Aug 15, 2016

Reflection on my time at GiveWell

March 2021 update: This post no longer reflects my current view of what it's like to work at GiveWell. Please get in touch if you would like to hear about my current views.

I worked at GiveWell from August 2014 to May 2016. This piece is a reflection on my time there, on things I think GiveWell does well as an employer, on things I think it could do better, and why I decided to leave.

I envision two functions for this piece: (1) as an exercise to help me process my time at GiveWell, and (2) as a resource for people considering working at GiveWell. When I was considering taking a job at GiveWell, I found Nick Beckstead's reflection on his internship at GiveWell to be very helpful. Outside of Nick's piece, there isn't very much substantive information available about working at GiveWell. Many people consider employment at GiveWell; I hope some of those people find this reflection to be useful.

Some background

I learned about GiveWell in Spring 2014, after reading Peter Singer’s Famine, Affluence, and Morality in a college ethics class and encountering related topics on the internet. By the time I took the ethics class, I knew that I did not want to go to graduate school immediately after my undergraduate, but I was very taken by academic ethics and wanted to continue serious thinking about the topic. GiveWell seemed like a good opportunity to continue thinking carefully about ethics outside of academia.

After a somewhat protracted application process, I decided to take a job at GiveWell. (A brief history of my application process: I was rejected two times, first because GiveWell was not hiring when I inquired, then because I failed the resume screen for a research analyst (RA) position. After my second rejection I was encouraged to apply for an admin position, during the admin application process I also did a case study interview for the RA position; I received an offer for the RA position after doing well on the case study.) Going into the job, I expected that it would help me think through several ethical questions, provide an opportunity to do impactful work, and be a good (i.e. well-regarded) first step out of college. Broadly, I think GiveWell delivered on all three.

While at GiveWell, I worked primarily on its top charity investigations. I spent time staying up to date on the progress previously recommended charities were making, as well as looking into promising programs run by charities we had not previously recommended. Specifically, I worked on GiveWell's investigations of the Against Malaria Foundation (which was a GiveWell top charity for several years prior to my joining), Development Media International (which we recommended as a "standout" charity for the first time shortly after I joined), and New Incentives (which has yet to receive a top charity or standout recommendation). I was not involved in Open Philanthropy Project investigations, nor did I do a substantial amount of intervention research (research focusing on what interventions are most helpful, rather than on what organizations are most effectively implementing priority interventions). The reflections below are based on my experience at GiveWell – I cannot speak from direct experience about what it is like to work on GiveWell intervention reports or the Open Philanthropy Project.

What I got out of my time at GiveWell

Worldview building

The main benefit I got out of my time at GiveWell was a more settled worldview. Going in, I was in exploration mode: I had many unsettled questions about the world, and few beliefs that I strongly endorsed. In college, I had been exposed to many areas of inquiry, but I didn't have opinions about most of these topics.

At GiveWell, I settled on answers to many of these questions. I now feel like I have a basic framework for thinking about the world and how I should operate in it. GiveWell enabled me to build this framework by (1) allowing me to work and converse with peers who had thought very closely about these topics, and (2) presenting me with real-world problems whose solutions required selecting and applying an ethical framework.

Here is a brief sketch of some of the worldview beliefs I now hold, all of which I decided on during my time at GiveWell:

For most ethical questions, I apply a consequentialist framework (though I remain suspicious of both thoroughgoing preference and hedonistic utilitarianism; when pressed, I lean hedonistic).

My meta-ethics are antirealist and pragmatic; I believe that, at base, a person’s choice of ethical framework is an aesthetic preference.

For figuring things out I am an empiricist (i.e. I believe that beliefs should be held lightly, should yield falsifiable predictions, and those beliefs that are falsified or can't be falsified should be discarded). I think skepticism is the correct default to hold when encountering untested knowledge.

In general I believe in keeping my identity small.

Professional skills

I gained several professional skills at GiveWell. I went from knowing very little about quantitative social science to being able to read a complicated paper and form a view about its quality (I am by no means an expert in this, but I am much more capable at it than I was prior to joining GiveWell).

On the object level, I learned a lot about developing world public health and some about development economics. I now have a good idea of how to answer empirical questions in these fields. For empirical questions in other social science fields, I now have a rough idea of how to assess the literature and arrive at an answer, though it would take me longer to spin up on the literature due to lack of previous experience.

I learned basic statistical concepts, especially those that apply to assessing the results of a randomized controlled trial. Note that I made a special effort to learn statistics while at GiveWell and it is not clear that I would have learned much statistics without making this effort.

My skill with Excel greatly improved at GiveWell. I had only basic knowledge of Excel prior to joining; building and working with GiveWell's cost-effectiveness analyses strongly developed my ability to use Excel for quantitative modeling.

My writing skill was strong prior to GiveWell. I think that it improved slightly, especially along the axis of precision (GiveWell writes in an idiosyncratic way that is very concerned with precision). My writing also may have become more verbose at GiveWell (another trait of GiveWell's writing style).

Finally, at GiveWell I became more persistent when trying to figure things out, and better at assessing arguments on their merits. Internally, I became more skeptical of my prior beliefs and began to develop a habit of thinking carefully about why I believed the things I did (as well as a habit of explicitly changing my mind when encountering new evidence or new arguments). Externally, I became more dogged when asking questions, and less likely to be satisfied with vague, underspecified answers. Along with this, I also lost almost all respect for arguments from authority, and now I try to not privilege arguments presented by experts over arguments made by laypeople.

Networking within the Effective Altruism community

I was unfamiliar with the Effective Altruism (EA) community prior to joining GiveWell. GiveWell provided an excellent platform for engaging with this community. This had both social and professional benefits: I now have many friends in the EA community, and when I was considering leaving GiveWell, three of the opportunities I seriously considered (including the offer I accepted) were enabled by my EA network. (I had multiple rounds of interviews for all of these positions. If I had not worked at GiveWell, and thus not had connections at these places, I doubt I would have received callbacks from any of them.)

Humanitarian impact

I think that my time at GiveWell yielded a positive humanitarian impact, though it is difficult to accurately assess the amount of impact appropriately attributed to my individual contribution. Parsing out the portion of an organization’s impact attributable to an individual contributor is a problematic exercise, and I’m personally skeptical of its usefulness, but some readers may be interested in such an estimate so let's give it a go.

As an institution, GiveWell is enormously impactful. I think in the final analysis, the bulk of GiveWell's impact will probably be due to indirect effects from changing how philanthropic incentives are set up and modeling an altruistic career path for young, high-potential professionals. However, it is especially hard to get a read on the expected magnitude of impacts like these, so for simplicity let's disregard them here and focus only on GiveWell’s direct impact.

A good proxy for GiveWell's direct impact is the money it moves to its recommended charities. In 2015, GiveWell tracked $110.1 million to its recommended charities. Most of this money moved was the result of work done in previous years, and the work done in 2015 will be used to inform recommendation decisions in future years, so there is not a direct relationship between work done in 2015 and 2015 money moved. However, I expect that GiveWell’s money moved in the next few years will be either about the same as it was in 2015 or higher, so for simplicity I’ll use 2015 money moved as an estimate of the impact of the work done in 2015; note that this is a conservative estimate.

According to GiveWell's 2015 year in review, an entry-level staffer in their first year is considered to add 10% of the value of a co-founder. I don't have access to the algorithm GiveWell used to model the relationship between co-founder equivalents and work experience, so let's assume that in 2015, my second year at GiveWell, I added 20% of the value of a co-founder. Further, as a research analyst, I was replaceable – let's assume that my working at GiveWell in 2015 had 10% more value-add than the person that would have been hired in my absence. GiveWell calculates that it employed 9 co-founder equivalents total in 2015, so the portion of 2015 money moved attributable to me would be: ($110.1 million / 9) * 0.2 * 0.1 = $244,700. Note that all the inputs to this calculation are very rough, and other values may arguably be more accurate.

It's tempting to assume that this figure is equivalent to me causing $244,700 of impact, but that would be double-counting. All of GiveWell's donations are made by donors, and many of those donors would have given to charity regardless of GiveWell's existence. I don't have good data on what proportion of donors would not have given charitably in GiveWell's absence, nor on what proportion of GiveWell's money moved would gone to less impactful charities in its absence. It's more appropriate to treat the figure as a proportion: i.e., whatever GiveWell's direct impact was in 2015, 0.22% of that impact was attributable to me ($244,700 / $110,000,000 = 0.22%).

If I wanted to put my proportion of GiveWell’s impact into dollar terms, I could make a rough estimate of the proportion of GiveWell’s money moved that ought to be attributed to GiveWell’s work (e.g. perhaps I estimate that roughly 40% of the impact of money moved should be attributed to GiveWell), then I could multiply through. (Following our example, $110,000,000 * 0.4 * 0.0022 = $97,900 of GiveWell’s 2015 direct impact would be attributable to me; from an impact perspective, I would be indifferent between working at GiveWell in 2015 and working at an impact-neutral job in 2015 and donating $97,900 to GiveWell’s recommended charities.)

Things I think GiveWell does well as an employer

Flexible working schedule. GiveWell suggested that its employees to work a "full-time week", which is intentionally vague but essentially means putting in around 35 to 45 productive hours a week (note that 35-45 productive hours a week equates to roughly 50-60 hours in the office per week). Where and when to put in these hours is left up to the employee. This flexibility is nice in many ways: it accommodates early risers and night owls, it lets employees avoid the rush hour commute (GiveWell's office is in downtown San Francisco), it makes it easy to schedule chores during the week, and it accommodates lazy mornings when going into the office feels unreasonably hard.

Access to co-founders and senior staff. GiveWell grew from 13 to 32 people during my time there. As a result of this growth, more levels of hierarchy were instituted (before I joined, the co-founders directly managed everyone; this shifted to distinct teams, each with a manager). Even at its larger size, GiveWell did a good job of maintaining access to senior staff. I had weekly, mandatory 1-on-1's with my manager, and co-founders held weekly office hours during which anyone could drop in to talk about anything. Less formally, it was easy to ask to meet with a co-founder or senior staffer, and these requests would almost always result in a 30- to 45-minute meeting (usually in the form of a walk along the Embarcadero).

Receptiveness to staff input. I think that GiveWell leadership were very good at soliciting and hearing feedback from other staff, including feedback on major decisions and on how GiveWell was doing as an employer. However, in general I think that senior staff were unlikely to change their minds as a result of staff feedback, and I think this was primarily because they engaged in very thorough thinking prior to soliciting staff feedback about decisions.

Clear communication about performance and trajectory. GiveWell communicated clearly about how I was performing and how it expected me to progress as an employee. I had ample opportunity to discuss performance feedback, and my manager made a point of periodically checking in about how things were going.

Hard-working culture. In general, people at GiveWell have a tendency to work very hard. This produces a cultural norm of working very hard, which in turn reinforces people's tendency to work very hard. Though it carries some risk of encouraging burnout, I think that overall this hard-working culture was a good thing. It felt good to work at a place that held you to an exacting standard, and I think spending time in GiveWell's work culture made me a more productive person.

Things I think GiveWell could do better as an employer

Top-down decision-making. During my time at GiveWell, most major decisions were made by the co-founders, or the co-founders in collaboration with senior staff. Decisions were then announced to all staff and staff feedback was used as a means of stress-testing the decision. In general, staff input had a low "hit rate" of changing decisions (note: my view here may be biased by my low personal hit rate of changing GiveWell decisions, other employees may have different opinions about this). I think this low hit rate is indicative of very solid decision-making by senior staff, but it could feel disempowering. Decisions often had the appearance of being made beforehand, and while decisions would be discussed after being announced, these discussions usually didn't change the plan very much, and the main contribution of research analysts was to execute on the original plan. As a research analyst, I generally felt that I had limited autonomy in my work (i.e. I usually felt that I couldn't do more than four hours work in any one direction without checking in with my manager), and this could be frustrating.

Time tracking. Everyone tracked their time at GiveWell. Most people opted to use Toggl, a desktop timer, though some people used more freeform methods (e.g. estimates logged in an excel sheet). Employees sent timesheets were sent to managers at the beginning of each week, as well as a more detailed report at the end of each quarter. Timesheets were used by senior staff to (1) get a sense of many hours different types of projects took, and (2) notice when a project was taking far longer than projects of its sort generally took, which could indicate a problem. Time tracking caused me a fair bit of anxiety during my first months at GiveWell, but I eventually acclimated to it. At my current job I am not required to track time, which I much prefer, though somewhat ironically I still track time on my volition (I think requiring time tracking introduces a bundle of small frictions and anxieties which cumulate into a substantial productivity cost; whereas voluntary time tracking doesn’t introduce many of these costs).

Performance evaluation. GiveWell’s work assignment philosophy is to give more challenging work to staffers who are performing very well, while keeping adequately performing staffers on the work they have been doing. After my first 8 months, I was assessed as an adequate performer (note that this is not used as a formal category at GiveWell), and once this assessment had been made I found it difficult to change. In general, poor performance on an assignment would result in a fairly large negative update on the staffer's overall performance, whereas consistently solid performance would be considered adequate, but would not result in a large positive update on overall performance (as far as I could tell).

I think GiveWell's method of performance evaluation works well for people who are already confident in their abilities and direction (i.e. people in execution mode), but can be quite costly for people who are still figuring out what they are good at and what they want to do (i.e. people in exploration mode). This is especially problematic because GiveWell has a tendency to hire young, high-potential, self-skeptical people, and high-potential, self-skeptical people have a tendency to believe that they can do less than they actually can, if challenged. I was very much in exploration mode when I started at GiveWell, and I think this resulted in a worse performance assessment than I would have received if I was more settled in my beliefs and more confident in my abilities. Once assessed as an adequate performer, I found it difficult to grow quickly, as I was not consistently given assignments that felt challenging.

Team building. I felt a lack of "teamness" during my time at GiveWell. I didn't realize how strong this lack was until I started at another job – I've been with my current company for about two months now, and I already feel a greater sense of team than I did at GiveWell. I have difficulty articulating precisely why a sense of team is important, but I strongly believe that it is. During my time there, GiveWell identified this issue and made efforts at increasing people's sense of connectedness. I think these efforts were moderately successful – things were better when I left than they were in late 2015 – but there is more room for improvement on this axis.

Lack of explicit emphasis on self-care. Many people at GiveWell maintained a strong division between their private and professional lives, and people were especially careful about infringing on their perceptions of other people's privacy. There are definite benefits to this, but one cost is that the organization did not explicitly encourage staff to take care of themselves.

I think that an emphasis on self-care is especially important in American work culture, where employees frequently feel that they should be working a lot and that time they spend not working is time they should feel bad about. Additionally, I think an emphasis on self-care is especially especially important when, as at GiveWell, the hiring pool selects for young people who might be convinced that they are morally obligated to do as much good as they possibly can with their time. Two things to note here: first, my impression is that most employers are pretty bad at encouraging self-care, and I have no reason to think that GiveWell is exceptionally bad in this direction; second, not explicitly encouraging something is quite different than discouraging it – in my experience, GiveWell was good at giving employees space and time when they asked for it, just not very good at encouraging people to make the ask.

Why I decided to leave

I decided to leave GiveWell because the object-level work didn’t align well with my interests, and because I found an opportunity which I thought would encourage faster personal growth (at Wave, a mobile money startup focused on African markets, which I joined as a non-engineer generalist). As I said, one of the main benefits I got out of GiveWell was worldview building – once I settled some of the open questions I had been chewing over, there was less building to be done.

I imagine that if I had a stronger interest in public health, or if I prioritized personal growth less highly, or if my disposition was slightly different (such that I found GiveWell's working culture to be more compatible with my preferred working style), I would have found medium-term (5-10 year) employment at GiveWell to be fulfilling. As things stand, however, my best guess is that I would have grown restless and unhappy at GiveWell if I had stayed for a few more years.

I got a lot out of working at GiveWell, and I have no regrets about my time there. I was able to work with amazing people who cared deeply about what they were doing, and I learned an immense amount. Through GiveWell, I was exposed to a community of people and ideas that aligned well with my own beliefs, and I was able to contribute to a sizeable humanitarian impact. As a first job after college for someone interested in an impactful career, or for someone interested in a career in public health or global development, I heartily recommend it.

Thanks to Michael Griffes, Ben Kuhn, Scott Weathers, Kit Harris, Vipul Naik, Holden Karnofsky, and Elie Hassenfeld for reading drafts of this post and providing feedback. And to Timothy Telleen-Lawton for kindly pointing out a calculation error after publication.

[rereads: many, edits: multitudinous]