The Effective Altruism Equality and Justice Project
Effective Altruism (EA) has successfully grown to become a global community of people who have made helping others a fundamental part of their lives. Its core idea — using evidence and analysis to take actions that help others as much as possible — appeals to a range of people across nationalities, political ideologies and religions. What its proponents have in common is they accept that we have limited resources at our disposal and that these resources should be directed where they can do the most good, downgrading projects that may be well-intended but would ultimately accomplish less in relative terms. While the EA community does not explicitly rely on utilitarianism, their emphasis on welfare maximisation, rationality and cause prioritisation can make EA a difficult sell to people whose ethical views are non-utilitarian. Recognising this obstacle was the motivation for this project, which sought to answer the following questions:
How can people with non-utilitarian ethical views, such as egalitarians and justice-oriented individuals, find a place in the effective altruism community?
And are effective altruism methods helpful when we seek to reduce systemic inequalities and social injustices?
To answer these questions, Sam Hilton from EA London sought to gather a group of 15-20 people who would meet 6-8 times during the summer of 2017. The group would work together to give away £1000 of Sam’s money to the best organisation it could find through effective value-led exploration, problem-solving research and conversations.
The idea was suggested at an EA London strategy meeting and then advertised in the EA London Facebook group and on Meetup.com. In the end, only seven people (none of whom knew one another prior to the project) committed, although the initial meetings had a few more attendees. Each meeting was led by an experienced member of the EA community.
Naturally, some constraints were present from the onset. First and foremost, we were a group of working professionals with limited time available to prepare for and work on the project. Furthermore, a lack of existing research on potential charities we would evaluate meant we would have to rely to a larger extent on intuitions than what would normally be the case in EA charity evaluations. We were not discouraged by these constraints, however, because while they would likely cause us to rely more on intuitions than we would prefer, we would still make a donation to an effective charity (even if not the most effective), provide a meaningful answer to a very important question for the future of EA and make a contribution to the community-building efforts of EA London.
The group defined its mission as “For all humans, minimise the extent to which circumstances outside their control limit access to personal need, starting with the most basic needs.” We arrived at this mission statement by consensus after various discussions and exercises to expose our core beliefs. Interestingly the group kept reverting to utilitarian principles of welfare maximisation when pressed to choose between basic needs and equality/justice, but by the same token the members were willing to trade off some satisfying of basic needs to make a significant improvement to equality and justice. This would create a conflict if we were to evaluate charities using a single metric or a quantified best guess measure, as that would ignore the subtleties of this trade-off and likely overstate or understate the relative importance of basic needs versus equality and justice. Therefore we decided instead to devise a rubric that considered multiple factors to guide our decision-making process.
The rubric consisted of six categories, each worth up to five points. The categories were Basic needs, Impact (on one individual), Number of people impacted per £1000, Quality of evidence, Levels of disadvantage, and Actively preventing discrimination. This approach seems entirely in line with EA if your cause is equality and justice; if one charity more effectively promotes equality and justice than others, without sacrificing basic needs and welfare maximisation to an unacceptable extent, we should dedicate our resources toward that charity. (Of course many EAs will argue that any compromising of welfare maximisation is unacceptable but this project is not targeted toward those people beyond showing them that EA can make non-utilitarians more effective).
Populating our rubric and evaluating the effectiveness of charities we discovered that most of them were health-focused charities that were already recommended by various EA organisations (see below for details). After shortlisting our finalists, No Means No Worldwide won by one vote in a voting process and received the £1000.
For the purpose of answering whether or not non-utilitarians can be meaningful participants in and make use of the EA community and framework it seems less important what charity received the £1000 than the fact that we were able to evaluate charities focusing on equality and justice while paying attention to the primacy of basic needs. Our charity evaluator was highly useful in the process and we were able to rely on research conducted by the EA community to work toward our mission. We believe these lessons demonstrate that the EA community can have impact on people who are not traditional utilitarians and equally that equality and justice-oriented people are able to incorporate EA methods without at all abandoning their core beliefs.
While the project had a number of shortcomings, we do not believe they invalidate our positive answer to our initial questions. We are certainly not claiming that our approach is the best approach for non-utilitarians concerned with equality and justice, and we do not want to convince all EAs that they should incorporate those considerations when evaluating charities to the same extent that we did. What we do hope we have been able to demonstrate is that there is a place for non-utilitarians in EA that the community ought to embrace as it continues to grow.
(Keep reading for a detailed account of the various stages of the project. 1856 words. Estimated reading time: approximately 10 minutes)
How we did it
Stage 1: Mapping values and defining the project
The first stage of the project centred around outlining the values of the group members and examine what we really mean when we talk about equality and justice. We considered a number of thought experiments designed to expose our core beliefs. This exercise served two purposes. First, it helped the group members who are not trained in or used to discussing philosophical or ethical questions to become familiar with how such questions should be approached. Second, it allowed us to see where the group as a whole fell on a spectrum ranging from completely utilitarian on one side to egalitarian/justice-oriented on the other, which would help us decide how to proceed in the later stages of the project. To visualise the result, we created a mind map of values (see https://realtimeboard.com/app/board/o9J_k0DKdME=/). As expected, most members of the group displayed non-utilitarian first principles.
Digging deeper into what we mean when we talk about equality and justice, we placed access to healthcare, education and employment opportunities centre stage and did not rank any concepts in order of importance at this point. We defined our common mission to be:
“For all humans, minimise the extent to which circumstances outside their control limit access to personal need, starting with the most basic needs.”
Stage 2: Narrowing the scope
After defining our values and mission, we held initial discussions on charity types, methods, locations and operations that could help us best work towards our mission and eliminated a few options from our search. We did not want to choose a charity focused on sport, arts, culture, science, promoting religion or rescue/emergency as these focus areas deviate from our core mission. In setting this direction we also narrowed the scope, especially where there was group consensus around an issue. In particular we decided to focus on humans. This focus was a result of a consensus decision of the group upon considering each attendee’s values and the current state of the world. (We recognise that setting the direction to exclude animals could be seen as a criticism of this project by some EAs. See: https://www.facebook.com/groups/effective.altruists/permalink/1471898846199801/) Furthermore, we decided to focus on a single-action charity and not a multi-action charity because for the latter, it is difficult to establish which of their interventions are the most effective and it is difficult to reliably measure impact. We also spent part of one session discussing how we could avoid cognitive biases when assessing our charities, agreeing that we would assess all charities according to the same set of criteria.
Key factors in charity evaluation include cost effectiveness, room for more funding, transparency, impact and management. We decided to selectively apply some of these considerations to our project focusing mainly on cost effectiveness and impact, as those are comparatively easy to assess and arguably the most important.
As well as eliminating a number of charity areas and deciding our evaluation criteria, we ranked the values we had associated with equality and justice on a scale of 0-10, where 0 represents death and 10 represents no material impact on the quality of your life. Prior to doing this we had researched the effects of promoting these values through charitable donations. The exercise exposed something most EAs already maintain: basic needs are perceived to be more important when measured up against secondary needs. For example, we assigned a 2-weighting to healthcare, but a 7 to education. Our challenge from here onwards then was to find out whether we were able to identify a charity that could work toward reducing systemic inequalities and social injustices without neglecting basic needs.
Stage 3: Using a rubric for making decisions
Having noted that everyone who cared about equality and justice also cared about well-being and that utilitarian considerations were crucial to the members (all members were willing to allow some inequality to save a large number of lives, for example), it was difficult to use a single metric assigned to charities to measure effectiveness. EA charity evaluation work has often sought a single metric (e.g. QALYs) and/or a quantified best guess figure for different charities, but has avoided incorporating factors like equality and justice that make comparisons less accurate. Given the added complexity that arises when equality and justice also need to be considered, we decided to create a rubric that considered multiple factors.
This part was challenging. We needed to devise a rubric that incorporated all the considerations outlined above. At the same time it had to be simple enough that we could employ it in a meaningful way with the limited time and resources we had available but sophisticated enough for us to succeed in showing non-utilitarians that EA methods are highly effective also if you are not a utilitarian. In other words, our rubric had to be a successful Proof of Concept.
At first, our group thought it helpful to research some charities that could qualify and see what types of charities they were to figure out how to best evaluate them. Of course, this process was insufficient to identify the most effective charities according to our criteria, although we were able to establish that some charities, like Transparency International, would be too difficult to evaluate despite their significant contributions to promote equality and justice across the globe.
The rest of our efforts at this stage concentrated on creating the rubric. As we had clearly established, it needed to 1) measure impact and 2) cost-effectiveness, 3) not ignore basic needs, 4) pay attention to the quality of evidence available, and prioritise charities that effectively 5) prevent discrimination and 6) offset people’s levels of disadvantage. By determining how a charity scores according to these six categories, inputting the data into a spreadsheet, we were able to quantify the performance of the charity and establish its effectiveness.
In our charity evaluator, all categories count equally and earn a charity 5 points, meaning charities are ranked from 0 to 30, where 30 represents the most effective and 0 the least effective.
The categories are as follows:
3. Community bonding / Mental health
Impact (on one individual):
1. Useless or harmful
2. Marginal gains
3. Noticeable improvements
4. Transforms life
5. Saves life
Number of people impacted per £1000:
1. Less than 1
2. 1 to 20
3. 21 to 200
4. 201 to 2000
5. More than 2000
Quality of evidence:
1. No evidence
2. Reasonable assumptions / good theoretical framework
3. One or a few rigorous studies
4. One randomised controlled trial (RCT) or several quasi-experiments
5. Several RCTs
Levels of disadvantage:
1. Recipients not disadvantaged
2. Recipients disadvantaged in 1 way (e.g. they are poor)
3. Recipients disadvantaged in 2 ways (e.g. they are poor and female)
4. Recipients disadvantaged in 3 ways
5. Addresses many layers of disadvantage at systemic level
Actively preventing discrimination:
1. Entrenching current injustices
2. Addresses non-justice issues (e.g. giving bed nets to the poor)
3. Addresses non-justice issues through empowerment (e.g. giving money to the poor)
4. Prevents passive injustice
5. Prevents active injustice
Of course, the charity evaluator only works to the extent that people have faith in the data that goes into it. And naturally our spreadsheet is a prototype that contains omissions. However, already at this stage in the project we had illustrated that it is both feasible and entirely rational to measure how effectively a charity promotes justice and equality objectives. If we can establish that one charity more effectively creates equality and justice than another, we should not focus our resources on the one that is less effective.
Stage 4: Choosing a charity
Because we wanted our spreadsheet to contain as much detailed information as possible, we reached out to the EA community (on various EA Facebook group pages) to fill in information using a shared public Google document.
During the two weeks we made the document available for editing, we did not receive the attention we wanted and the responses came mainly from within the group. One reason the spreadsheet did not get the attention we expected could be that we had not advertised our project well enough in advance and did not provide adequate background information when asking people to help us fill in the information. Another reason could be that it was in the middle of July and people had other priorities. Either way, we decided to organise a co-working session and the group met to work together to fill in the spreadsheet ourselves. That way, we were able to question the input data and explain why we nominated certain charities and not others.
To assign values to the impact and quality of evidence evidence categories, we relied mostly on research conducted by GiveWell, thus placing a high level of confidence in the accuracy of our data.
The charities that scored well according to our evaluator typically focused on health, in particular children and women’s health, and gender equality. The charities that came out on top included Population Services International, Project Healthy Children, Maternity Worldwide, Against Malaria Foundation, Give Directly, The Deworm the World Initiative (Evidence Action) and No Means No Worldwide. (Due to an error, The Schistosomiasis Control Initiative (SCI) disappeared on the version we were using during the co-working session, something we did not discover until after the project. We cannot know if SCI would have been chosen had this error not occurred. We extend our apologies to SCI for this error.)
With a shortlist of finalists following our co-working session, our final meeting was dedicated to decide on a charity. While the evaluator provided a useful indication of which charities were effective and not, we wanted to keep the final decision democratic and we revisited our mission: “For all humans, minimise the extent to which circumstances outside their control limit access to personal need, starting with the most basic needs.”
Although we voiced our opinions on the finalists, we decided to keep the decision democratic and choose our charity through dotmocracy, giving each group member five dots to distribute between our final charities. While this is an effective method of deciding between many options, the process would have been better blinded to avoid anchoring.
When making the final decision, only five members were present. We had narrowed down our choice to two charities: Population Services International (PSI) and No Means No Worldwide (NMNW). We believed that their missions, “[making] it easier for people in the developing world to lead healthier lives and plan the families they desire” and “create a rape free world”, respectively, actively addressed our own mission and the charities were demonstrably highly effective. We held a final vote between the two giving each of the five members one vote.
Ultimately, No Means No Worldwide won with 3 against 2 votes, earning it the £1000. One key argument by those who voted for them was that PSI is a multi-action charity and that we could not be certain that donations would be going specifically to improve equality and justice to the same extent that NMNW would ensure. The group may post a separate write-up on the estimated effectiveness of NMNW.
We look forward to your feedback.
Written by Pouya Jafari, with Enrico Calvanese, Kirsten Horton, Dr Colin McClure, Ellie Karslake, Naim Sheikh and Ruth Stokes. Many thanks to Samuel Hilton, Sanjay Joshi, Holly Morgan and Saulius Šimčikas.