Hide table of contents

Authors: Vaidehi Agarwalla & Arjun Khandelwal

Key Takeaways

  • You may want to read this post if you’re interested in understanding the engagement of EA community members and how it might evolve over time. This post is especially relevant to community builders.
  • The most prominent model of community building, and implicitly, EA engagement, is CEA’s funnel model. This model is suitable in specific contexts, such as when looking to recruit members into an organisation. (Read More)
  • The funnel model’s approach is contribution-based (where engagement is defined largely by someone’s contribution to the movement itself, rather than the world), and linear (it implicitly suggests that there is only a linear, non-circuitous path for dedicated EAs to take). (Read More)
  • As a result, it does not consider certain groups of people in the EA movement, and thus using only this model to inform strategy may be harmful to movement health. (Read More)
  • I propose an individual-focused non-linear approach. (Read More)
  • In this approach stages are defined not by an individual’s contribution to the movement but by an individuals’ relationship with EA and multiple possible paths are mapped through the stages.
  • I believe that such a model is a more robust model for EA engagement. (Read More)
  • I provide evidence for this approach by presenting selected data from ~22 in-depth interviews with EAs. (Read More)

Thanks to Nathan Heath for reviewing a draft version.


EA community building aims to influence individuals to achieve the movement’s goals (to figure out how to do the most good, and do it), by engaging with EA principles and ideas and being influenced by them to take actions or change their behaviour. Since community building can encompass a wide range of possible activities, we need ways to narrow the scope down. One way to do this is to model different stages or levels of engagement. This can be a useful way to identify target audiences and to understand what their needs might be to narrow down the possible activities.

There have historically been relatively few models of community building prominent in the EA movement.[1] I believe that the most prominent of these, the funnel model, has important limitations. I outline its approach, uses and limitations, and then propose an alternative approach, and the uses and limitations of this new approach. In the final section, I propose a tentative model based on this approach.

The Funnel Model

CEA adopted the funnel model some years ago to think about how they could support the EA community. This model comprises a series of narrowing segments, shown in the image below, which are defined by the contributions a person makes to the movement’s goals, especially in the final stages. For example, contributors “are willing to make significant sacrifices to act on those ideas” while the core members are defined as having “have devoted most of their resources to acting on those ideas”. You can see their funnel model and concentric circles model posts for more details.[2]

The approach

The funnel model, like all models, acts as a set of headlights - illuminating what it is optimised for, and leaving much else in darkness. Below I identify the aspects the funnel model focuses on, the consequences of the focus, and what this approach, and model, is well-suited for.

Contribution-based definition

The model has a contribution-based approach. It looks at individuals primarily in their capacity to contribute quite directly to the movement - such as being recruited to an aligned organisation, or as a thought leader.  Thus the model is “inward-facing” in that it focuses on flow into the movement, rather than flow out into the world, or flow between the two.


The funnel model emphasizes that people flow through the funnel in one direction, going through a series of stages in a linear progression. Thus it doesn’t account for nonlinear journeys, which may be very common. It may also implicitly suggest that the straightforward progression through the stages to the core is the “best” one - something that individuals should strive for, and that community builders should encourage members to do.

Limitations of this approach

Due to the contribution-based and linear approach,  a number of groups are excluded from the model. The model cannot therefore account for the impact of community-building actions on the community as a whole. For example, there have been several discussions since 2018 on the frustrations and difficulties of many members. The groups excluded are:

The EA network

David Nash points out a shortcoming of the funnel model is that “people often link engagement to impact, thinking that those who are more involved in the community, go to lots of events or work at EA related organisations are having more impact and so put on events and tailor content towards those activities”. However, he suggests that the vast majority of people who are influenced by EA ideas or have impact may not be actively engaged in such a manner, and that “it may be more worthwhile to provide them with value that they are looking for, whether that’s donation advice, career ideas or connections to people in similar fields.” The EA network may have a nonlinear journey (see David’s inverted funnel model of engagement).

The EA network may be harder to track, which would mean we have greater uncertainty about the EA movement has influenced them. However, if they can help the movement achieve its goals, then this seems like an important group to consider.

The “middle”

There are many active EA community members, who benefit from being a part of the community but may not currently be on a path to the end of the funnel, or ever will be. Some may need to be connected with EA for some time before they make larger changes in their life, as they learn more and test out the best things to do. Others may only ever make small changes, but they may contribute in other ways to the movement which are less typical and higher variance. A welcoming community would help such members learn about EA, stay abreast of recent developments, be motivated, and feel supported. For those who want to get more involved, the community could help them find new opportunities to test different career paths.

Regardless of their contributions to the movement, if such members feel frustrated or disillusioned, this could have negative impacts on the community. Thus, an approach that does not factor in this group may not fully appreciate and mitigate this risk.

The people who leave

The model encourages us to measure what is most easily measurable - those who stay in the movement rather than the ones who might leave, either temporarily or permanently. It is also important to understand the reasons that might drive people to leave the EA movement or experience (apparent) value drift.[3]

It seems prudent for a movement that aims to survive generations to consider the long-term health of using models that explicitly consider and invest resources on this group of people, even though it may be costly, difficult to track, and possibly hard to justify.

The funnel model is well suited for

  • An organisation hiring a new candidate, intern, or volunteer
  • A group evaluating members’ contributions to their group (e.g. to run events)
  • A group looking for members to become more actively engaged in the EA community or EA organisations
  • A subgroup of people with homogeneous and linear journeys.

An individual approach

The approach

Individual’s journey

The focal point of this model is on an individual’s journey and relationship with EA over time, through learning, thinking, and doing. The focus is on the development of an individual and helping them have the highest impact they can, rather than their relative contribution to the movement as compared to others.


This model would try to closer map reality by not suggesting a single linear path through the stages. Instead, it suggests that EA journeys may be varied and offers actions to support an individual’s growth and development.

Limitations of this approach

  • It is scope insensitive to contributions of the individual members. There is a balance that community builders need to strike when determining their strategy. On the one hand, a community that ruthlessly prioritises can be seen as elitist, unwelcoming, and induce things like imposter syndrome. On the other hand, some people can likely have more impact than others. However, it can sometimes be difficult to identify such individuals or accurately predict their impact in the long-term. For example, someone who seems promising today may not be part of the movement in the future, or vice versa. It seems important that the community support members not because of the value they bring, but rather because we are a supportive community.[4]
  • It provides a more complex model that risks becoming too noisy, making it harder to track and drive decision making. Groups with very specific goals (such as those outlined for the funnel model) may be better off using a narrower model that provides more clarity.
  • May not help identify how to prioritize people, who to help, or when to help them. For example, members of the EA network and the core would likely require different resources. Groups trying to narrowly prioritise specific strategies for these groups would find this model less useful.

What the individual approach is well suited for

  • Groups looking to support individuals on their journey to have more impact with their actions
  • Understanding the possible journeys of heterogeneous community members across the EA movement or across a particular sub-community
  • Identifying or focus on reasons for a drop off in engagement with EA
  • Identifying obstacles to engagement

Mapping the Stages of an EA’s journey

The above insights come from a series of ~22 in-depth interviews with EAs seriously considering or in the middle of a career change conducted by Benjamin Skubi and myself between July - Sep 2019. We followed up with participants in Feb 2020.

While coding and analyzing the data, we tried to identify some common stages that interviewees went through to try and draw observations across participants. This turned out to be quite challenging because each individual’s journey was so complex. We’ve tried to map an estimation of what we observed in the simplified diagram below, but note that it is far from perfect or complete. We expect further work in this field could change the specific mechanisms and stages, or the relevant importance of them.

For each stage, I have provided 5 journeys of the participants we interviewed to illustrate the diversity of experience, even amongst participants who end up making similar choices.

  1. Discovery: Someone makes an initial discovery of key EA ideas and their interest is piqued.
    1. Andy learnt about effective giving from a colleague at his workplace. He is convinced by the basic arguments, and tells his partner about EA.
    2. Bob was part of the atheist community at his university. He first discovers the student EA group through a few collaborations and is intrigued by the proposition of EA.
    3. Carla learnt about EA through the rationality community in Berkeley and wants to get more involved to figure out how to do the most good in her life.
    4. Daniel came across the early writing of 80,000 Hours in 2013/2014  and was immediately excited by the idea. Much of what 80,000 Hours said resonated with him. At the time, he was working for a prominent local NGO in a major US city.
    5. Edward learnt about EA through various touch points between 2007-2011 including an New York Times (NYT) article, Overcoming Bias, and a series of posts on efficient charity. He didn’t start earning to give 10% (as the NYT article stated), but did start making larger donations.
  2. Learning: The individual learns about the core frameworks, views, and shared values of the EA community.
    1. Andy’s partner became very involved in the local EA community, and told him about various Facebook Groups. Andy also found 80,000 Hours and resonated with the ideas there. He began to be interested in direct work.
    2. Bob starts organizing the EA group at his university and attends monthly socialis. An avid reader, he reads many books on the moral philosophy behind EA.
    3. Carla learnt about EA through her conversations in the rationality community. She didn’t need to go online much, as EA is “in the social waters” of Berkeley. She also attended EA Global each year.
    4. Daniel did a lot of detailed research outside of 80,000 Hours on the future of automation during 2013-2015. While he was inspired by EA ideas, he didn’t find 80,000 Hours’ concrete advice at the time useful because he noticed how quickly their priorities were changing.
    5. Edward was aware of EA on the periphery for many years. He heard of GWWC in 2012, but didn’t take the pledge. He started splitting his donations between MIRI and GiveWell. He wanted to wait until he had time to do some proper research, so spent a few years “on autopilot”.
  3. Exploration & Engagement: The individual begins to more seriously engage with EA ideas and may get more involved with the EA community. They may start taking small to medium-sized actions as well.
    1. Andy had been involved in promoting effective giving at his workplace. Although his partner was very involved in their local EA group, he found that he got less from that community because they were mostly researchers from the local university.
    2. Bob started organising his university EA group, and became very good friends with the other community members. His friends were Earning to Give around 50% following the anything above the poverty line approach. He felt they motivated him a lot and he took the GWWC pledge during this time. He aimed to also donate 50% eventually.
    3. Carla had worked as a technical AI researcher through an affiliation at an AI organisation, but found that research was a poor fit for her. She preferred jobs with more where she can study people. She felt guilty that she was a perfect fit for a priority path, but didn’t feel motivated to do it.
    4. Daniel didn’t really engage with the EA community at all, or have a defined exploration stage. Based on his research, he decided to get a high-paying job with a low risk of automation replacement where he could Earn to Give about 50% of his income.
    5. Edward began to delve more deeply into EA in Jan 2018. He felt he needed to be more deliberate in his life in general, but considered EA to be one of the most important things. He started taking actions, prompted by the EAG SF application, such as attending a local group, reading 80,000 Hours and other EA articles. He also attended EAG SF that year. This resulted in some important effects on his thinking, such as considering the long-term future and animal welfare more seriously, and becoming a reducetarian after EAG SF.
  4. Entry/Transition: The individual begins a major life change - an entry or transition into what they believe is a higher impact trajectory. This could include changing jobs, taking a giving pledge or seriously exploring a new focus area.
    1. Andy left his consulting position after 1.5 years there and took a job at a large tech company while his partner finished her studies for about 1 year. After that point he planned to apply for direct work jobs at EA organisations and was willing to relocate to the US (from the APAC region).
    2. Bob really wanted to do direct work at an EA org, but also has a backup plan to find a high paying job related to his PhD. He spent many months applying for both jobs. He found networking for the non-EA jobs was the most challenging aspect for him. During this time he helped organise his local group in a small city in the US.
    3. Carla took a while to get over the negative feelings associated with not being able to do AI alignment research. After some time, she decided to focus more on mental health and start doing therapy work.
    4. Daniel applied and completed a short (3 year) Nurse Practitioner’s (NP) program and planned to donate 50% of his income to charity. Due to his busy schedule he couldn’t connect much with the local EA community during this time. Daniel started working as a NP and was struggling with job satisfaction, due to issues with the US healthcare system out of his control. He had initially thought he would be able to feel satisfied with the small improvements he could make to his immediate workplace and team, and perhaps the larger improvements he may be able to make as he progressed in his NP career.
    5. Edward wanted to transition from his day job as a software engineer to direct work in AI alignment. In Oct 2019 He attended an AI workshop that he found useful, but ultimately needed to improve his chronic injury before he could take action.
  5. Long-term Retention: The individual continues to be aligned to EA values and ideas and makes (possibly discontinuous) decisions influenced by EA thinking in the long-term (> 5 years).
    1. Andy, Bob and Carla have not been involved in the movement longer than 5 years at the time we followed up with them. Andy and Bob are somewhat concerned about value drift and motivation, although both believe the counterfactual impact of Earning to Give is very high. Carla did not seem that concerned about this when asked about her current challenges.
    2. Daniel has been a part of the EA community, although he’s had limited interactions with the local community, for 6+ years. He has made major life changes based on EA principles and stuck to them.
    3. Edward has known about EA for almost 10 years and had been donating for some time but only started to become more seriously engaged in 2018, which is also when he took the GWWC pledge.

Although I've presented only two approaches here, I don't think they are sufficient. I believe we would need different approaches depending on the specific strategy question and target audience at hand. However, I believe the individual approach may be important for general community building purposes for the community overall, and in that sense it is a little broader than the funnel model.

I hope this post might start a discussion on different models of EA engagement and what their limitations and benefits are. I’d be very interested to hear other approaches to creating models of EA engagement that are not outlined here. I’d also be interested to hear how your engagement with EA fits into the stages, especially if it doesn’t fit nicely. This is a very preliminary mapping, and I’m keen to hear people’s thoughts on the examples provided.

  1. See a comprehensive list of models here. The 3 models discussed under "Models of an EA group" are either the funnel model, or practical applications of it which draw on the same theory. The international models do not appear to have been discussed widely, and Jah Ying Chung's framework is relatively recent. These models also don’t quite deal with the same questions. The other sources discussed on this page are mostly strategies rather than models or frameworks. ↩︎

  2. I make no claims about CEA’s current thinking on the funnel model or their theory of change for community building. The funnel model hasn’t been updated in many years. However, the funnel model is a good example of a general strategy that has been prominent in the EA community in past years. In the absence of an alternative I have proposed this one, the one which I have been using for some time. ↩︎

  3. For the curious reader, the 2019 EA Survey asked about reasons for declining interest in EA. ↩︎

  4. An example of this is my alma mater, Haverford College, providing a full scholarship to an ordinary freshman whose parents passed away around when he started college. Years later, he became a billionaire and is now Haverford’s largest individual donor. Haverford had no way of knowing that this particular student would become significantly wealthier than his peers and then become a large donor. Providing this scholarship had multiple positive effects for the college. One, it engendered a strong sense of loyalty from the student himself, one which eventually vastly repaid the cost it took to earn it. Two, this student’s story became widely known, and seemed to be positively received by students. More on his story here. H/T Arjun ↩︎

Sorted by Click to highlight new comments since:

Thank you for this article. I really like the idea of your model and how you highlight how it compares to the funnel model. I think it is a good idea for the EA community to focus more on the "middle" and how to keep people motivated who are not the most highly engaged. This is especially important as it is unclear what causes might be the most important in the future. Right now highest engagement is only really possible for people who are working in the currently most valued causes areas. However, those might change in the future and we might need people with different skills and perspectives and retaining those people is only possible if we offer everyone ways of participation and growth. 

Thanks for writing this; I found it interesting (especially with the diagrams)!

I can't seem to find a comment/message I wrote some time ago (perhaps on the EA Organizers Slack), where I talked about the two-pyramid model, which emphasizes the potential disconnect between someone's beliefs and actions (and that stronger beliefs/actions tend to be less common). I wanted to bring it up so as to expand on it and apply it to the discussion here.

Maybe this is overly simplistic, but it seems that two of the most important goals/targets of community leaders/organizers tends to be "persuading and supporting people to engage in more-effective actions" and "growing the community (mainly to indirectly support the first goal)." Given that, and partially  in building off of/in regards to the individual model, I was wondering if you have considered some kind of model that emphasizes the shifts in an individual's different characteristics in relation to EA/the EA community?

For example, one characteristic could be "belief in/alignment with EA principles": to what extent does a person believe the research/arguments regarding cause prioritization and/or the ability to reasonably estimate impact? This could importantly be different from a characteristic like "Action alignment with EA principles": to what extent does a person actually act on EA principles (e.g., donating to effective charities, pursuing high-impact careers)? This could also be different from something like "level of engagement/interaction with the EA community," such as "to what extent do they attend events, etc.?" I make that distinction particularly because I have organized a group where some people would attend somewhat regularly largely for the social-intellectual atmosphere (and, probably, the free pizza) but did not seem to really express noteworthy changes in belief or enthusiasm for action. Additionally, one could assess a characteristic like "action to support the community": to what extent do they help recruit others, support events in financial/facility/planning/execution/etc. terms, and so on. 

It seems that all of these characteristics can be present to varying extents: there are probably some meaningful correlations between the first two (EA belief and EA action), but they may not always align. Additionally, someone may have EA beliefs and take EA action (with regards to career and philanthropic choices), but not be very involved in terms of the latter two characteristics (community engagement and community support)--and further research may find this to be related to relevant trends like value drift, etc. You might also just have some people similar to what I had: low interest in EA beliefs or actions, but some engagement. And so on.

Ultimately, I definitely can see this as being more complex/noisy, but I think it could potentially be a helpful "background/advanced tool" to have in conjunction with an easier model like the one you describe. Of course, I'm not engaged in the community organization literature (so I hope this post isn't totally duplicative or missing the point), but I would be interested to hear your thoughts!

This was a very interesting comment! 

Tangentially related comment:

General model of movement building

I've been working on a model somewhat similar to the aspects you mention, but with a bird's eye view of the movement to try and figure out coordination challenges. It's not finished yet but here's a relevant section:


1. Willingness to coordinate

a. Belief in Goals

b. Belief in the success of actions

c. Sense of belonging to the community

2. A suitable opportunity structure

Together I term 1 & 2 "organisational discipline" within the movement - the conditions necessary for the movement to coordinate and achieve it's goals. 

Mapping your suggestions loosely

"belief in/alignment with EA principles" -> 1a/b

"Action alignment with EA principles"-> 2

"level of engagement/interaction with the EA community," such as "to what extent do they attend events, etc.?" -> Not modelled explicitly because this isn't a community focused model, but probably 1c) may come closest.

"action to support the community"-> not modeled explicitly

I think this touches on some good points, such as the "willingness to coordinate" being influenced by motivation/perceived value in coordination. I am a bit confused/unclear about what you mean by "suitable opportunity structure" and/or how it relates to action alignment; does it refer to ideas/questions like "do the opportunities/platforms/networks that are necessary for coordination exist (such as Slack, narrow-topic groups, etc.)?" (It's probably clearer in the context of the larger post/writing, I just wasn't 100% sure here.)

More broadly, does this model employ a community-centered decision approach along the lines of "1) Does the community want to coordinate; 2) Is the community able to coordinate?" I mainly ask for clarification but also because it vaguely reminded me of a simplified rational-actor-centric decision model I know/like, which basically focuses on three main factors: beliefs, values/preferences/goals, and options/capabilities. Would I be correct in thinking that "beliefs" is similar to 1b, "values" is similar to 1a, and "options" is similar to 2?

The other question/comment I had was with regard to 1c. When trying to figure out "why don't people want to coordinate," I think that's a good point to include in a shortlist of questions to ask for troubleshooting. If I were to go a bit deeper, though, and look at it on a semi-rational-actor choice level (as I like to do), I think 1c strikes on / could be expressed as an alternate motive for coordination: "to what extent do people enjoy coordination for the process/journey (e.g., socializing with others, performing/affirming my values) as opposed to just the outcome/destination (i.e., success)?" --The contrast being that 1a/1b are more focused on "what is the outcome: how likely is it and how much do I value it?" In contrast, I think one key factor/dampener for coordination (at least on the individual-choice level) are the drawbacks in terms of opportunity cost, stress, financial or other resources (perhaps), etc. Thus, I was wondering if you were planning to include such "coordination costs" as part of the model?

Could the Individual Approach be considered a complement to the Funnel Model?

The Individual Approach explains "Entry/ Transition" as a major life change. To me, that sounds a lot like moving deeper into the funnel

Long-term retention sounds like staying at your current stage. And drop-off sounds like moving back a stage, or leaving the funnel entirely

Hi Geoffrey, and apologies for this abysmally late reply. I think it is definitely  a complementary model.  I think that the transition stage can be a transition to directly contributing to the movement, but it may also be a transition that doesn't appear very direct / may not be perceived as such by the community. So perhaps it partially overlaps in that way. 

Curated and popular this week
Relevant opportunities