TLDR: Splitting your 8-week fellowship into 2 4-week parts, and marketing only the first half, potentially allows more people to sign up. Motivated participants can opt-in for the second half, unmotivated ones can drop out, and busy ones can do it later. Potential drawbacks are discussed, such as risks of decreased retention and epistemic erosion.
Why and How
If you haven’t heard about EA or AI safety, signing up for a fellowship might be a big commitment. You will have to spend the next 8 weeks reading about and discussing something you are not sure you will like, with people you are not sure you will like. At the beginning of every semester, university groups are hustling to get enough fellowship applicants, and marketing is no fun.
I don’t think splitting your program is a silver bullet to this, but I will describe why I think it might be useful to do, especially if someone is mindful of the potential drawbacks.
By splitting your fellowship I mean that you split your program into two halves, and only market the first half so people need to commit less before they become familiar with EA/AIS. People who have time and like the program can continue right after[1] 4 weeks with the second half of the program.
Many of the pros of this approach are already outlined in Lisanne’s post on “Why a 4-week fellowship is better than an 8-week fellowship”, (which is a great post and is what largely got me to start thinking about this!), while trying to mitigate the drawbacks of people learning less about EA/AIS due to attending a 4-week course instead of an 8-week one.
Below I will outline how I see the pros and cons of this approach and what I think happened in our case. Note that whether this is preferable for you will to some extent depend on your own theory of how CB should work, so I recommend evaluating each item for yourself. To help with this, I also added what I see the cruxes are. I welcome you to write a comment if your cruxes are different!
(The PROs will be outlined with numbers while the CONs with letters to make it easier to refer to them)
The PROs of doing this:
1. A shorter course allows for more signups
Crux 1: Does it really?
The obvious crux here is whether it indeed increases signups and completion rates. We have some evidence from PISE and EA Utrecht that marketing a shorter course indeed boosts application numbers. My subjective experience with our group is that this has helped us out as well, but not to a crazy extent. It would be pretty hard to measure this in a reliable way. My best guess is that it can increase applications by around 10-20%, but I would be interested in other people’s experiences.
2. Highly motivated people can still take the whole course
For the people who really like the ideas, things can just go business as usual and they can finish the 8-week course. If somebody liked the course but can’t commit to the second half, then they can finish it in another fellowship round. (All of this sounds very simple in theory, but it’s not. See the discussion of cons below.)
3. Allows you to run intensive fellowships for university students
By intensive fellowship, I mean that instead of having 4 sessions stretched out over 4 weeks, you can blitz them out in just one week before the semester starts or during spring/fall break.
We only tried this twice (once with our EA and once with the AIS course), so take this with a grain of salt, but I really like this so far, and my rough sense is that this also boosts fellowship applications, perhaps because people can more easily commit to freeing up time for a given week, as opposed to planning ahead for 4-8 weeks.
I plan to write about intensive fellowships in more detail later, but for now, here is a big caveat:
Intensive fellowships are intensive indeed! People told us that they were pretty exhausted by the end, so I think it is better to have intensive fellowships if your course allows for in-session readings, as opposed to people reading at home.[2]
The CONs of doing this:
A) Decreased retention?
Crux A.1: Of those finishing the 1st 4 sessions of the fellowship, how many will continue and finish the second half?[3]
Well, this is where things get tricky. In an ideal world, people who seem very motivated but can’t commit to the second half would come back in the next semester to finish. I’m not sure if they do though.
Here is the data I have so far:
Just please take it with a big grain of salt, as it is not the highest quality. See footnotes for the caveats.
I found that out of 82 people who finished the first 4 sessions of the fellowship, 51 went on to finish the 2nd 4 sessions, resulting in a conversion rate of ~62%.[4] Going forward I want to do a much better job of keeping track of the exact numbers.[5] Based on my current impression, I expect a conversion rate of 65-85% for fellowships that run weekly from the beginning, and 40-70% for those that start with an intensive week and then switch to weekly sessions. I will make sure to update this section once I have more and better quality data!
Crux A.2: What happens to those who don't continue right away? Will they finish the fellowship at a later round?
Unfortunately, we didn't do a good enough job of keeping close track of this, but my rough impression is that while a few people do, most don't, even if they seemed quite excited about EA in the first half of the program. I think one explanation of this is planning fallacy, ie. people thinking "I'm very busy right now, but next semester I will make sure to finish the second half" - and then they are even busier in the next semester. For this reason, I think it is very important to personally follow up with people who you are very excited about, and try to include them in the community even if they did only the first half of the program.
B) More operations work
Running weekly fellowships is already plenty of work in themselves, and once you get complicating things it’s easy for things to be forgotten and fall through the cracks.
Here are a few things you need to keep in mind (and I often failed to) in addition to what you already have to do in coordinating fellowships:
1. If you have been running 8-week fellowships before, you have to make sure all facilitators clearly understand the “new system” - meaning that the participants are advertised the 4-week course.
On your website, you should already mention that motivated participants can attend additional sessions, and the facilitator should start encouraging people to sign up for those around weeks 2-3. The way we did this is that at week 4, we had people fill out an “opportunities form”[6] at the beginning of the session, where they can opt-in for the additional other 4 sessions, sign up for newsletters, ask to be added to Facebook chats etc.
The idea here is that even if they can’t continue right away, we keep them in the loop and introduce them to the community so they are more likely to finish the fellowship and join later events.
The two failure modes here for us were some facilitators not being aware that we changed to this split version, so during the early discussions they were already referring to sessions between 5-8 weeks which confused people.
The other one is the facilitator not mentioning soon enough that there will be an option to sign up for sessions 5-8, so people don’t have time to decide if they want to commit or not.
Another thing I would recommend that in communicating about the additional sessions, you should make it clear that while it is optional to join, if someone decides to do so, they are “expected to” attend all 4 sessions, in a similar way we expect them to take part in sessions 1-4. The failure mode here is that if you don’t communicate this clearly, then people will think that given that this second half is optional then any of the sessions within it are optional too, so they have a lower bar for missing a discussion or would only attend the topics they are particularly interested in.
2. You have to keep track of who finished the first half only, and who finished the whole course.
This is easy in theory, but in the middle of coordinating your fellowship, you already have so much stuff to pay attention to, that this is easy to forget. The way we try to do this is that after every session, I ask facilitators to fill out a reflection form,[7] which among other things asks about attendance.
When you are marketing the in-depth fellowship to your group, you should also include a question that asks if they have done the full intro fellowship or just the first half. If the latter, I would still have them for the in-depth course, but make sure they read up on sessions 5-8 before or during the in-depth course (ideally with someone having a 30-minute 1-1 with them about each topic)
Additionally, one should also make sure to note down when a given person finishes part 1 and part 2, so it's possible to keep track of people who do the first half but don't continue right away.
C) Epistemic erosion?
Worlds in which I don’t see this approach working well are the following:
1) If most people sign up for the first half of the course, but don’t continue because they are busy and think that next semester they will have more time (which they won’t, classic planning fallacy, eh.). Even if they don’t disconnect from the group altogether, they will attend events with less knowledge than someone who has done the full fellowship, which might affect the quality of discussions during meetups.
2) People who have only done the first half of the fellowship will apply to conferences and will be more likely to get rejected, which discourages them from learning more. Alternatively, they might get into EAGx conferences without fully understanding what EA is. About the latter issue, I don’t necessarily think this is a bad thing in itself. As I understand EAGx admissions are more hit-based than EAG ones, and this way talented individuals would get to see that “EA is a real thing”, not just a group of students reading about philosophy.[8]
Concluding thoughts
Apart from one’s group’s operations capacity, whether one thinks this approach is desirable will likely depend on how narrow vs. broad one thinks EA community building should be, as well as whether one is willing to potentially trade off the median local group member’s knowledge against increased fellowship signups.
If you think that most of your group’s impact will come from a few members who become very committed to EA and pursue EA careers, then I would think you should try to make your course as accessible as you can,[9] so that more people sign up and make sure you give good support to the most promising attendees. Needless to say, one could argue that the people who will become very engaged would have signed up for an 8-week fellowship anyway, I really don't know!
Overall I think it was worth experimenting with this for us and I will likely continue this approach due to the increased flexibility and more signups. If you would like to try this for your group, feel free to reach out to me, however, note that you should first definitely check in with your mentor! (Learn who is the CEA contact person for your group, or request help, here.)
- ^
Or you can have a week-long gap, see the section about intensive fellowships. If you can though, I recommend continuing right after to keep the momentum going!
- ^
My sense is that most AIS groups do in-session readings already, so adapting it for an intensive fellowship should be pretty easy. We recently piloted a zero homework /in-session reading EA fellowship, which I plan to publish about soon, but for now, I will say that I'm much less excited about having EA fellowships with in-session readings compared to AIS ones.
- ^
The exact wording is important here! Eg if you are looking at "what % of people will finish part 2 out of those who apply", you will get a lower conversion rate, as some people drop out even before the fellowship starts
- ^
Note that we had very big differences for different programs, with the lowest conversion rate being 30% for an intensive fellowship and the highest being 83% for a weekly fellowship).
You can look at the raw data here, see the green and orange sections of the Excel. - ^
Going forward we will collect these metrics, if you plan to split your fellowships please consider doing the same and sending me your data! :)
Applied Found a group Started part 1 (min. 1 session attended) Finished part 1 Conversion rate of starting starting part 1 and finishing part 1 Started part 2 Of this, joining from the current fellowship were Of this, joining from a previous fellowship were Finished part 2 Of this, joining from the current fellowship were Of this, joining from a previous fellowship were Conversion rate of starting part 2 and finishing part 2 Conversion rate of finsihing part 1 to finishing part 2 Conversion rate of starting part 1 and finishing part 2 Conversation rate of applying to finishing part 2 - ^
If you would like, you can make a copy of either the EA Hungary or Budapest AI Safety opportunity forms.
- ^
I don't know who we originally got the idea from, but I think this is a great practice to help keep your ear on the ground, especially if the coordinator is not facilitating fellowships themselves. You can use our template if you want.
- ^
Note that my personal experience might overly influence me, as we had 2 students who did the first part of our AIS course but couldnt commit to the second half. However, they got into EAGxPoland which motivated them to learn about EA and start their own EA group at their university.
- ^
This may sound paradoxical as narrow EA usually refers to having fewer people in groups/programs. As long as you have the capacity though, I would let everyone in the program who meaningfully fills out the application form, as it might be hard to predict who is going to be very motivated in the end.