Hide table of contents

Employers usually prefer to hire employees with credentials. The most common example of this is having a university degree: employees with university degrees receive significantly higher compensation.

Universities are very expensive, and it would be nice to replace them with some cheaper way of identifying talent. For example, maybe Google could just go to high schools, give everyone some two-hour long exam, then offer those who pass the exam six months of training at Google, after which they become Google software engineers.

The fact that Ivy League universities still exist seems like prima facie proof that this isn’t possible: corporations could make billions of dollars if they could do an end run around universities and hire people right from high school, so the fact that they don’t do this must indicate that it’s a hard problem.

Even if it’s hard in the general case, though, there might be specific subsets of employees or jobs for which we can avoid universities. Coding boot camps, for example, seem to demonstrate a credentialing system for software engineering jobs which is just as effective as existing ones but massively cheaper.

In this article, I explain the concerns with existing credentialing systems and try to speculate on which idiosyncrasies of EA would enable us to avoid them.

Why Credentialing is (Sometimes) Bad

One simple solution to identifying talent is expert deferral: if I’m hiring a software engineer, I could just say something like “if Google offers you a job then I will also offer you a job,” thereby deferring to Google about hiring standards.

This is not a terrible strategy, but it has the downside that we are, by definition, competing with Google for every employee. Similarly, if we recruit only from Ivy League colleges, we are competing against a bunch of other employers who recruit from Ivy League colleges.

We might prefer to find someone with good open-source contributions who didn’t finish college (as I almost did), or graduated from a state school (as I did). These people might be better fits (in the specific case of programmers, open source contributions are arguably more predictive than educational pedigree), but even if they are less skilled than the average Ivy League graduate, we have much less competition for recruiting them.

Secondly, credentials don’t exist for many things that EA’s consider important. Someone researching existential risks may come across infohazards, and having a PhD in no way certifies “this person will avoid publishing infohazards even if publication would advance their career.” Therefore, we need an alternative certification process.

Finally, the EA community currently contains a lot of people who are early in their careers and have not had the time to build a track record of success. There are legitimate concerns with putting people into positions of authority without credentials, which means that these early-career people have a lot of difficulty contributing.

What are some ways EA could take its unusual requirements into account, in order to hire more effectively?

Hypothesis 1: Hits-Based Hiring

Established companies are very risk-averse. Boards will pay a large premium to hire a provenly mediocre CEO versus taking a chance on an untested CEO who might be really good but also might be really awful.

Certain types of EA work are possibly more hits-based, such that the organizations doing that work might prefer the “untested, high variance” employee to the “tested and mediocre” employee.

As an example, some types of research might fall into this category: if the research project is done poorly, it will just go into the dustbin as an un-cited paper, but if it succeeds, it could define a new field.

For these positions, it seems like we can rely less on credentialing in favor of a “spray and pray” approach where we hire a lot of people with the expectation that most of them will be bad fits but the few who are successful will cover their costs.

Rocket Internet’s model of starting vast numbers of companies and then shutting down the ones which aren’t successful is one for-profit example of this strategy. (It’s worth noting that Rocket has a horrible reputation as a result – it turns out that repeatedly laying off massive numbers of people within six months of hiring them doesn’t make you popular.)

Hypothesis 2: Altruistic Internships

Suppose a corporation feels that someone with no relevant background could be brought up to speed after completing a two-year training program. The corporation might think of some strategy like: train someone for two years (during which the company loses money on them), but then have them employed at below-market salaries for three years (during which the company makes money on them). This would seem to benefit both the corporation and the employee.

In practice, this doesn’t usually happen because there’s nothing to prevent the employee from leaving after they complete the training program and going to a competitor which pays market rates. Corporations sometimes try to put a “poison pill” in their contracts, forcing employees who leave before a certain date to pay money back to the corporation. But it’s hard to legally enforce these contracts.

Altruistic employers might not have this problem: if I train someone to become awesome at corporate outreach campaigns, and then they go do corporate outreach campaigns somewhere else, that’s kind of okay?

More specifically: altruistic funders would be okay with this. If The Humane League has unusually high HR costs because their skilled employees leave to go work at Mercy For Animals, donors to THL might be happy to subsidize that cost because the donor is almost indifferent between work at THL versus MFA. Or a funder might support a grant for someone to work on a problem apart from either organization.

Hypothesis 3: Avoiding Negative Salaries

The US has a cultural and legal norm that employees must be paid if they provide value to their employer. University students, for example, do not need to be paid for their studies (and indeed usually pay for the privilege of studying) because their studies do not benefit the University.

This prevents a lot of mutually beneficial apprenticeships: young apprentices might be willing to work for free or even pay to help more skilled workers, and the skilled workers might be happy to have free labor. But this is culturally and often legally prohibited, e.g. by minimum wage laws. For example, shortages of surgeons are sometimes blamed on the fact that hospitals are required to pay surgical residents, even though the market equilibrium would be for the residents to pay the hospitals.

Nonprofits are not as shielded from these requirements as you might expect – there are laws preventing volunteers from doing the work that employees would otherwise do. But, as in the previous section, nonprofits can coordinate with funders in ways for-profits can’t. For example, a donor could subsidize someone’s employment through a donation to their employer, letting the nonprofit avoid taking a risk on their new employee while still complying with minimum wage laws. The donor is thus effectively investing in the apprentice’s employment, with the hope that the apprentice will provide positive value later on.

Another approach is entrepreneurship: the self-employed are generally exempt from minimum wage laws and often lose money on their companies – if EA has valuable projects which can be worked on independently, people could do that instead of more expensive credentials.

Conclusion

It seems like there is some momentum in EA behind these approaches. BERI’s Individual Level-Up Grants is one relevant program; some EA Grants recipients have been funded to gain important skills through independent study. Eliezer Yudkowsky, the founder of MIRI, is an extreme example of success without credentials, having attended neither high school or college.

Given that EA has a lot of people with raw capabilities but relatively few credentials, I would be excited for more people to explore this.

I would like to thank Aaron Gertler for reviewing a draft of this. Any remaining mistakes are mine. Views expressed are my own, and do not represent my employer’s. I have not been involved in CEA’s hiring, nor in any rounds of EA or Community Building Grants. This post is related to several recent ones such as EA is Vetting Constrained but was written several months ago and not influenced by them (it just took me a while to get around to publishing).

Comments8


Sorted by Click to highlight new comments since:

Having a reliable work-sample test and interview process allows me to hire fairly confidently without much regard to credentials.

At basically every company in the world, someone who comes in with a 10-year-long track record of success is going to be way more likely to be hired than someone fresh out of college, even if they are equally skilled. It would be pretty surprising to me if you are able to outperform the hiring practices of all these companies - do you have a sense of why you think you can? Many orgs use work-sample tests, so that seems too simple of an explanation.

I can't reply on behalf of Peter, but I would imagine the following:

  • Individuals at companies choose to hire for reasons other than expected performance (e.g. it's a publicly defensible decision to hire someone with recognized credentials and track record, whereas it's not publicly defensible to hire someone who lacks those but who otherwise seems like they'd perform really well). See general discussion of the signalling value of education.
  • Individuals at companies are bad at hiring for expected performance: e.g. relying on things which the evidence suggests don't predict job performance well (such as subjective impressions in an unstructured interview) and (possibly) credentials.
  • Many companies in the world can in theory hire people with 10 year track records doing similar roles in similar companies. People hiring for EA researcher roles typically can't find anyone with a 10 year track record in similar work- and even if you relax the assumptions somewhat, can still find far fewer people with any kind of track record in similar work.
  • The competencies Peter is hiring for may be more test-taskable than many that companies are hiring for. e.g. creating a cost-effectiveness model may be a better predictor of performance at creating cost-effectiveness models than the best available test tasks for "be an executive" or "manage HR."

I can reply on behalf of Peter and I approve this message.

Individuals at companies are bad at hiring for expected performance

Fair – an implicit assumption of my post is that markets are efficient. If you don't think so, then what I had to say is probably not very relevant.

Fair – an implicit assumption of my post is that markets are efficient. If you don't think so, then what I had to say is probably not very relevant.

I assume you're not arguing for the strong EMH here (markets are maximally efficient), so the difference to me seems to be a difference of degree than kind (you think hiring markets are more efficient than Peter does, Peter thinks hiring markets are less efficient than you do.)

If you are arguing for the strong version of EMH here I'd be curious as to your reasoning, as I can't think of any credible economists who think that real world markets don't have any inefficiencies.

If you're arguing for a weaker version, I think it's worth digging in to cruxes... Why do you think that the hiring market is more efficient than Peter does?

I very much like this approach to finding ways to deal with credentialism, however I'm also unsure how much of an impact credentials are having on current EA hiring. That is, my impression is that current EA orgs are hiring folks more based on work experience rather than credentials, and in fact EA orgs are unusually willing to consider candidates without traditional credentials (EA-orgs within universities being an exception due to their hiring processes being tied to those of the host institution). This suggests your premise may not apply (EA orgs not hiring folks because they lack credentials), but regardless I think your solutions apply anyway because they also address the issue where candidates lack experience rather than credentials.

Thanks for the feedback! I meant "credentials" to include things like work experience, and perhaps should have used more examples like that to be clear.

Curated and popular this week
 ·  · 11m read
 · 
Does a food carbon tax increase animal deaths and/or the total time of suffering of cows, pigs, chickens, and fish? Theoretically, this is possible, as a carbon tax could lead consumers to substitute, for example, beef with chicken. However, this is not per se the case, as animal products are not perfect substitutes.  I'm presenting the results of my master's thesis in Environmental Economics, which I re-worked and published on SSRN as a pre-print. My thesis develops a model of animal product substitution after a carbon tax, slaughter tax, and a meat tax. When I calibrate[1] this model for the U.S., there is a decrease in animal deaths and duration of suffering following a carbon tax. This suggests that a carbon tax can reduce animal suffering. Key points * Some animal products are carbon-intensive, like beef, but causes relatively few animal deaths or total time of suffering because the animals are large. Other animal products, like chicken, causes relatively many animal deaths or total time of suffering because the animals are small, but cause relatively low greenhouse gas emissions. * A carbon tax will make some animal products, like beef, much more expensive. As a result, people may buy more chicken. This would increase animal suffering, assuming that farm animals suffer. However, this is not per se the case. It is also possible that the direct negative effect of a carbon tax on chicken consumption is stronger than the indirect (positive) substitution effect from carbon-intensive products to chicken. * I developed a non-linear market model to predict the consumption of different animal products after a tax, based on own-price and cross-price elasticities. * When calibrated for the United States, this model predicts a decrease in the consumption of all animal products considered (beef, chicken, pork, and farmed fish). Therefore, the modelled carbon tax is actually good for animal welfare, assuming that animals live net-negative lives. * A slaughter tax (a
MarieF🔸
 ·  · 4m read
 · 
Summary * After >2 years at Hi-Med, I have decided to step down from my role. * This allows me to complete my medical residency for long-term career resilience, whilst still allowing part-time flexibility for direct charity work. It also allows me to donate more again. * Hi-Med is now looking to appoint its next Executive Director; the application deadline is 26 January 2025. * I will join Hi-Med’s governing board once we have appointed the next Executive Director. Before the role When I graduated from medical school in 2017, I had already started to give 10% of my income to effective charities, but I was unsure as to how I could best use my medical degree to make this world a better place. After dipping my toe into nonprofit fundraising (with Doctors Without Borders) and working in a medical career-related start-up to upskill, a talk given by Dixon Chibanda at EAG London 2018 deeply inspired me. I formed a rough plan to later found an organisation that would teach Post-traumatic stress disorder (PTSD)-specific psychotherapeutic techniques to lay people to make evidence-based treatment of PTSD scalable. I started my medical residency in psychosomatic medicine in 2019, working for a specialised clinic for PTSD treatment until 2021, then rotated to child and adolescent psychiatry for a year and was half a year into the continuation of my specialisation training at a third hospital, when Akhil Bansal, whom I met at a recent EAG in London, reached out and encouraged me to apply for the ED position at Hi-Med - an organisation that I knew through my participation in their introductory fellowship (an academic paper about the outcomes of this first cohort can be found here). I seized the opportunity, applied, was offered the position, and started working full-time in November 2022.  During the role I feel truly privileged to have had the opportunity to lead High Impact Medicine for the past two years. My learning curve was steep - there were so many new things to
Ozzie Gooen
 ·  · 9m read
 · 
We’re releasing Squiggle AI, a tool that generates probabilistic models using the Squiggle language. This can provide early cost-effectiveness models and other kinds of probabilistic programs. No prior Squiggle knowledge is required to use Squiggle AI. Simply ask for whatever you want to estimate, and the results should be fairly understandable. The Squiggle programming language acts as an adjustable backend, but isn’t mandatory to learn. Beyond being directly useful, we’re interested in Squiggle AI as an experiment in epistemic reasoning with LLMs. We hope it will help highlight potential strengths, weaknesses, and directions for the field. Screenshots The “Playground” view after it finishes a successful workflow. Form on the left, code in the middle, code output on the right.The “Steps” page. Shows all of the steps that the workflow went through, next to the form on the left. For each, shows a simplified view of recent messages to and from the LLM. Motivation Organizations in the effective altruism and rationalist communities regularly rely on cost-effectiveness analyses and fermi estimates to guide their decisions. QURI's mission is to make these probabilistic tools more accessible and reliable for altruistic causes. However, our experience with tools like Squiggle and Guesstimate has revealed a significant challenge: even highly skilled domain experts frequently struggle with the basic programming requirements and often make errors in their models. This suggests a need for alternative approaches. Language models seem particularly well-suited to address these difficulties. Fermi estimates typically follow straightforward patterns and rely on common assumptions, making them potentially ideal candidates for LLM assistance. Previous direct experiments with Claude and ChatGPT alone proved insufficient, but with substantial iteration, we've developed a framework that significantly improves the output quality and user experience. We're focusing specifically on
Relevant opportunities