Entry Question

What are the next concrete steps I could take to become a researcher in one of the EA organizations.

Disclaimer

I am “heavily” paraphrasing what people have told me about what skills are important. Any misunderstanding from my part is my mistake alone.

Current level

The following table shows when I applied, which organization I applied to (and to what position), how far I went in the rounds and the number of people who qualified to that round.

Year-mon Org Rounds passed “Ranking”
Apr-2020 CE-Intern S-FTFI-WT ~20/~40
Jun-2020 HLI-Intern S-WT -/30
Jun-2021 OpenPhil-Program-Assistant S-WT -/-
Jun-2021 Elizabeth-Assistant S-FTFI 30/30
Jul-2021 RP-Researcher S-WT1-FTFI 20/>100
Jul-2021 FP-Researcher S-HRI 25/>100


S –> Initial Screening
FTFI –> Face to Face Interview
WT –> Work Task
WT1 –> First Work Task
HRI –> HR Interview

CE –> Charity Entrepreneurship
HLI –> Happier Lives Institute
RP –> Rethink Priorities
FP –> Founder’s Pledge

How to get to the potential options?

  1. It appears that I can look at what successful people (who are “similar to me”) did.

    “Similar to me” meaning, people without a Harvard degree, people without an economics PhD, people who at some point of time didn’t get into EA organizations and then later managed to get in.

  2. I could offer my services as an intern, or as an assistant to people who are already in EA organizations. And that way perhaps I can find out more about the sort of work being done and/or the sort of “skills” I would actually need.

  3. I could ask people in the business “what steps I need to take” to become “GiveWell/OPP/RP material”

What did the successful people do

A while back I asked a question to the EA community about how people who got rejected in EA organizations, “made it” eventually. And the following section is based on the answers from that question in the EA forum.

It appears that the successful people wrote a lot in the EA forum prior to getting into an EA organization. As shown in the table below, all of the people who answered my question have at least a few posts on the EA forum before they got into an EA organization. I on the other hand wrote one post on ‘Talent Constrainedness of EA’ which was really bad (wherein I butchered concrete thinking and massively got confused with definitions. Thanks to an STM for pointing it out).

Name Org Posts before EA org Total Karma Comments
Me N.A 1 154 Baseline
Peter Wildeford RP ~35 7k+ 2017-Jan
Saulius RP 3/18 3k+ 2018-Oct
Gregory Lewis FHI 9/17 3k+ -
Micheal St. Jules CE ~10/22 1k+ 2016
Micheal Arid CA 7/76 7k+ 2020-jan


RP–> Rethink Priorities, CA–> Convergence Analysis, FHI–> Future Of Humanities Institute, CE–> Charity Entrepreneurship.

Some people have made “thoughtful” and “intensely researched posts on the EA forum”, after being rejected by an EA organization and before getting into a Researcher position in an EA organization. Saulius, for example, after not making it to ACE as a full-timer (after his internship there), published 2 “well researched” articles, where one of them even made it to ACE’s Blog and the other on EA Forum (with 50+ karma points). Apparently this sparked the interest amongst Rethink Priorities’ recruiter to invite him for an interview. Note: One has to also do well in the tasks and interviews to get into the organization.

Another example is Micheal St. Jules who did not make it in 2016 to ACE or GiveWell. Before getting his internship at Charity Entrepreneurship in 2020, he had won an “EA forum prize” for his post on Biases (which grossed an awesome 90+ karma points; contrast that to the 30 points I got for my Talent-Constrained post 🤦).

In addition to “making thoughtful posts”, some people also wrote criticisms, on research put forth by EA organizations. Both Saulius and Micheal St. Jules, seem to have been providing feedback voluntarily to the organizations they ended up joining for internships.

Some other people seemed to have applied like crazy and eventually made it. Micheal Arid applied to ~30 organization positions before getting in eventually into 2 organizations. He suspects that doing all these interviews might have helped him understand what the interviews required him to do. He shares some of the things he thinks might have helped him here where the main point seems to be that he read and consumed “practically everything EA” (Less wrong articles, books recommended by top EA people, all EA conference videos, etc.). He now has a staggering 7k karma points in just 3 years. The most awesome thing about Arid is that in 3 years he went from ~28 rejections to a Future Of Humanities Institute Researcher.

In Summary, writing “many well researched posts” on the EA forum, voluntarily providing feedback on the research to EA organizations, applying to more than 20 organizations, and consuming a ton of EA content, seem like good things to do.

Offering services as an intern or assistant

Another way of learning how to become an EA Researcher is probably offering services as an assistant to a researcher or taking up an internship in EA organizations.

Saulius seems to suggest that after his first internship (at Animal Charity Evaluator) “he knew what sort of articles would be useful”. He also gained relevant connections, with whom he collaborated on two articles which eventually got him noticed at the Recruiters of Rethink Priorities.

Another example is Nicole Zok, who started in GiveWell as a conversation notes writer and grew to the level of a Research Analyst in 2 years at GiveWell.

Asking people what steps/skills to take

I asked people who are Researchers in EA organizations, the following: “What steps could I/someone take to become GiveWell/Rethink Priorities/Open Philanthropy material?” or “What skills do I need, to become GiveWell/Rethink Priorities/Open Philanthropy material?”

For this I interacted (through direct messages or video call) with Peter Wildeford, Micheal St. Jules, DW (Researcher at a prominent EA org), AL (Junior Research at a prominent EA org) and ST (Senior Researcher at a prominent EA org). The following are the steps that they have suggested (and I paraphrased). More information on who said what is in the appendix. I have also (for lack of a better way) sorted the skills below by number of people suggesting it.

  • Read and write independent research (>=2)
  • Practice making cost-effectiveness estimates (>=2)
    • Practice discounting for various uncertainties (e.g., what is the effect size for the target population compared to the one described in the paper) (>=2)
  • Read social science papers and identify the strengths and weaknesses based on data quality, methodology, internal and external validity, assumptions (>=2)
  • Do MIT courses on GH&P: The Challenges of Global Poverty and Foundations of Development Policy: Advanced Development Economies (1)
  • Be good in Statistics and Causal inference (do courses) (1)
  • Be good in writing conversation notes (1)
  • Be good at quickly evaluating interventions (1)
  • Forecasting (1)

Note: The above skills are not representative of the view of the all EA organizations (as the sample people were not selected randomly). But amidst that uncertainty and the lack of a better way to get this knowledge, I accept the above information and will act on it.

How to get “better” at skills

Reading academic papers

When thinking about something like reading academic research and commenting on the internal validity, I can make comments all day, but I don’t think there is any point unless I get feedback. Am I doing it right or am I doing it wrong, and how could I do better! And if I need to send my work to somebody in order to get feedback then that takes weeks. Is it possible to get faster feedback?

Here is a great blueprint on Deliberate Practice called the Ben Franklin method. Basically, the author wants to improve his writing. He takes an example paragraph from someone he reveres (Paul Graham in this case and his post on “Schlep Blindness”). The author makes some notes on a paragraph and then writes his own version of “Schlep Blindness”. Next he evaluates his and Paul’s performance against certain dimensions (such as “do the sentences begin with subjects and verbs?” etc.). This way he gets feedback on how good he is compared to Paul, and what he can do to improve his performance. A key aspect about Deliberate Practice is that the practice can be repeated a lot. Paul has written more than 100 essays which means the author can pretty much practice like this forever.

When thinking about “reading academic papers and commenting on say their internal validity”, a similar approach might work well, I think. The Deliberate Practice would thus involve:

  1. Internal validity comments from someone who is revered (aka ‘the reference’),

    For this, Salius’s posts mentioned above (1 and 2), High karma posts in the EA forum, and the huge pool of Intervention reports from GiveWell (and also this), can be used.

    Each of these posts/reports cites many academic papers, and also has some commentary on internal validity of that respective paper.

  2. My comments on internal validity for those very same academic papers.

  3. Dimensions to evaluate against,

    Here there is a list on what constitutes good internal validity, such as “low attrition rate”, “Random sample collection”, “Randomization of participants to treatment and control groups”, “has there been experimental manipulation” etc.

The plan is to use these dimensions to evaluate the internal validity commentary of both my comments and the reference. This way instant feedback can be obtained after evaluating each paper and it is possible to see what the reference solution looks like as a way on how the commentary should be done.

I’ll admit that it’s hard for me to tell for sure if I am evaluating the dimensions correctly. For example, when I think of the dimension “experimental manipulation”, most likely I wont spot it. But there is also the reference to compare with. Hope this is good enough.

The same idea can probably be used for many of the skills listed such as “reading academic papers and commenting on external validity, methodologies, data collection” and perhaps even “quickly evaluating interventions”. However, the dimensions to evaluate the work needs to be identified. I think there should be some general links on these I can find online and use them (Will be done later).

Cost-Effectiveness Analysis:

There is another model of Deliberate Practice called the “case method” (also the “chess method”). In chess, isolation of an aspect seems a little out of hand, as key aspects are probably hard to find. But, with chess, thousands of books on different positions, different games have been published. The Deliberate Practice routine would be to look at a particular position and choose a move and compare it to the move chosen by the master (‘the reference’). Followed by analysis.

For cost-effectiveness practice, I think this method will serve well as I am not sure how to isolate different aspects of it. I would in this case, look at an intervention report and for each cited article, predict a factor in the cost-effectiveness analysis. And then immediately compare with the actual cost-effectiveness analysis. And repeat.

The cost-effectiveness practice session would involve the following:

  1. Obtaining content that is revered (‘the reference’). This can be obtained from GiveWell. There are enough intervention reports and associated Cost-Effectiveness estimates such as this and more are also available here.

  2. Reading intervention reports and predicting factors/values for the cost-effectiveness analysis.

  3. Comparing the actual vs prediction for each factor.

I think this practice would also cover “how to discount for uncertainties” as we are going to be doing the discounting in the cost-effectiveness analysis.

Writing/Doing Independent Research (with “good reasoning transparency”)

For getting better at writing independent research, first I need to know what good research looks like. I suspect the “concrete thinking practice routine” will help here. Concrete Thinking is about thinking in terms of claims and checking them with at least one example. In this post, I start with looking at text by someone (me, 80000hours or EA forum post). I look at each claim that is made, and find an example for it. I check this to the best of my abilities across dimensions such as “Does the example given match the definition of the claim”, “Is it possible even to give an example” etc.

A limitation of this method is that it is completely up to my own discretion to note if I am right or wrong. To counter this it is often useful to send it to a knowledgeable friend for feedback (“longer-feedback loop”). This way, I do have my immediate feedback and I also check how good this feedback is, over some time.

I hope to apply these principles to “writing independent research” as well. For this I would need some text to evaluate across some dimensions. For “good independent research” text, (as mentioned above) I can look at Salius’s posts (1 and 2), High karma posts in the EA forum, and the huge pool of Intervention reports from GiveWell (and also this).

For dimensions across which I measure the success of such posts, I hope I can pick up points from good reasoning transparency, “how we evaluate a study”, and some notes by Micheal Arid on high quality research.

Now using the text and dimensions I can evaluate “how good an article is”. I can thus evaluate how the greats do it. I can also do the same for posts I have written and see how much I fall short on these dimensions.

After doing a “few” examples of this practice, I expect to write short pieces of research and check it against the above dimensions. As for the “longer-feedback loop” I can also post these pieces to EA forum or ask specific feedback from particular people of the EA community.

Statistics, Causal inference and GH&P courses

As for GH&P, a research analyst from a prominent EA organization (AL) told me that she has done these courses from MIT in person and that she recommends doing them online. Some of these courses are by Nobel Laureates (e.g., Esther Duflo teaches Data Analyses for Social Scientists). I think I will take at least one of these courses (for lack of time) to get introduced to the “framework of doing research for GH&P” as it seems like it will be useful.

For statistics I don’t want to plan anything specific now. If I am going to be reading academic papers anyway, then I will know what sort of statistics I would need to do, and I hope I can prepare from there on.

As for causal inference, I really don’t know what I can do. I shall check with the person who suggested it.

Writing conversation notes

For writing conversation notes of talks, and to make a Deliberate Practice session out of it, I would need conversation audio or transcripts, and the conversation notes, from a source we revere. Unfortunately, conversation notes like these, only have the final notes and not the transcripts of the conversation.

I can maybe extract dimensions from several of these conversation notes. Following which, I can practice on my own on other EA talks on Youtube and check against the dimensions. Perhaps I should reach out to someone say from GiveWell, to give me feedback (“longer-feedback loop”). Perhaps!

Forecasting

I don’t think I will do anything about this now. Should I? I don’t know. I’d rather focus now on writing independent research on GH&P or animal welfare.

However a quick look at Metaculus shows that there are tons of past forecasts and future forecasts to practice on. Smell ‘Deliberate Practice’, anybody?

Other things I could do

My plan to get better

So far it is clear am not already GiveWell/Rethink Priorities/Open Philanthropy Material from my applications. So it seems like I need to do something to get better than I am now, before I apply.

So I propose to work on the things people from EA organizations have suggested, i.e., to start with practicing reading academic papers and making comments on various aspects, followed by working on cost-effectiveness analysis practice, and slowly climb my way to writing independent research on the EA forum (2 posts at least). (However, AL seems to have suggested that independent research might not be necessary. Then maybe it is wise to apply before starting independent research).

On the side I want to do one course from MIT on GH&P and continue consuming content (e.g., try reading Eliezer Yudkowsky’s series for example or take apart top forum posts in ‘concrete thinking’).

After spending a few months (3-6 months) on the above, I hope to re-apply (while maybe providing feedback on the works of other organizations), and also considering volunteering opportunities and/or internships.

Perhaps one thing to do would be to get feedback on this plan from someone like Micheal Arid (who can be reached on a call here) or the EA forum. Micheal is the guy who went from 28 rejections to a researcher at Future Of Humanities Institute in 3 years (Of course that is not telling the entire story). But still, wow!

How sure am I that I know what skills to train?

I am “sort of sure” about the skills I need to work on, as it came from the mouths of some of the people in the business. Looking at people like Saulius and Micheal St. Jules, and the time it took them to blossom, I think 1-2 years seems “realistic” before I get any internship or a full-time job in EA organizations as a Researcher. I guess!


Appendix: What did people do to get into EA

Peter Wildeford worked for over 5 years doing EA research by himself after which he was able to co-found an EA org. One thing Peter seems to have done is write a lot of Forum posts (while “aiming to emulate good reasoning transparency” and the style of highly up-voted posts). He has a very high karma of 7000 points. (Contrast that to me with 154 points). He has written around 38 posts in the EA forum before becoming Co-founder at Rethink Priorities.

Saulius seems to have started of with an internship in ACE after which it didn’t convert to a job. A year later he works for Rethink Priorities. He was rejected by other organizations too. Meanwhile with his experience and connections from ACE, he wrote two “intensely researched” articles (one published on ACE website and the other on EA forum), which seemed to have grabbed the attention of recruiters at least. He also has a high karma in the EA forum of about 3000 points. Prior to getting into Rethink Priorities he had written 3 posts. Note: One has to also do well in the tasks and interviews to get into the organization.

Gregory Lewis applied for a research role in GWWC and got rejected back in 2015. Since mid-2017 he works for FHI. In the meantime he continued to work as a Public Health Doctor and also appears to have been writing on the forum. He suspects his current role at FHI was “mainly because” of his public health background than his “forum ouevre”. He also has a high karma 3000 points unsurprisingly.

Micheal St. Jules applied to GIVEWELL and ACE in 2016 and didn’t get the roles (when he was new to EA). In the recent past he started working with his local EA group, started commenting and writing in the EA forum which was still not enough. He got an internship in Charity Entrepreneurship, following an EA forum prize, attending EA global and meeting people from Charity Entrepreneurship. He has over 1000 points and currently is a visiting fellow (internship) at Rethink Priorities.

Micheal Arid got ~28 rejections in 2019 before getting 2 job offers. After a year of experience he was able to apply for 11 positions and got 4. In the mean time he seems to have written many posts on EA forum. He shares some of the things he thinks might have helped him here where the main point seems to be that he read or consumed “practically everything” (Less wrong articles, books recommended top EA people, all EA conference videos, etc..). He now has a staggering 7k karma points in just 3 years. The most awesome thing about Arid is that in 3 years he made it to FHI. :)

In summary,

  • write “well” (high karma posts) in the EA forum and “a lot”.
  • “Consume EA content”
  • Attend EA global

Appendix: Suggestions on skills/steps to be taken

Peter:

Peter basically told me what’s written here.

Peter Wildeford suggests that the best thing applicants can do to improve their odds of being a successful researcher is to read and write independent research for the EA forum with good reasoning transparency and get feedback from the community. In addition he says it may also be good to get relevant credentials to perform research at such organizations e.g., “top forecaster on Metaculus or Good Judgement Open”.

DW is a Researcher at prominent EA org

DW is pursuing an economics PhD and was part-timing at a prominent EA org. His suggestions were:

  • Be able to “read” social science papers (academic papers in Lancet, Nature or Science) and “comment on and understand” their methodologies, data quality, internal external validity, weaknesses and understanding assumptions.
  • Have a solid understanding of statistics and causal inference
  • knowledge how to discount for trustworthiness of a paper

And perhaps consider the following steps to get there:

  • Consider getting a relevant degree
  • Can do online courses for statistics and causal inference, textbooks and material where you have lots of questions and feedback.
  • Start a reading group where you can discuss with other peers about academic papers for example.

Micheal St. Jules suggests the following:

Micheal St. Jules is currently a Research Fellow at Rethink Priorities. He suggested the following:

  • Try starting with an internship (instead of a full-time job).
  • Provide feedback on the research of organizations, publicly or by asking for opportunities to review unpublished work.
  • Do a literature review or Cost-Effectiveness analysis on some EA-related topic (like that of Saulius’s posts) and post on EA forum
  • Sign up to organizations that allow you to be a reviewer of their research (e.g., Charity Entrepreneurship seems to have some opportunities now).

For “Data Analysis type” roles:

  • For survey related roles use Mixed-Models and reproduce “EA Survey Analysis” and find other insights from the data.
  • There is another survey data set which needs analysis, and he can help me with the getting the data.

AL is a Jr. Researcher at a prominent EA org

The Researcher job at GiveWell-type-organizations involves trying to gauge evidence, piecing together poor evidence, and working with open-ended questions. She thinks it is a “good idea” to practice with the different reports on GiveWell (look in “Prioritized List of Programs” and “other charity reviews”), the following skills:

  • Understanding academic research (how we evaluate a study), the strengths and weaknesses and guessing the effect size for the target population.

  • Cost-Effectiveness analysis (e.g., “simpler” Cost-Effectiveness estimates such as this and more are available).

  • Communicating in a transparent way probably what is promising and what you are uncertain about (again GiveWell reports show how it is done and there are at least 61 of them).

  • She has done courses from MIT in person and said she “benefited” from doing courses like this (available in edX).

  • Also the ability to state the next steps, communicate clearly the uncertainty during work trials, is important. This will be very important also in the work tasks at GiveWell.

  • “I think The Challenges of Global Poverty and Foundations of Development Policy: Advanced Development Economies would be helpful for having some background on development programs and the evidence bases behind them.”

  • “I think deliberately practicing based on the GiveWell material is a good start, and I don’t think it’s necessary or particularly helpful to do independent research. Once you’ve looked at some of the materials and feel like you have a decent grasp on how GW generally thinks about programs, I’d encourage you to just go ahead and apply!”

ST is a Sr. Researcher at a prominent EA org

Skills:

  • Ability to read Cochrane meta analysis, literature pubmed, scholar, econ papers
  • Cost-Effectiveness (botec –> back of the envelope calculations)
  • Doing science with uncertainty
  • Quickly evaluating interventions (how many people have the disease and how much would it cost per person)
  • writing conversation notes
  • Critical thinking and communication in writing.

And perhaps take the following steps,

  • Practice writing conversation notes like these, although the recordings are not available, try to emulate the style
  • There seems to be lot of material on GiveWell website which I can use to perhaps train on.

SJ is a Sr. Researcher at a prominent EA org

Skills:

  • Gain more experience in Bayesian reasoning, e.g., thinking in terms of priors and making small updates
  • Cost-effectiveness analysis

  • Also he doesn’t think based on my current application to his org that I am “likely” to get into his org (in the “future”).

  • Also consider research-adjacent roles such as advisory roles etc.

Summary

  • read and write independent research
    • Write in EA forum with “good reasoning transparency” and get feedback from the community. (Peter)
    • write reviews, Provide feedback on research of organizations (e.g., Charity Entrepreneurship seems to have some opportunities now. (Micheal St. Jules)
  • Practice doing Cost-effectiveness estimates (Micheal St. Jules, AL and ST)

  • Forecasting
    • Get relevant credentials in Metaculus or Good Judgement Open (Peter)
  • Read Social Science papers
    • from Lancet, Nature or Science and “reflect” on methodologies, data quality, internal and external validity, weaknesses and assumptions. (DW and AL)
    • Cochrane meta analysis, pubmed, scholar and econ papers
  • Practice discounting for uncertainty
    • discounting for trustworthiness of a paper. (DW)
    • guessing the effect size for the target population from the papers (AL)
    • doing science with all the uncertainty i.e., you will not have the values you need (ST)
  • Be “good” in Statistics and causal inference (DW)
    • Do online course to get solid understanding of it.
  • Get a relevant degree in Economics for example (DW)
  • communication
    • Write with “good reasoning transparency” (Peter)
    • Communicating in a transparent way (AL)
  • Critical thinking (ST)

  • writing conversation notes (ST)

  • quickly evaluating interventions (ST)

Statistics

5k words
40 hours
>4hrs per day on average