A guide to effective altruism fellowships

by jtm 10mo19th Jan 20196 comments

34


Introduction to the guide

This document is based on three semesters’ worth of experience running effective altruism fellowships at Yale University, though most of what I write here relates to the Fall 2018 semester, which was novel in several important ways. The document is intended to help groups run successful fellowships or similar projects. The document can either be read in its entirety or used as a encyclopedia. The appendices are comprehensive and are especially useful for group organizers. For any questions, don’t hesitate to reach out to joshua.monrad@yale.edu. If you wish to reach Yale Effective Altruism (YEA) in order to collaborate, get involved or inquire about anything, contact yaleeas@gmail.com.

Summary

The Yale Effective Altruism Fellowship is a semester-long program, in which 15-16 people are selected to learn a great deal about effective altruism over a few months. The fellows submit a written application and are interviewed prior to acceptance. The fellows meet once a week for dinner, where they discuss a range of EA topics. Before each meeting, the fellows are assigned between 30 and 45 minutes worth of readings, podcasts or videos. Dinner discussions are moderated by a group leader and usually do not involve professors. Beyond the dinner discussions, the YEA Fellowship involves an impact challenge, wherein the fellows are tasked with deciding where to donate a significant amount of money. The fellowship also includes a couple of workshops, social events, and optional events with speakers and professors. Fellows were also invited to take part in organizing our GWWC pledge drive.

Some keys to a successful fellowship seem to be:

  • Only admitting the right people, who are committed and genuinely interested
  • Setting high expectations for fellows
  • Making the fellowship seem and feel like a serious and well-organized program
  • Only involving professors who are exceptionally good speakers and/or speak on topics that are directly related and aligned with effective altruism ideas
  • Doing enough outreach that you can get a large amount of applicants
  • Planning well in advance of the fellowship

If you plan to do a massive Giving Game, it might be good to:

  • Focus on comparing organizations within one cause area (e.g. global health)
  • Let fellows choose from a predetermined list of candidates, instead of starting from scratch
  • Have one or more sessions of collaborative, in-depth research
  • Focus on teaching fellows to evaluate, rather than produce, research on effectiveness

In the 2018 Fall Fellowship, we accepted 16 out 65 applicants. We had an astonishingly high level of retention and attendance throughout the 12 weeks of weekly dinner discussions and workshops. Among them, our fellows spent, in total, more than fifty hours researching and deliberating as part of the impact challenge and ultimately decided to allocate the $3,000 donation (provided by CEA) to the Against Malaria Foundation. One of the fellows took the GWWC pledge during the fellowship, with several others indicating that they strongly considered it. Fellows’ views on topics in effective altruism updated in a positive direction during the fellowship. (See appendices for results of post-fellowship evaluation survey.)

Acknowledgments

Although I (Joshua) wrote this document and formally was in charge of our fellowship for the past couple of semesters, it would be both wrong and dishonest if I failed to recognize the hard work of other YEA board members; in particular, Frankie Andersen-Wood, Jessica McCurdy, and Mojmír Stehlík, each of whom have been essential in making the fellowship what it is. I should also express our deepest gratitude to CEA for their monetary and strategic support.

Goals of the fellowship

The fellowship was mainly aimed at people who did not have a lot of prior familiarity with effective altruism. For those, more experienced, people, we had other events and activities such as a weekly discussion group.

We had several goals with the fellowship. Looking back, I recall these being the main ones:

  • Bringing members into our “funnel.” The fellowship appears to be a great way of attracting potential core members and we hoped that at least a couple of fellows would go on to join our board and that even more would remain members of the general EA community on campus.
  • Direct impact of affecting fellows’ views. 12 weeks isn’t a lot when it comes to deeply affecting someone’s views, but we hoped that we could at least teach the basic tenets and concepts of effective altruism to almost all fellows and that this could affect donation and career decisions for a good portion of them. Moreover, we hoped that this might leave a huge impression on 1-2 fellows, who could at some point drastically change career paths as a result of being introduced to EA. We assumed that a large part of the expected value of the fellowship came from this possibility.
  • Strengthening YEA as a prominent group on campus. The fellowship acts as a “high-profile” activity, which can be advertised widely and which board members can highlight when asked “So, what does your group do?” Furthermore, the fellowship can become well-known on campus through word-of-mouth and attract attention to the group.

Recruitment of fellows

Reaching the right number and the right type of people will be essential to a good fellowship.

How to properly advertise the fellowship. One of the most important things I ever learned about running the fellowship is this: You get what you advertise for. Before the Spring 2018 semester, I would often advertise the program by saying things like “It’s not a huge time commitment, you just have to come for some dinners, and you’ll be eating anyway right?” and “It’s definitely NOT like an extra class!” My reasoning was that I wanted to bring in more applicants. The problem is that if you say that the fellowship doesn’t require a lot of commitment, you’ll get exactly what you asked for: people who aren’t very committed.

In Fall 2018, I instead said things like “We’re looking for people who will be really engaged” and I would make it clear that the fellowship would require 4-6 hours a week. This made it so that people who weren’t that excited would self-select away before the application process, making it much more likely for us to admit very engaged fellows. (At the same time, you don’t want to go too far. College students are busy and they need to know that the fellowship can fit into their weekly schedules.)

Channels for advertisement. We had a two-pronged strategy for recruitment: broad & targeted.

      • Broad recruitment: We aimed to make sure most first-years (at least) knew about the Fellowship, through a broad recruitment strategy. We hoped that well-thought-out messaging (accurate - not just focus on donations) could eliminate the risk of spreading low-fidelity conceptions of EA. The reasoning was that if we achieved this goal, hopefully only people who were truly interested in EA ideas would apply, and the selection process would be much easier. I believe that was the case.
    • Posters on campus (See Appendix 1 for Fall 2018 poster designed by Frankie Andersen-Wood)
    • Website
    • Facebook posts to people who ‘like’ our page, posts in relevant groups, & targeted ads
    • Fresher’s fair/extracurricular bazaar, handed out flyers in Appendix 1
    • E-mails to specific newsletters (especially to relevant departments and graduate schools such as economics, computer science, philosophy, school of management, etc.)
    • Word of mouth
      • Targeted recruitment: referrals. For fall 2018, we also made a “Refer-a-friend” form that we sent to former fellows, where they could input an email address and we’d contact their friend about applying. We aimed that all former fellows and board members are aware of the referral form and we would receive at least ten high quality referrals.

How large should a cohort of fellows be?

In Fall 2018, we opted for a cohort size of 16 fellows. This seems to be a really good size for discussions, allowing for talking time to everyone (if well moderated). In the past, we’ve tried with slightly larger cohorts (18-20), reasoning that some people inevitably would drop out and bring us to the ideal size of 15-16. However, if 20 people show up for the first discussion, it will be very crowded and leave people with a poor first impression of discussion quality. Moreover, if you succeed in selecting only very committed people (see Advertisement and Selection) and running a really good fellowship (see Running the fellowship), you can achieve high attendance and remain at the initial cohort size.

One or two cohorts?

If you have many applications, it is crucial to only accept the people who seem like they’re are very committed and a good fit for the program.

In earlier semesters, when we started receiving 40+ applications, we were reluctant to turn anyone away who had an interest in EA. Instead, we chose to accept almost all of them into two cohorts of 20 people. The downside to this was that, in each group, only half of them were really committed and/or a good fit for the fellowship, leading to declining attendance and quality of discussions.

In Fall 2018, we resisted the temptation to take everyone and decided instead to go for just one cohort. (Later on, we created a new cohort specifically for graduate students, which was a distinct and less extensive project, based on weekly discussions of an assigned reading.) This definitely seemed like the right move, as it allowed us to get a cohort entirely comprised of engaged fellows. Rejected applicants were still strongly encouraged to get involved with our community, though did so to a discouragingly low degree.

This isn’t to say that an EA group couldn’t run multiple fellowship cohorts at once, only that you shouldn’t accept every single applicant. It seems plausible that 30-40% of applicants (at Yale at least; rate may be higher or lower elsewhere) will be a good fit. So, if you wind up with 100 applications, you might have enough good candidates to fill two cohorts. Just make sure that you also have the capacity to run two programs, then.

Selection process

We received 65 applications for 16 spots. Once the deadline had passed, we (four people from the leadership of YEA) met. Each of us were assigned 25% of the applications, which we then read and graded according to selection criteria (see below). We blinded the names of the applicants to mitigate biases and conflicts of interest. After everyone had graded their portion, we ranked the applicants from highest to lowest score. Then, we went through from the bottom up and decided on the 20 least competitive applicants, whom we weren’t going to interview. We talked each applicant over, just to make sure that the selection wasn’t just a function of one grader having different standards than the others.

Once that was done, we invited about 40-45 people for 30-minute interviews. I would strongly encourage you to do interviews, even if you don’t have that many applicants. For one, it gives you a much better picture of the applicant pool and how much they would gain from the fellowship. Second, it sends a strong signal to prospective fellows that the fellowship is a serious thing with high standards.

For the interview, we asked them to read the Introduction to EA from CEA and asked them to “come prepared to share your reactions, disagreements or uncertainties.”

During the interview, we discussed the article and asked questions which we thought might help us rate them according to our selection criteria.

Note: Young people take interviews very seriously and may be a lot more nervous than you think they are. It’s incredibly important to be kind, patient, and encouraging. Intimidating them will be bad for them, for your group, and for the reputation of EA as a whole. Also, be mindful that people have different styles of thinking and talking, and it’s important not to be dismissive of those styles which are different than yours.

Selection criteria

When evaluating applicants for the Fall 2018 cohort, we rated them on the basis of a set of criteria that had also been outlined on our application:

    • Altruism: Passionate about helping others
    • Effectiveness: Ambitious in their altruism, with a drive to do as much good as they can. Potential to be aligned with the central tenets of EA.
    • Potential: Excited to dedicate their career to doing good or to donate a significant portion of their income to charity
    • Open-mindedness: Open-minded and flexible, eager to update their beliefs in response to persuasive evidence
    • Enthusiasm: Willing and able to commit ~3-4 hours per week
    • Fit: How good a fit are they with the fellowship format? Will they be good in discussions? Will they do good work for the Impact Challenge?

The questions we used to judge these criteria can be found in Appendix 11 (the Fellowship Application.)

We rated the applicants from 1-5 on each of these criteria. We discussed some of the applicants together and shared how we’d each rate them, to ensure that we had similar standards.

Ultimately, the selection process was very time intensive and extremely difficult. (It should be noted that our ability to interview 40+ was very much a product of the fact that we had someone working full-time, on a CEA grant, on running our group during this semester.) The end result was certainly acceptable and of all the 16 admitted fellows, there was only one person whom I maybe would have preferred to substitute for another applicant. Still, it is clear that there’s some room for improvement and I’d be happy to hear if someone knew of a more systematic approach to admissions.

Other important considerations: Diversity and homogeneity in levels of experience

Diversity

Beyond the above criteria, I believe it is important to consider two kinds of diversity when creating a fellowship cohort.

First, diversity of race, gender, and socioeconomic background should be seriously considered. The EA movement generally has a huge problem with diversity. Among about 1,000 respondents in the 2017 EA Survey, almost 90% identified as white; less than 1% identified as black; and 70% identified as male. The full causes, consequences, and remedies for this lack of diversity are serious but beyond the scope of this piece. Instead, I will briefly note that the community could stand to benefit a great deal from more diversity and that this should be kept in mind when putting together a cohort of fellows.

One (small) way of addressing is, for instance, to have non-male or non-white members of your group advertise the fellowship publicly. In our group, we are so fortunate to have several women in our leadership (as of Spring 2019, both Co-Presidents are female), and they did (and do) a great job at marketing the fellowship in a welcoming and inclusive way.

We also made sure to include an encouragement to apply, regardless of one’s identity, on the application page.

Another implication of this issue is to be careful about judging applicants too much by their prior familiarity with EA jargon or terminology. Some people may have great potential to become core members of the community and do great things, even though they have not yet been introduced to the “rationalist vernacular” or haven't happened to hear about Peter Singer when doing debate in high school. Aim to evaluate candidates for their potential and their passion, not their current level of familiarity with EA.

Second, it is a good idea to get a diverse range of academic backgrounds represented in the cohort, if possible. The discussions benefit a great deal from contributions from a variety of disciplines.

Homogeneity in levels of experience

On the other hand, there is one area where homogeneity is desirable: levels of familiarity and involvement with the EA movement. While some heterogeneity is OK, it will be a disadvantage for everyone considered if there are some fellows who are veterans in the EA community while others are total newcomers. Since our goal for the fellowship was to serve as an early stage of the “EA pipeline,” we erred on the side of getting newcomers. That said, we do also consider the fellowship a viable option for people who are somewhat familiar with the movement, but who haven’t yet had the chance to really study it, such as newly joined members of our board. We had a few applicants who were obviously overqualified, whom we encouraged to join our board instead of doing the fellowship.

Putting together a program and curriculum

Content of the Fellowship

The fall 2018 fellowship contained the following elements:

Mandatory

  • Weekly dinner discussions from 6pm-7.15pm (10 meetings)
  • A social initiation event on the first Friday evening of the fellowship
  • A 3-hour introductory workshop on EA concepts during the first weekend of the fellowship
  • A 3-hour Research-a-thon

Optional

  • Dinners with professors and guest speakers
  • YEA social events
  • Other YEA public events
  • A workshop on career strategies

Topics

In my view, the fellowship should give an introduction to several important topics in EA, but shouldn’t necessarily attempt to dive deep in to any one of them. In 2018 Fall, the discussion topics were as follows:

Discussion #1: Methods and Approach of Effective Altruism

Discussion #2: Effectiveness in Global Health and Development

Discussion #3: Impact Challenge I

Discussion #4: Impact Challenge Discussion II. Conclusion of Impact Challenge.

Discussion #5: The Philosophical, Empirical, and Historical Case for a Long Term Focus

Discussion #6: Global Catastrophic Risks from AI, Biosecurity, and More

Discussion #7: The Moral Case for Animals & Animal Welfare Interventions

Discussion #8: Career Choices and Effective Altruism

Discussion #9: Effective Altruism Community Building and Living As An Aspiring Effective Altruist

Discussion #10: Wrapping Up the Fellowship and Discussing Remaining Issues

Too little focus on the long term? You might notice that only 2 out of 10 topics are directly related to existential risks and “long-termism.” This might seem at odds with the “mainstream” view within EA. (Whether this really is the mainstream view, and whether it should be, are perfectly interesting questions that are beyond the scope of this piece.)

Here’s the reasoning behind this decision:

A lot of people in the EA community (including myself and other student group leaders as well as prominent people such as Will MacAskill and Toby Ord) began their involvement with the movement with a focus on global poverty and health. Gradually, many transition into caring more and more about the long term trajectory of humanity. I don’t think the order of this transition (global development then long-termism) is coincidental and I do think it’s extremely important to keep in mind when trying to bring new people into the community.

Let’s face it. Long-termism is not very intuitively compelling to most people when they first hear of it. Not only do you have to think in very consequentialist terms, you also have to be extremely committed to acting and prioritizing on the basis of fairly abstract philosophical arguments. In my view, that’s just not very appealing - sometimes even off-putting - if you’ve never even thought in terms of cost-effectiveness or total-view consequentialism before.

Accordingly, I believe that starting off with abstract arguments for future generations or apocalyptical risks from AI has a very high likelihood of turning people off from EA - people, who could come to support those very same causes down the line, if only they stuck around long enough to begin understanding and agreeing with the core principles that lead to those views.

On the other hand, by starting with global health and development - which often is compelling even to newcomers - people get used to the methodology and philosophical assumptions of effective altruism. Once you start thinking “Hm, maybe I should care about people, even though they are far away in space and maybe I should have a maximizing approach” then you’re one step closer to thinking the same thing about people far away in time as well.

In sum: I’m not saying that it shouldn’t be a priority to get people to care about long-termism and x-risks. But I believe pretty firmly that, for most people, the path to that point goes through more “conventional” cause areas such as global health and development.

Obviously, this is up for discussion and ultimately a judgement call. But I should add that my “model of EA-newcomer psychology” is based on not only my own experience and the anecdotal experience of EA thought leaders, but also on the experience of talking to tens, if not hundreds, of people about effective altruism and seeing what they respond well to.

Readings

I think it is reasonable to assign between 30-90 minutes of readings for each week, especially if you have a cohort of very engaged people. Readings should be informative and useful as talking points during discussion. They shouldn’t require too much former familiarity with a subject.

See Appendix 2 for the assignments for the Fall 2018 cohort.

Involvement of professors

For the Fall 2017 and Spring 2018 fellowships, I was convinced that it was essential to the fellowship that it included weekly dinners with professors. I had two main reasons for believing this, but I no longer believe in either of them.

First, I thought that well-known names, such as Paul Bloom and Laurie Santos, would attract more applicants. While this may be true, you can still obtain that effect without having to have professors for every dinner. Moreover - as discussed above - it’s not just a matter of getting more applicants. You want more applicants of the right kind - i.e. people who are interested in the ideas of EA, not just looking to meet famous professors.

Second, I thought that professors would make for more interesting conversations. This can be true, if it’s an amazing professor or if you have a poor group of fellows. However, not all professors are amazing and hopefully you have a great group of fellows.

But the biggest issue with professors is probably that it’s rare that their work is directly related to EA and sometimes their personal principles are at odds with core EA ideas. You might think that a professor in global health would be perfect for an EA discussion. But that won’t be the case if they focus on niche research in a niche (not-so-EA) area or perhaps haven’t done much work on cost-effectiveness, cause prioritization etc. This means that while conversations with the average professor might be super interesting, they will only be tangential to an education in EA. Most of the time, the fellows will learn much more about EA ideas if they have a free conversation moderated by someone who’s very familiar with the movement.

If you do want dinners with professors...

All of this said, professors can add tremendous value to your group. For this reason, we decided to have a series of “Dinners with Professors” parallel to the fellowship, which was open to board members and fellows. If you run these, here are some helpful tips.

Pick professors carefully and never bring a professor without meeting them in person first

Some professors are surprisingly awful at giving talks and leading discussions. Others seem like they do interesting work, but then it turns out that their opinions are antithetical to core effective altruism ideas. Few semesters ago, we managed to bring a professor who was an awfully dull speaker and was convinced that it was ridiculous to even be the slightest worried about AI alignment. To be sure, it can be could to have professors push back on mainstream EA thinking, but it’s not very useful if they don’t even agree with the very basic principles that make EA what it is.

Assign a reading or video

Discussions are usually better if people are on the same page in terms of the topic of discussion. Talk to the professor about choosing a good assignment that isn’t much longer than 20 minutes worth of reading/watching/listening.

Get participants to sign-up affirmatively

Students are busy and will flake from optional events if they feel like it. For this reason, it’s good to get people to commit to showing up. Usually, I first send out an RSVP form (using google forms or similar) to determine which 15-16 people get a spot for the dinner. Then, I send an email to those people, asking them to respond with an email confirming that they are able to come and that they will do the reading.

Beyond that, all you’ve got to do are the basic logistics:

  • Agree on a date that works for the professor. It can also be a good idea to fit it with the fellowship calendar. (E.g. in Fall 2018 we had a conversation with Shelly Kagan about ethics of future generations during the same week that this was the topic for the fellowship).
  • Book a dining room well in advance
  • Pick up the key (if applicable) for the dining room at the day of the dinner

Social events

One of the main purposes of the fellowship is to integrate new members into the core EA community at Yale. To this end, social events are pretty essential.

Initiation. As part of the fellowship, we had an “initiation” event on the first weekend of the program. This was framed as being “mandatory” and was pretty well attended. The initiation was a chill mixer taking place in a dorm common room. Beverages and snacks were served and we played a few games, including name games and ice-breaker games like “two truths and a lie.”

Other social events. Throughout the semester, we invited the fellows to all YEA social events but it was pretty rare that any fellows attended. In their evaluation survey responses, fellows seemed to suggest that actually wanted to be more of a part of the community, but just didn’t have the time. There’s some room for improvement on this front!

Workshops

As part of the fellowship, we had an introductory workshop on important EA concepts. The purpose was the introduce useful concepts and terminology to guide future discussions. The workshop was largely based on version made by Huw Thomas of Oxford EA, who deserves much credit for his work.

The workshop was fairly interactive, lasted 3 hours, and took place during the first week of the fellowship.

See Appendix 10a for workshop content and Appendix 10b for workshop slides.

Running the fellowship

Once a cohort of fellows has been selected, the work of actually running the fellowship begins.

Location

You need a recurring booking. Preferably somewhere central and somewhere that allows for a good conversation held by 15-18 people. At Yale, The Saybrook-Branford room and Branford Mendell Room are both excellent options. I prefer a room that is separate from the dining hall (unlike e.g. the Silliman dining annex), because this choice discourages getting up in the middle of the conversation to get more food, which is disruptive.

Sending weekly emails

Every week, I usually send out two emails. See Appendix #4 with all the emails sent out during the 2018 Fall Fellowship. If the fellowship dinner is on Wednesday, the first email should ideally be sent out during the weekend. The second email is a reminder email sent on the day of the dinner. An email usually includes:

Introduction to next week’s topic. How is it relevant to EA? What will we discuss and why?

Assignment. Include links to every assignment (double check that links work before sending) and a note about whether the reading is optional or mandatory.

Other information. Mention other group events of the week, sign-ups for special events, etc.

You should definitely use Mailchimp, as it makes for some nice-looking emails.

Running discussions

Make sure everyone has a chance to speak. Some people are more talkative; louder; better at “seizing the microphone”; and quicker at formulating their ideas than others. This means that you’ll sometimes have discussions where a few people are doing most of the talking. This is bad for two reasons. One, it can be less exciting and interactive for the people who don’t talk. Two, you get an unrepresentative sample of people’s opinions and people might think that no one believes in position X, just because no one happens to be arguing in favor for it.

If this happens, it might be useful to sometimes say “How about some of the people who haven’t really had the chance to talk yet - is there anything you’d like to add?” Another trick is to pay close attention to people who aren’t saying much. Sometimes, their face will reveal that they actually have something on their mind, in which case you can gently say “Laura, you look like you’re [nodding/disagreeing/whatever]. What’s on your mind?” Sometimes, they’ll chime in and sometimes they’ll politely say no.

Table unresolveable or tangential discussions. Every once in a while, the conversation will move towards really big picture questions like “capitalism versus other economic systems” or “radical versus moderate political platforms.” When this happens, I like to interject and say something like “This is an extremely interesting and important issue. However, I think we should table it for now and stick to the topic at hand. The conversation should definitely be continued outside the fellowship, though.” The main benefits of this approach are:

  • Keeping focus on the topic of the night
  • Making progress on narrowly defined questions instead of aimlessly discussing extremely difficult questions that could not possibly be resolved in the span of one hour
  • Avoiding heated ideological debates

Making sure everyone is on the same page. Every once in a while, someone will refer to an advanced concept, either from their coursework or from the reading, and go on to discuss that. In some of those cases, it might be a good idea to interject and say something like “Would you mind defining that term?” or “Can you remind us what her argument is?”, which allows other people to partake even if they don’t know the term or concept.

Be respectful of people’s opinions and interests. Many fellows will undoubtedly - as they should - care deeply about issues other than those commonly considered in EA. Of course, you want to present sustained arguments for priority EA-areas. However, being dismissive towards other causes and organizations will do you no benefit and can cause a great deal of harm. Not only will it make for a less enjoyable experience for the fellows but you also risk that fellows become defensive and even less likely to consider EA arguments. Even more troubling is the risk that you affect the reputation of EA negatively by making the community and philosophy appear hostile to and incompatible with a concern for other, important issues.

Educate, don’t argue. Occasionally, someone will present a claim that is somewhat at odds with mainstream thinking within EA. Often, I believe that the right course of action is to resist the temptation to refute their argument right away. You’re not there to prove how much you know about EA or to convince everyone that EA is 100% right on 100% of issues (as it most likely isn’t). Instead, you are there to offer insights and arguments to people for their consideration. Some people won’t change their mind on some issues, and it can be a waste of everyone’s time to try and convince them that they’re wrong. One of the best pieces of advice I ever got was to think of EA advocacy as being about “an offer of information, not about persuasion.”

Approaches to keeping attendance high

Arguably the most clear measure of success for the 2018 Fall fellowship was the very high attendance rate spanning 12 weeks of weekly dinner discussions and workshops. Most of the absences were “excused” in the sense that the fellow had contacted me prior to the dinner to let me know. So, what are the keys to ensuring high attendance?

Recruit the right people. I think a lot of the things that can be done to achieve high attendance happens before the first meeting even takes place. Attendance will start at a high baseline if you accept fellows who seem committed; who aren’t scared off by the expectation of spending 3-5 hours on the fellowship per week and who are genuinely interested in EA ideas.

Make it clear that the expectations are high. Again, “you get what you tolerate.” Resist the temptation to say things like “The readings are optional” or “It’s totally okay if you arrive a bit late.” Instead, emphasize that this is a fellowship with high standards.

We tell fellows that they must attend 80% of meetings in order to be “recognized as graduating fellows” by the end of the semester. (You don’t necessarily have to enforce this, but it’s a good idea to set the bar high.) At the beginning of each mandatory event, take attendance, noting if people are “Present, Excused absence (if they’ve notified you in advance), Absent without excuse, or Late”. I told them that I'd take attendance, and they must have seen that I pulled out my computer to do it, but I didn't inform them that I also noted if they were late. This was mostly to keep track so that I could nudge someone of they were consistently late.

Make the fellowship feel like a big deal. People are more likely to invest a lot of time and energy if they feel like they’re part of something serious. For this reason, it seems to have a good effect to do things that increase the feeling of being in a real program. Some things we did:

Webpage. Create a page on your student group website (e.g. the Yale EA site) where you post photos and information about the fellows. Instead of just having the fellows send in random photos, spend fifteen minutes before dinner discussion taking photos with a good quality camera.

Handbook. Print out a handbook with a calendar, a statement of the fellowship’s purpose, a list of norms and expectations for the fellows etc. See Appendix #5 for the Fall 2018 handbook.

Useful readings for discussion moderators

I believe anyone who moderates EA discussions should read these guidelines on depicting poverty, this piece on empathic communication of effective altruism and this piece on why & how to make progress with diversity and inclusion in EA.

Impact Challenge

As part of the fellowship, we launched our first-ever “Impact Challenge” (IC). This was essentially a massive Giving Game, spanning three weeks. During the IC, which was generously funded by the Centre for Effective Altruism, fellows were asked to choose one of four organizations to receive $3,000 in real cash. Ultimately, the fellows voted to donate the $3,000 to the Against Malaria Foundation.

This project was similar to the Oxford Prioritisation Project, though there were several significant differences. (The OPP’s director, Tom Sittler, wrote a nice review similar to this one.)

Content

The IC consisted of:

  • One dinner meeting spent going over useful concepts in health and development
  • One dinner meeting spent discussing initial impressions of the candidate organizations
  • A 3-hour “Research-a-thon” spent doing research of the organizations in small groups
  • An optional lunch with Mushfiq Mobarak, professor in development economics
  • One dinner meeting spent discussing and voting on the organizations

Goals

Before the semester, we formulated the following goals for the Impact Challenge:

  • Fellows learn the core concepts of Effective Altruism and charity prioritisation
  • Fellows become more dedicated to making Effective Altruism a big part of their lives, by transitioning our discussions from theory to practice
  • Fellows feel engaged and we see a high retention rate
  • The quality of the pool of applicants increases, as the fellowship appears even more serious and professional

Candidate organizations

Prior to beginning the IC, I selected five organizations from which the fellows had to choose:

  • Against Malaria Foundation
  • GiveDirectly
  • Evidence Action
  • Deworm the World Initiative
  • StrongMinds

I did this for a few reasons:

Keeping the focus on one cause area. On the face of it, one might think that fellows should compare organizations working in different cause areas, such as global health or AI safety. After all, one of the most important parts of effective altruism is to compare different such areas. However, comparing organizations in different areas poses several challenges.

For one, it is very difficult to make apples-to-apples comparisons of the pros and cons of organizations in different fields; especially considering the lack of quantitative methods in our IC.

Second, the choice between cause areas does often not come down to analysis but rather to moral and philosophical judgments. These judgments are extremely important and considering them is a crucial part of being an aspiring effective altruist. However, I do not believe that the IC is the ideal forum for discussing these philosophical issues. I suspect that people are much less likely to update their moral intuitions through a few hours discussion, than they are to update their empirical beliefs about what’s effective. For this reason, I think that within-area comparison would allow fellows to genuinely learn a lot through open-minded research, whereas I fear that between-area comparison might risk leading the fellows to argue for whichever cause area they felt most intuitively compelled by from the beginning.

Finally, I have an idea that a lot of people in the EA community began with an interest in global health and development and only later gained an interest for e.g. existential risks (as I discuss elsewhere.) On this view, it makes sense to avoid pitting cause areas against each other so early in people’s EA trajectory. We gave this a lot of thought prior to launching the IC and it also seemed to have been the best choice after the fact. If anyone wishes to discuss this topic further, feel free to reach out.

Ensuring a minimum impact. Since a considerable amount of money will be donated, I reckon it is desirable to eliminate the risk that the funds are donated to a poor choice. This also helps with applying for funds: you can guarantee the donor that the impact of their decision will be equal to the value of donating to a highly effective organization plus whatever the value of the IC is.

Focusing the impact challenge. It is tempting to think that the fellows should generate the list of candidates from scratch. However, this is an extremely difficult task and would likely be too much of a mouthful, unless the IC had been scaled up to span several weeks and to require a whole other level of commitment. Limiting the options allows for more in-depth research of each candidate and for higher quality discussions.

Voting

I sort of improvised a voting system, where each fellow would rank the organizations from 1 through 4. If an organization got ranked first, it got 1 point, if it was ranked second, it got 2 points, etc. In the end, the organization with the fewest points won. This system seemed to get the job done alright, but there’s likely a better solution. I encourage organizers to listen to the 80,000 Hours podcast interview with Aaron Hamlin on voting systems.

Some fellows expressed interest in being able to spread the donation across all four organizations or to at least have the allocation be proportional to the number of votes. While this is also an option, I felt that the winner-takes-all model was better at creating incentives for rigorous research and serious deliberation.

Research-a-thon

After the first dinner, the fellows met on a Saturday for a session of collective research. Due to a few people not being able to make it to the research-a-thon, I decided to drop one organization, so that the number of groups would fit better with the number of organizations. I chose to drop StrongMinds, since it was the one that I had the lowest prior confidence in and also was the one that had met the most skepticism in our initial discussion.

The attending fellows were split into groups of three and assigned two out of the four candidate organizations. (If someone fails to attend, you’ll have to improvise a bit with the group arrangement.) The organizations were assigned such that each was covered by two groups. Furthermore, each group was assigned one organization for which they had the responsibility of writing a research summary.

Each group spent the time as follows:

  • 5 minutes spent preparing to research their first organization and assigning research questions among each other;
  • 45 minutes spent researching their first organization
  • 10 minutes summarizing, discussing, and writing down their initial findings
  • 60 minutes rinsing and repeating for their second organization
  • 20 minutes comparing and discussing the two organizations
  • 40 minutes writing up a summary of findings for their second organization. See Appendix 6 for example of research summaries.

Although this structure worked fairly well, it was pretty much improvised on the day of the session, and it is possible that another structure could work just as well or better.

Evaluating the organizations

You could imagine an Impact Challenge where the goal was to have extremely high standards for rigorous decision-making, in accordance with EA methods and values. Such a project would likely employ quantitative models, such as the ones used by GiveWell or in the Oxford Prioritisation Project. However, this would require considerably more time and it would require methods that are unfamiliar to many fellows. One of Tom Sittler’s key lessons from the Oxford Project seems to be that too high ambitions for the evaluation process can get in the way of a successful project and I agree with this assessment.

Moreover, it could be argued that the benefit of having such rigorous methods might actually not be that great. To be sure, we want fellows to recognize that charity evaluations require rigorous methods. However, we are not so much aiming to train people in evaluation as we are hoping to teach the importance of doing such evaluation in a good way. To this end, exposing fellows to the rigorous literature of GiveWell seems perfectly adequate.

Instead of quantitative methods, we relied on much less rigorous, more qualitative assessments based on literature review and discussion. This approach seemed justified in part by the fact that we were choosing from GiveWell-recommended organizations, which gave us confidence that each alternative was largely in the same ballpark in terms of impact.

We tried to include a somewhat quantitative (though non-rigorous) component by explicitly asking “What does this organization achieve with an additional $3,000?” but the discussion was mostly concentrated on more general pros and cons of each organization.

Evaluation of Fellowship

Before and after the fellowship, we had fellows complete surveys. This had the dual aim of measuring our impact and learning ways to improve the fellowship.

See Appendix 7a-7c for survey design. See Appendix 8 for survey results. (This appendix is fairly comprehensive and I encourage readers to take a look.)

Overall, based on my own subjective experience and a qualitative assessment of the survey results, I’d say that the fellowship was a success; that it was an enjoyable experience for everyone involved; and that it was net-positive in terms impact - even if we consider the opportunity cost of the 100+ hours I spent organizing and running it.

In terms of the goals set out earlier in this document:

Bringing members into our funnel. This goal was not achieved as well as I had liked. As I write this, it’s a bit too early to tell whether alumni from the fellowship will be involved with our group going forward. My guess is that at least 30% of fellows will go on to attend our events and that at least 1-2 of them will join our board. Still, interest in joining the board hasn’t been appeared to be quite as strong as I had hoped, so there is room for improvement.

Direct impact of affecting fellows’ views. This goal was achieved. Judging by the survey results as well as the general change in tone and viewpoints throughout the term, I believe this goal was achieved. We didn’t have appear to affect many radical changes of personal philosophy or career plans - though there were at least a couple of people who began reconsidering their paths as a results of the fellowship - but I would also think that this is generally difficult to achieve in just one semester.

Strengthening YEA as a prominent group on campus. This goal was achieved. The fellowship continues to be our flagship activity on campus. It helps us reach more people and contributes to the feeling among board members that we are doing significant work with clear outcomes.

Beyond the fellowship

Fellows who have participated to a satisfactory degree (in the handbook, this was 80% attendance, though in practice we tolerated slightly lower rates than that) were given a printed certificate. See Appendix 9 for example certificate (designed by Frankie Andersen-Wood). Fellows were also strongly encouraged to join our group’s board and to stay involved with our activities on campus.

Appendices

These appendices are not intended to be read from A-Z. Rather, they might be useful to answer particular questions like “What readings should I assign this week?” or “What are some talking points I can draw on in case tonight’s discussion comes to a halt?”

Appendix 1: Poster for recruitment (On request, we can also send you the canva template to edit)

Appendix 2: Reading list

Appendix 3: Discussion talking points

Appendix 4: Emails sent to fellows

Appendix 5: Handbook

Appendix 6: Research summaries examples

Appendix 7a: Pre-fellowship survey design

Appendix 7b: Post-fellowship survey design

Appendix 7c: Notes on survey design

Appendix 8: Survey results

Appendix 9: Example certificate (On request, we can also send you the canva template to edit)

Appendix 10a: Intro workshop content

Appendix 10b: Intro workshop slides

Appendix 11: Fellowship Application