows then it'll be important to shift more towards macro-style strategies.
Obviously, it's more of a spectrum.
Not with every individual as different individuals will desire different levels of engagement, but engaging deeply with people who would appreciate and benefit from such engagement.
Some of these example are based on things I did or wish I had done in relation to the Sydney AI Safety Fellowship; others are purely hypothetical.
This would probably be a bad idea if you thought the other participants would hate it/get nothing out of it, but let's assume you think they'd gain some benefit too.
This may make sense if you have access to additional funding beyond what the program will cost, where you may have originally have want to avoid spending if it wasn't required.
There are some very specific circumstances where the perception of success or failure can be even more important than actual success or failure. For example, for the Sydney AI Safety Fellowship because the idea came to me quite late, by the time we'd secured funding many potential fellows had already committed to other opportunities, so it was hard to find participants. It was quite plausible to me that the program could fail in a way that wouldn't provide strong evidence about whether the program would be viable in future years and that such a failure could result in an attempt to run the program next year from also failing. I'm not suggesting that people should engage in deceptive behavior or dishonesty in these circumstances, just that in these circumstances it is even more crucial to achieve at least a moderate level of success and to do so visibly.