EA Geneva’s fellowship: a fellowship model for non-university groups

In this post, I vary between the use of ‘I’ and ‘we’. When I use ‘I’, I’m referring to myself, Naomi, the author of this post. When I use ‘we’, I refer to EA Geneva’s community builders (Konrad and Naomi) as we developed and executed the program together.

With this post, I would like to share our experience with EA Geneva’s new fellowship during the first half of 2020. The aim is that others can learn from our experiences and mistakes.

We started with the fellowship in February 2020, and even though we will continue with the program for at least half a year longer, we are doing our first evaluation of the program this summer (2020) and I thought that sharing our experience early might benefit others who are considering a similar program.

I hope that the reader will keep in mind that I’m sharing our first conclusions. Since the program is only running for half a year (and is influenced by the social dynamics that stem from the Covid-19 pandemic), these are only anecdotal data and could be highly dependent on factors such as the composition of the group and the quality of the one-on-ones with the group leaders.

Introduction

At the end of 2019, EA Geneva ran a series of strategy sessions to (re)formulate our theory of impact, evaluate our previous work and assess which of our ideas for potential future products could be impactful. After creating an MCDA, where we ranked our current activities and possible prospective projects, a fellowship program scored among our highest ranking products.

The idea of the fellowship came from a wish to encourage community members to follow through on their insights more often and an observation that people are sometimes confused about how to apply effective altruism concepts. We wanted to better support motivated members on their journey towards doing as much good as possible and hoped to address the mentioned bottlenecks with the fellowship program. Part of the fellowship’s objective is to learn about effective altruism, as understanding ‘how to do as much good as possible’ is a crucial aspect of having an impact.

In order to accelerate the growth of the fellows, we grouped the majority of our activities into a coherent and continuous program, which created a subgroup of highly engaged community members. We thought that integrating driven newcomers into this sub-community could facilitate a steep learning curve for them. To focus the fellows’ efforts, and to facilitate coordination among the participants, every fellowship member formulates a goal that is evaluated after approximately three months.

By formulating a goal we expected that:

Fellows develop themselves quicker;
Community members can hold each other accountable for their progress because it becomes easier to coordinate among this sub-group of the community;
Common projects can be identified more easily; and
We can organize invite-only meetups and other events more in line with the interests of the most promising community members.

Description of the fellowship

The main characteristics of the fellowship are:

The formulation of a fellowship goal with a deadline at most 3-6 months into the future, consisting of a goal-oriented element and a continuous learning element on topics related to effective altruism.
Regular 1-1’s with one of the community builders (approximately once every two weeks or once every month, depending on the fellow), where we discussed the progress on the goal, addressed possible bottlenecks, shared relevant resources, and provided feedback.
Optional fortnightly fellowship evenings, consisting of a mix of workshops, presentations, discussions, and collective work (e.g. collective ABZ planning session). We started the evenings with an hour of coworking. As the group sessions started in March 2020, we have had only virtual meetings so far.
EA Geneva’s fellowship is a continuous program. Together with the fellow, we regularly (~every three months) evaluate the progress on the fellow’s goals, and decide whether it is valuable to continue with the program. It is possible for a fellow to stay in the program for a long period of time (e.g. while working on an ambitious and potentially impactful project).

The public description of the fellowship:

“Our fellowship is designed to support you on your journey as an aspiring effective altruist. We highly encourage anyone to apply who is committed to doing as much good as possible and who aims to continuously learn more about how best to do so. Our personalized and result-oriented support helps you make progress, while the other fellows provide you with an inspiring and challenging group of peers.

The program consists of regular one-on-one meetings and optional-but-highly-encouraged “fellow sessions” every two weeks. These sessions include co-working, knowledge exchanges, and discussions.

As a fellow, you work on your personal goal and learn more about effective altruism. We help you to clearly define your goal(s) at the start of the program. Your goal can be anything related to EA, from planning your career to answering a research question, organizing an event, figuring out where to donate, skilling up, developing a project idea, or simply to gain more insight into effective altruism.

Your goal is defined for maximally six months, usually two to three. How much time you invest depends on your availability. Being able to invest at least two hours per week has helped others make regular progress.

The fellowship is free of charge, does not entail financial support, and is open to anyone who is concerned with the question ‘how can I do the most good?’ No previous knowledge of effective altruism is required.”

Depending on the audience we adapted this message.

We advertised the fellowship in our existing community, through word of mouth and by inviting members personally to apply for the fellowship. We invited 15 members, of whom 10 applied. We also announced the fellowship during the general assembly in January; on our internal chat server; our February newsletter; and on our Facebook page.

Furthermore, we included the call in an email that was sent to all the students of the University of Geneva and in an email that went to all students of EPFL (a research institute and university in Lausanne). We don’t exactly know how many applications came through these two mailing-lists, but we know that the majority of the new people found out about the fellowship through these emails. From the people we accepted into the program at least 7 decided to apply after reading one of these emails.

Applications

You can find our application form here.

By July 2020 we have had 34 applications for the fellowship. Out of them, we have accepted 22 into the program. 5 of the fellows decided to stop for different reasons. 1 because they did not have enough time; 2 people stopped replying to messages; 1 decided the program was not a good fit; and 1 person reached their goal and wanted to prioritize other things.

We were surprised by a large amount of applications from people who were not familiar with our community. Out of the 22 fellows, 8 people are new. It seems likely that they would not have engaged extensively with effective altruism ideas without the fellowship in the foreseeable future.

We tried to select fellows based on their reasoning and motivation. Knowledge of effective altruism is not a prerequisite, and we have accepted several people into the program who did not have much previous interaction with EA content. If people are open-minded and motivated to think things through, they will acquire the desired knowledge through the fellowship and interaction with the community.

To have an indicator of the applicant’s open-mindedness and reasoning, we ask: ‘What do you think are the most pressing problems in the world? And why?’ and the next question is: ‘Assume you are somebody who disagrees with your previous answer. Explain why your stated problems might not be the most pressing ones.’ These questions are intended to be an ideological turing test and give us a first impression of their thinking style and how well they understand the other side.

To assess an applicant’s altruistic motivation we asked: ‘What are your personal values and why do you have those values?’. All applicants, however, seemed motivated to do something good and we did not reject anybody for this reason.

The main reason why we rejected applicants was usually a lack of a link to the ideas of effective altruism in combination with an absence of intellectual humility. We sometimes rejected people when they did not give sufficient arguments for their positions on the ideological turing test (e.g. straw-manning or not giving an argumentation at all). When we rejected people, we tried to explain our reasoning and invited them to apply again, incorporating our feedback in their new application. So far we have rejected 32% of the applicants and nobody has reapplied (or reacted to their rejection).

Participants

As of July 2020, we have 17 active fellows. The group is a mix of students, and early- and mid-career professionals. 34% work in academia, 18% are active as an entrepreneur, 6% work for a UN agency, 18% are consultants in the private sector, and 24% are students. Fellows have a diverse background. We have people with an education in biology, computer science, physics, law, philosophy, engineering, and translation.

Fellowship goals

Of the 22 fellows we accepted over the last half-year, 5 formulated a goal related to ‘learning more about effective altruism’, 11 are working on career planning, and 6 work on an EA-aligned project. Some people have formulated two fellowship goals, one related to career planning and another one related to a project. As the fellowship goals about career planning and working on a project also consists of a learning element on topics related to effective altruism, I scored ‘learning more about effective altruism’ only when this was the main element of the fellow’s goal.

The fellowship goal is formulated during the first 1-1 and is evaluated after approximately 3 months. The goal can exist of a long-term goal (e.g. “transition into a high-impact career”) and is broken down into goals that can be attained in a few months (e.g. “create ABZ plan by 1 June” or “develop a framework for evaluating future career options by 31 May”)

Examples of fellow’s goals

Learning more about effective altruism:

“We will break down your goal of ‘improve scientific and rational thinking’ and understanding the concept of ‘maximizing impact’ into:

Diminishing returns & marginal impact for the 23rd of April, Counterfactual reasoning for the 7th of May, Expected Value for the 21st of May, Heavy tails, effectiveness, and cost-effectiveness for the 11th of June

Applying all the above concepts on the problem of climate change and write a short (blog)post for the other fellows for the 25th of June”

“Gain an overview of EA knowledge basics in probability, longtermism, and cause prioritization and write up a short blog post summarizing insights on one of the topics by June 30th.”

Career planning:

“By May 31st have an A/B/Z plan draft”

“By April 24th: make a text document listing skills, availability, diverse information, and thoughts.”

Development of an EA-aligned project:

“By September 4th, develop an overview of the EA fundraising opportunities for projects in the …….. area, mapping funders, amounts, grants, requirements & personal connections.”

“Write a theory of impact by 1 June and define a clear focus for your project for our call of 23 May.”

“Organise giving games in your organization in fall 2020. Discuss the follow-up steps during our call of 8 October.”

Group sessions

Approximately every two weeks we organized a fellow group session. As we had to switch to online meetings at the beginning of the fellowship, we slightly changed our intended program. For the first online sessions we started with this schedule:

19.00-20.00 online co-working

20.00-21.30 knowledge exchange/discussion

21.30-22.00 online board game

We dropped playing a game after a few sessions, as it was too late for the majority of the fellows.

Subjects of the group sessions so far:

Introduction of fellows + workshop: collective action as a community
Workshop & discussion: decision making under uncertainty
First 30 min: random 1-1′s between fellows + presentation of a project of one of the fellows
A fellow was leading an evening on ‘the scientific method & rational reasoning’ (workshop + discussion)
Collective ABZ career planning
Discussion in small groups: longtermism & cluelessness
Workshop: intro to applied rationality
Presentation of Ozzie Gooen on Foretold
Discussion: crucial considerations & moral uncertainty

We chose these topics based on a combination of the interests of the fellow’s, important subjects for everybody to think about, inspiring examples of projects others are working on, and the learning goals of fellow’s. We think that a good way for a fellow to learn more about effective altruism can be to lead an evening about a topic they are interested in.

Results

To measure the impact of the fellowship we conducted a survey among the fellows. A summary of the answers to the survey can be found here.

The most interesting outcomes of the survey

“Estimate: how much more good will you do as a result of joining the fellowship?”

Average answer: 25% Median answer: 20%

The survey was a mix of open questions and statements about the fellowship. For the statements, we used a scale ranging from 1 (disagree) to 7 (agree).

Three highest scoring statements:

Ranked with the average response

The 1-on-1 calls have helped me advance faster on my journey as an aspiring effective altruist: 6,4
I understand Effective Altruism better: 5,7
I am better connected to the EA network: 5,7

Three lowest scoring statements:

Ranked with the average response

I have gained more clarity about my donation strategy: 3,5
I have a better understanding of key impact multipliers: 4
I am more motivated to do good: 4,5

We did not expect a high score on ‘I have gained more clarity about my donation strategy’ as none of the fellows had formulated a goal related to their donations. As an explanation of their answer for ‘I am more motivated to do good’ some fellows stated the following:

“I was already very motivated to do good, so this was unimportant, I was just struggling with how to do so.”

“I wouldn’t say that my levels of motivation for doing good have changed. However, if this was rephrased as “I’m more motivated to prevent bad—e.g. existential risks”, then it would definitely be a 6 or maybe 7.”

As motivation for doing good was one of the selection criteria of the fellows, and we asked people to compare their answers to the counterfactual of not joining the program, it wasn’t surprising that motivation to do good hasn’t improved much.

Other outcomes of the program

I think this quote summarizes the impact of the fellowship program well: “The fellowship has allowed me to make progress faster, and in a better direction for impact.”

Because of the fellowship, 11 members have formulated an ABZ career plan and are actively working on transitioning into a higher impact career. 2 fellows have come halfway in hiring processes at 80,000 Hours, Effective Giving, and Rethink Priorities. One of them has now accepted a non-EA job, the other continues applying.

The fellows whose goal was to learn more about effective altruism have the tools to better evaluate their impact and are better equipped to do good in the future. 1 of them has formulated a new fellowship goal related to career planning and is planning to switch into AI safety research.

The members who are working on EA-aligned projects reported that their projects will probably have more impact, that they are incentivized to continue working on their project, and better connected to the EA network. Examples of projects that fellows are working on: a start-up in cellular agriculture, a research project about the future of consciousness, and independent AI safety research.

Given the changes made through the program, we consider the program successful and will continue for the rest of the year.

Lessons learned

While running the program we also learned a few lessons that I would like to share:

We planned to use a learning grid for the formulation of the fellowship goal but did not use it much. Usually, the learning element was shaped by the goal-driven part of the objective and the learning grid didn’t add much value. The grid is therefore not perfected and I mainly use it as a collection for references that I can send to fellows.
We tried to create connections between the fellows during the online fellowship evenings and by creating a separate channel for them on our internal chat server. Nevertheless, we got the feedback that some fellows would appreciate being more connected with other fellows. Originally we had planned to have a common dinner during the fellowship evenings, but since the pandemic didn’t allow us to socialize this way, we could have invested more into actively connecting fellows.
As the fellowship is a continuous program, it can be harder to keep up the motivation to work on a goal than when the program would be limited by time. One fellow also stated in the survey that more structure could be an improvement to the program. At the moment we are considering adding a new version to the fellowship that is more structured and time-bound and where fellows will work on their goal (likely related to career planning) for ~3 months and afterward graduate “into” the continuous fellowship group.

Acknowledgment

I would like to thank Konrad Seifert for the feedback on this post and our collaboration during the fellowship program.