Edited 02.06.2021: Minor edits to the "Tentative Defintion" section for clarity.
A Tentative Definition
Meta Effective Altruism is one of the main focus areas of the Effective Altruism movement. It can include research to help direct efforts (Global Priorities Research) and efforts to build or support the EA movement and its members (EA Movement Building). Some meta EA projects can also be focused on specific cause areas, professions or more (Within-Cause Meta). Typically, meta effective altruism is at least one step removed from direct impact.
This post aims to clarify the different kinds of work that is counted as “meta effective altruism”. Examples are non-exhaustive, and many are listed under multiple categories of movement building because the categories defined are not mutually exhaustive.
Map of meta Effective Altruism
This diagram was not made for this post, but I found it somewhat useful to get a broad overview of the elements I'm describing. To save time I chose not to update the image with the terminology that I've used in the post.
Global Priorities Research
Global priorities research (GPR) is concerned with the question ‘If our aim is to do the most good possible, from a totally impartial perspective, with limited resources, how can we do the most good?’ (Slightly altered from Source)
Foundational Global Priorities Research
"Macrostrategy" or "foundational" research that would inform cause prioritization in a less direct way (e.g., research into the Fermi paradox or whether we're living at the hinge of history). (From the Forum tag)
Examples include: Global Priorities Institute
Applied Global Priorities Research
Applied global priorities research is most commonly done for object-level causes, however, it can also include research on meta considerations.
Cause prioritization refers to efforts to find the most important causes to work on and compare possible resource allocation between different causes.
There is often overlap between Cause Prioritization and Intervention Evaluation since the availability of promising interventions can change the way we prioritise a given cause. I’ve separated them because they can be distinct projects.
Examples include: Open Philanthropy Project, 80,000 Hours, Future of Humanity Institute, Rethink Priorities
Local Priorities Research
A subset of applied global priorities research that is narrowed to a specific geographic region (a locale). This is a cross-cutting area that overlaps with cause prioritization and career advice research.
Examples include: Region-specific career advice
EA Movement Building
Any work to build, grow or develop the EA movement. Movement building does not need to involve using or spreading the Effective Altruism brand or name. The types of movement building are described below.
Developing groups or subcommunities (local & online) organised by geography, shared interests, careers, causes and more; events and conferences; career advice and more which spread and educate people about the ideas and principles of the EA movement and encourage people to take actions based on them.
Examples include: EA Groups, AI Safety Support, 80,000 Hours, CEA, Giving What We Can
The development of community-wide products and services that help intra community coordination. Examples include building tools or websites such as wikis, forums, tools, platforms; offering cross-cutting services like shared services to EA organizations; and offline events such as conferences, community houses, and regional networks.
Some community infrastructure may be limited to certain subgroups, such as events and services for leaders but still provide benefits to the wider community.
Examples include: EA Forum, EA Hub, EA Global conferences, Project incubation (e.g. Charity Entrepreneurship), EA Operations Slack
Developing new or influencing existing academic disciplines or fields through the creation of new organisations, advocacy or funding academics to work in this field.
Closely related to Professionalization.
Examples include: Scholarships and teaching buyouts (e.g. Forethought Foundation, several grantmakers), Charity Entrepreneurship’s Incubation Program, Future of Humanity Institute’s Research Scholars’ Program, Conferences (such as EA Global)
Raising money to help achieve movement goals.
Examples include: Giving What We Can, One for the World, Effective Giving, Founders’ Pledge, Raising for Effective Giving, High Impact Athletes, more.
Distributing movement resources towards where they need to go. This includes charity evaluators, foundations and individual donors who are making donation decisions.
Examples include: EA Funds, Open Philanthropy Project, Survival & Flourishing Fund, Founders’ Pledge Funds
Developing the EA network to include non-EA actors, organisations and communities. See Community vs Network by David Nash.
Often overlaps with Community Building.
Examples include: Effective Altruism Consulting Network, High Impact Athletes, Raising for Effective Giving, All-Party Parliamentary Group for Future Generations (London), promoting positive values (e.g. EA for non-EA People: External Movement Building by Danny Lipsitz), EA Asia, and more.
Spreading the ideas and principles of the EA movement and encouraging people to take actions based on them. Outreach can be general or targeted to specific groups. Outreach often involves content creation.
Examples include: Books on EA, Podcasts, EA Groups, 80,000 Hours, Giving What We Can
Giving an occupation, activity, or group professional qualities. This can be done by creating a career out of, increasing the status of, raising the qualifications required for, or improving the training given for an occupation, activity or group.
Closely related to Field Building.
Examples include: Providing training (e.g. Animal Advocacy Careers), WANBAM, Future of Humanity Institute’s Research Scholars’ Program, Various internship opportunities
Movement research is research done to help determine what kind of movement building to do, which interventions to conduct and which
A subset of EA movement building that is cross-cutting is research specific to the allocation of talent. Other effective uses of one’s time could be included within this such as volunteering, although this has not historically been a focus of this kind of research.
There are three main types of carer advice research:
- Movement-level career advice research: "Which career paths are especially impactful, overall, for people to join?"
This research identifies bottlenecks in top causes and makes recommendations on how to address those bottlenecks, such as specific actions (e.g. career paths, jobs) movement actors can take to pursue them.
Examples : 80,000 Hours cause profiles, Animal Advocacy Careers’ skills profiles, Probably Good, Local Priorities Research (see below), Studies predicting which jobs get automated
- Individual-level research: "How can a given person find an impactful career that is a good fit for them?"
This research identifies best practices, frameworks and tips on how to have a successful, fulfilling career. It could help people find a career that is the right choice for them: that is aligned with their values, that they can excel at, and that they are motivated to stay in in the long-term.
Examples: 80,000 Hours' 2017 Career Guide, Career Profiles, So Good They Can't Ignore You by Cal Newport
- Career advice intervention research: "What can we do to help more people find impactful careers?"
This research identifies specific interventions (e.g. career 1-1s, articles, events, headhunting) which can help achieve 1) and 2).
Examples: Animal Advocacy Careers pre-registering a study of their career 1-1 calls, Research on the effectiveness of coaching.
Some meta-EA work is done at the level of individual causes, rather than the movement as a whole. Intervention Evaluation and Impact Assessment can also apply to meta EA as a cause itself (for example, conducting a study to see if conferences positively influence altruistic behaviour).
See the categories listed above, but applied to different causes rather than EA.
Intervention evaluation is the comparison of different interventions, normally within a given cause area. However, some organizations may compare interventions across cause areas to decide where to give.
This may fall outside the scope of meta and more within certain cause areas. Some organisations listed below evaluate interventions across a number of causes, but many of the employees may be specializing in a particular area.
Examples organizations include: GiveWell, Charity Entrepreneurship, Rethink Priorities, Happier Lives’ Institute, Future of Humanity Institute, Center on Long-term Risk
Impact assessment measures "the effectiveness of organisational activities and judging the significance of changes brought about by those activities" (source), as well as cost-effectiveness analysis and cost-benefit analysis, which compare impacts and costs. The activities and interventions need not be tied to any specific organization. (from Forum wiki)
There is often overlap between Impact Assessment and Intervention Evaluation because charity evaluators will first identify the most effective interventions and then evaluate charities that implement them.
This intervention may not be meta effective altruism for the most part, but the development of frameworks for charity evaluation or encouraging evaluation more broadly could be considered meta effective altruism.
Examples include: SoGive, GiveWell, Animal Charity Evaluators, Founders’ Pledge, Giving Green, Lark’s annual AI Safety Literature Review and Charity Comparison
Thanks to many, many reviewers for providing valuable inputs on this post. I decided to publish because I felt a lot of the categorizations and definitions I offer haven't really been discussed that much and it wouldn't make sense to try to create a "perfect" version of this post. I hope this can be a living document, so please feel free to suggest edits, corrections and additions in the comments.
I found the distinctions you drew between types of career advice research interesting - I hadn't really thought of those distinctions before and expect to find that useful in future. I've also suggested that that section be drawn on for the new EA Wiki entry on career advising.
That said, I think it should really be a four-part (or maybe five-part) distinction, with the parts being:
Currently your "individual-level research" category kind-of implies that it includes a bit of work on what would make someone successful, but really that's something fairly different and something that can be researched in a more generalised way.
I say "or maybe five-part" because one could also add "Individual-level career success research". But I'm guessing that that wouldn't add much.
For some reason, I'm just seeing this now. I agree, that's a good distinction between 2/3. Building on this:
The very vague definition of "Cause Area" is making it hard for me to think about meta EA. It feels like GPR is a cause area and so working on it would be direct impact work but I am not sure. Same goes for EA Movement building. Also, it starts getting trippy if we claim meta-EA is also a cause area!
Maybe we can clarify the definition for cause area within this meta EA framework?
Thanks for your comment, I think it's a great point!
I think it depends on your level of analysis. I outline my thinking in the comment above for how I'm thinking about it for this post.
In most cases, it probably makes sense to separate out Global Priorities Research and EA Movement Building because practically the interventions can be very different (although there are some interesting areas they overlap - e.g. career advice research).
Also, there was a discussion on this you might find interesting.
I am a little confused about the purpose of this post, because surely meta-EA is just EA? I feel like the major innovation of EA is the idea that altruists can and should compare the value of different interventions (which you appear to consider meta-EA). In other words, EA is meta-altruism.
The content might be useful as a road-map, but I think that the terminology is a bit misleading. What these areas have in common is that they are indirect, as opposed to having some kind of abstract meta-ness property.
You're right that basically everything defined in this list can be referred to as EA work itself. However most of these things can be referred to as "meta EA" is used within the community.
Meta EA is not limited to comparing causes or interventions. The examples given in the EA Movement Building section are more action oriented. E.g. running a local group or field building. They are a higher level of abstraction (one step or more removed) from direct impact.
I agree that that the USP of EA is the concept of cause neutrality / prioritisation. However EA is more than just meta work - so some people may spend a little time on comparing, and then move on to direct work in the space (e.g. lobbying for animal rights, research on x-risk, implementing vaccine programs). I think meta work is sufficiently different that it's worth mapping out the possible things you could do.
I think these things have a meta-ness property in the sense that they influence the structure / composition / nature of the EA movement. GPR research influences the causes we focus on, movement building affects the people within the movement and what they do. One influences the other.
For example, if research on cause prioritisation suggests we should prioritise AI Safety movement builders may do active outreach to software engineers, thus changing the composition of the movement. Similarly, if fundraisers decide to fund certain cause areas, they may pull in new people who counterfactually wouldn't have joined the movement. On the other side, if movement builders start to quickly grow specific profession-specific networks, then there may be interest to research how people from say, a political background, can leverage their political capital - which may result in a very different prioritisation than if we are looking for causes that have the biggest funding gaps.
I think I had a similar confusion to bwildi, or more specifically I wondered while reading this what wouldn't count as meta EA. Your comment helps clarity that, but I think there's still an issue, which is essentially that (almost?) all impacts will only occur indirectly. Some examples:
I still think like there's a useful category in this vicinity, which includes the examples you give but doesn't include things like researching specific AI safety ideas or doing policy advising. But I don't think that the definition you give by itself makes it clear what's in and what's out of scope.
I think maybe I'd see it as cleaner to have a concept for "building effective altruism", and a concept for "global priorities research", and then everything else (e.g. technical AI safety research, policy advising). Rather than trying to merge building effective altruism and global priorities research under the meta EA banner and then explain why those things fit together but everything else doesn't fit as part of them.
(All that said, I found this post interesting, and I think this sort of mapping seems hard so I'm not saying I'd have done a better job.)
(I think neither of those things attempts to explain where movement building fits in.)
Is it a common use to consider GPR when talking about Meta-EA?
This came up amongst reviewers as well - whether meta EA is just about EA Movement Building or if it's broader. I personally think it's broader, and thus includes GPR as well because it GPR work will influence the other parts.
Hi Vaidehi - thanks for the post!
Would you see worldview advocacy as part of meta-EA? Moral circle expansion and moral advocacy more generally are examples often referenced within EA - but some worldviews go beyond morality into epistemology.
As an example, I work (amateurishly) on developing and building community around Sentientism as a combination of naturalism and sentiocentrism. I summarise it as "evidence, reason and compassion for all sentient beings." Some think of it as a corrective to the anthropocentrism of secular Humanism.
Meta-comment about writing this post: