Hide table of contents

EA-motivated specialists sometimes offer free or subsidized versions of normally expensive services to EA projects. I think this is often counterproductive and outline my reasoning in this post. 

The key problem with free services is that we don’t have market-based information about their quality, so the beneficiaries of a free service might be getting less value than it might appear they are getting. As a result, service providers waste their time providing expensive services to people who wouldn’t pay the full price (instead, providers could charge and donate[1] or spend that time on other impactful work). Additionally, community-level overestimates of the quality of free services are more likely and might lead people who need good services to use the free versions even when they’re worse suited to their needs. 

If you’re offering, taking, or advertising a free service like this, I think you should believe the situation is an exception to the general heuristic. (More on these problems and other issues, as well as exceptions and nuances.) 

Chart: a service is offered for free to EAs, leads to praise/advertising, leads to info cascades, leads to over-estimates of the value and quality of the service, and that leads to problems (waste of the service provider's time and people using a service that's worse than what they need).

Some services make sense as free services. For instance (see more):

  • Community infrastructure projects or other work where the benefits go through many people (there’s not a clear “beneficiary” of the service)
  • Services for which overhead costs are especially high
  • Services that are in large part beneficial to the provider, not the recipients of the service

Notes & caveats: 

  • Scope of the post
    • When I talk about “services,” I mean costly services that are often provided by specialists (coaching, consulting, etc.). I think more minor services are more often fine, and I’m definitely not talking about friendly support, like helping your officemate with an ughy task or having a call with someone you met at EAG who’s interested in getting into your field.
    • Certain types of volunteering don’t encounter the issues outlined here, particularly when the volunteer has the option to be paid a specified rate but opts out (so they know more about the “market value” of their free work).[3]
    • The concerns I discuss are also less relevant for services provided by for-profit organizations/groups that provide a free/subsidized version to e.g. nonprofits. 
    • I focus on free services here, but the arguments can probably be extended to cover subsidized services. 
    • I don’t discuss paid services that only serve EA customers, but I also have concerns about these services when it isn’t clear that the needs of EA projects covered are unusual.[4] My rough take is that:
      • People advertising infrastructure and support projects in EA spaces (e.g. newsletters) should not give preference to services that only have EA customers unless there’s something clearly special about the service that makes it especially useful for EA work.
      • People working on EA projects should find the available service that best suits their needs, which is generally unlikely to be EA-focused; most of the world is not in EA. (Maybe it’s easier to identify a good-enough service in EA,[5] but I still think we should be a bit more suspicious of our instinct towards going for EA things.)
  • I’m not speaking for my team or for CEA. 
  • I don't know how important the issues I outline here are relative to other issues in EA, but this topic has come up several times, so I decided to write this post. 
  • I don’t think people who provide free services are ill-meaning or anything like that. Also, I appreciate many people in the EA community for reasons that go beyond “I think they do impactful work.” I like to hang out with other people in the EA community for personal reasons. But if I’m thinking about doing something for EA reasons — because I think it’s an effective use of resources in order to help the world — I want to focus on supporting impactful projects.[6] I don’t want to support people in EA just because I like them. This is a relevant part of the premise of the post. 

Problems

It’s harder to evaluate the quality/value of free services, so (1) providers might keep offering free services that are a bad use of their time, and (2) people might use services that are not what they need

If a service costs $100/hour and is getting customers, we know that these customers are willing to pay at least $100 for the value they get. If a service is free, however, people would use it as long as it’s better than ~nothing (maybe they’d only pay $10) — even if it’s not worth the time of the provider.

A service provider who wants to dedicate some time for effective altruism might be choosing between providing their service for free to people in EA or charging for it in the normal market and donating the proceeds (or spending this time working on something else entirely). If the recipients of a free version of the service are getting less value than the provider could earn and donate (e.g. they’d rather get $30 than an hour of this service, and the provider could earn and donate more than that in the same timeframe), then the provider is wasting their time on the free version.[7]

Moreover, the service might get over-promoted in EA and people working on impactful projects might go for it instead of services that suit their needs better. 

As a result: 

  1. The provider might waste their time
    1. Sketch of how this happens: 
      1. People who get some value from the free service keep using it, even if they wouldn’t pay the sticker price. And they’re thankful to the provider and avoid criticizing the service. 
      2. The provider doesn’t realize that the value people are getting from the service is less than the value of their time. Instead of working on something else or charging for the service and donating the proceeds, they keep providing the service for free.
      3. Also, they don’t get good feedback about the quality of their service and don’t notice important improvements.
    2. Example: 
      1. Asimov offers free high-end cakes to people who run EA events. The cakes are expensive and Asimov is very skilled — he usually gets $300 in profits per order. He has a steady trickle of happy EA customers. However, these customers aren’t actually getting $300 in value from the cakes — they’re just happy to get free cake for their events. (Also, some of Asimov’s vegan cakes are much better than others, but his customers aren’t telling him that since they’re not paying and don’t want to criticize.) Asimov really cares about improving the world and would donate the money or go work on something he has the skills for (like a cultured meat startup) if he got the signal that his work wasn’t as valuable as it seemed, but he’s not getting that signal, so he keeps offering free cakes for EA events.
    3. Related: sometimes people get really excited about EA and default to finding an “EA version” of the work they were doing before that happened. I think this is often (but not always — related[8]) a mistake — see 12
  2. The service might get over-promoted and people in EA might use it instead of a service that fits their needs better
    1. Sketch of how this happens:
      1. The service is advertised in EA spaces, since it’s free for EAs.
      2. People who use the service feel extra grateful because the service is free. They also don’t criticize the service — it was a gift. 
      3. A praise/info cascade emerges: we hear mostly good things about the service, we hear that people have used the service, and we think we’re getting independent signals about the quality of the service — so we think it’s really good.
      4. People looking for this kind of service default to the free-for-EA-projects version. They might not really know what to expect, and if the service is overall not super useful, they might just conclude that this kind of thing doesn’t work very well. 
    2. Example: 
      1. Asimov offers free cakes for EA events. This is advertised on the EA Forum and elsewhere. Ursula is hosting an event for potential donors. She could look up reviews for the best cakes in the right location, but she’s heard of Asimov’s cakes (maybe she knows they were served at a recent conference) and defaults to asking him for cakes. But then the cakes are subpar, or there’s some chaos getting them in on time. They don’t add the value that she was hoping cakes would add to the event. She still sends Asimov a thank-you email and doesn’t tell anyone that there were issues; the cakes were free and she doesn’t have time for this kind of thing, anyway. She decides that caterers like this are probably not really worth it and doesn’t try serving cakes again at events like this.

It might be more effective to charge the full cost and donate the profits

We can try to estimate overhead costs of charging and donating and compare those costs to effectiveness decreases from offering services for free.[9] 

  • A generous (I think) estimate of overhead costs: Let’s say that you usually charge $100/hour for your service and would pay 15-45% in taxes and other overhead on this. Also, let’s assume that your alternative is charing in the normal market and donating proceeds to the EAIF,[10] and that an additional 1-10% of your donation would go to overhead costs (e.g. grantmakers and others’ time). 
  • Conservative estimates of effectiveness costs: Let’s say that recipients would actually pay between $30 and $100 for your service. You need to vet people who’d get your free service.[11] Your system is probably somewhat worse than professional grantmakers’ (maybe between “as good as” and “five times worse”).

According to this rough Guesstimate (you can input your own values if you want), charging and donating proceeds probably produces around twice as much value as offering the service for free. If you assume that grantees would in fact pay your sticker price ($100), you still need to believe that grantmakers are only 1.2x to 1.9x as good as you are at choosing projects in order to conclude that providing the service for free is the better option. I think this is probably not the case for most free service providers to EAs, but it might in fact be true for something like EA Global travel support (attendees are vetted). So I think that many free services would be better as donations.

Chart: Option 1: services are free (service is provided directly to the recipient). Option 2: proceeds from services get donated. Service earns money, donated to grantmaker, who provides money to a recipient, who can they buy the service or go for other options.

More minor/complicated issues with services that are offered for free

  1. Worse movement-level tracking of the costs of different initiatives
    1. If Project A applies for funding to grantmaker Alpha, they’ll say something like “look, our work led to these outcomes, and our costs are X.” But if they’re getting a significant amount in support via free services, those might not be taken into account for evaluating the true cost of Project A.
  2. EA partisanship and insularity
    1. I’m a bit worried about people forgetting that EA is for the world (not for EAs), so I’m concerned that initiatives that seem like they support EAs just because they’re into EA nurture something like “EA partisanship” (and insularity). 
  3. Issues with the “service market” in EA
    1. Some potential service providers whose services would be really useful to projects in EA might not be able to offer them for free. If people are already using free services, newer projects won’t be able to compete even if they’d actually be better. 
    2. Additionally, people might underestimate the importance or demand for services, because no one is paying for them. Relatedly, you might not take your work as seriously if you do it for free. 
  4. Power issues,[12] grifting
    1. If you’re distributing free services in EA (especially when they’re very valuable), you have some power, like a grantmaker. You might not notice this as acutely as you should. (Related: Power dynamics between people in EA)

Situations in which it’s potentially reasonable to offer/advertise free services to EAs

Several of these can be true about the same service. Please note that I'm not arguing that if a service has one of the following qualities, I think that it should be free — only that the quality could be a reason for why offering it for free is more reasonable than would be the case otherwise.  

  1. The service is a communal good — many people are getting a small amount of benefit
    1. People share research on the EA Forum for free, although it can take hours of work per post. Should we charge readers/Forum users? (Let’s ignore the fact that part of the “benefit” is going to the authors via name recognition, etc.) I think no, in part because there’s no specific recipient of the service — the benefits are diffused through many readers and via supporting the existence of a platform where others want to post (using the Forum in a productive way supports a virtuous cycle). 
    2. It’s hard to get lots of people to pay a little for the pretty small amount of value that work like this brings them, and I think it’s ok for these “services” to be free. 
    3. (I think it’s still quite important to develop things like cost-effectiveness estimates and evaluations of projects like this if they’re large enough, though. We’re working on updating/improving ours for the EA Forum right now.)
  2. The costs of overhead+taxes outweigh the benefits of donating earnings to be distributed by grantmakers/other systems
    1. Maybe you’ve done the math on this for your service, in which case it might make sense to keep offering your service for free. (Although beware lack of feedback and accountability if you’re volunteering.)
    2. This is more likely if:
      1. Your service is pretty cheap, which probably makes overhead costs of charging bigger (as a percentage of the total cost). 
      2. It’s impossible or illegal to charge for your service (or you have some required amount of time you need to spend offering a free service).
      3. You find it much easier to offer your service to EAs.[13]
  3. The service makes its “customers” more impartially altruistic
    1. It can be hard to identify the “true beneficiaries” of services for altruists. The benefits of work in EA are ultimately going to the world (or they’re meant to be), not to EAs. So while we can talk about “services for EAs,” the reality is that we’re all hoping that people will use services that make them more effective and more impartially altruistic. Especially for the latter kind of non-value-neutral (“value-oriented”) “service” (e.g. introductory courses on EA, free EA books, texts that argue for the moral relevance of some species, or career advising which involves talking through values and seeing if they’re aligning with EA) we can actually think of them as services for others (shrimp, people who want to spread their moral theories, etc.) through the people ostensibly getting the service. Charging for these makes less sense.
  4. The service provider evaluates applications (like a grantmaker does)
    1. 80,000 HoursProbably Good, or Successif, etc. evaluate applications for advising (as far as I know); they’re not just providing the service to anyone who identifies as EA. (Those examples might also fall under 3.) 
    2. The average project supported by a service like this is probably more impactful. 
  5. The cost-effectiveness of the service is evaluated (e.g. by funders) in order to keep it running
    1. If grantmakers don’t think a service is valuable enough, they probably won’t fund it. So I’m less worried about services that rely on grants (e.g. larger-scale projects).
    2. Related, if you get really good information on the value of the service (and other feedback) in some non-market way (e.g. you collect really rigorous feedback), you don’t face some of the issues outlined here. (Or: you also have non-EA customers who are still using your service, and you get good feedback/value baselines from them.)
  6. The service is valuable in large part because it helps the provider of the service, as opposed to the people getting the free service
    1. Maybe you’re looking to upskill in a certain area and wouldn’t be able to get paid for this kind of work, but could find opportunities to do it for free.[14] It could make sense for you to work for free as long as you’re tracking that this is the true motivation. 
    2. I’d still caution against traps like overestimating the direct benefit of the service (because people are still using it), and/or continuing to provide it when it stops being as valuable to you.

Other nuances and counterpoints

  1. Decentralization[15]
    1. Say you want to support efforts to build EA. You can charge for a service you provide and donate the proceeds to e.g. the EAIF, but then you’re giving up some of your resource-allocation power and adding it to a central pool. I think this is a reasonable concern. But I also think there are other solutions, like donating to smaller grantmakers (e.g. 12) or participating in a donor lottery.
  2. Beware of fake alternatives
    1. Maybe the alternative to someone providing an expensive service to impactful projects for free isn’t that the provider would donate proceeds to EAIF or elsewhere — they’d just keep earning money. In this case, it’s more likely that it’s useful for them to keep providing the free service. 
    2. Similarly, maybe people using mediocre or ill-fitting free services wouldn’t pay for a version that’s more suited to their needs — they just wouldn’t use a service at all. 
  3. A kind and unselfish community
    1. It’s just nice to have a supportive community that does things for each other without expecting something in return. My take is that we should accomplish this via things that mostly aren’t free versions of expensive services — related discussion here. But there’s probably more nuance here. 
  4. Valid concerns in theory, but don’t realize in practice?
    1. Maybe what I’m outlining here makes sense in theory but real examples of free services in EA don’t face these issues. I think this is plausible (I’m basing this post off a few second-hand experiences and some conversations, but I don’t go into specifics and haven’t tried to comprehensively list and analyze real examples). I’d guess that some of the issues I’m describing are real (hence the post). 

Thanks to Jonathan Michel and Lorenzo Buonanno for helpful comments and conversations! 

  1. ^

    It might be worth pointing out that a lot of projects and fields would benefit from donations right now. 

  2. ^

     Helping is good.

  3. ^

     You might be interested in “Volunteering isn't free” though. 

  4. ^

    It seems possible or likely that I’m missing something, though. 

  5. ^

    This post talks about how people go for Subway instead of local restaurants because they can assume that there’s a minimum quality level they’ll get from Subway. 

  6. ^

    and the careers of people who might do particularly impactful work (a key reason for internships)

  7. ^

    Note that it might help to straightforwardly ask people whether they'd prefer an hour of your service or a check for whatever amount you could donate. 

  8. ^

    In short: working in a field where you have amazing skills can often make sense, and it’s possible that the right “EA” opportunity will come along. And you have a lot of information about your fit and comparative advantage, etc.

  9. ^

    In case you prefer a janky visual representation: 

  10. ^

    if you’d donate somewhere else, that’s because you think that work is better, so this is conservative

  11. ^

    Not all “EA” things are good, so vetting would make the average project you support more impactful (probably). You could get people to submit thorough applications, but that costs time and might make it harder for you to get customers. You don't specialize in this and would probably need something relatively simple so I would guess that you’re probably at least somewhat worse at this than grantmakers are.

  12. ^

    Thanks to Jonathan Michel for this point. 

  13. ^

    Maybe you're energized after talking to EAs, so it's less costly for you; overhead is lower. (Although I'd tentatively recommend that some people read Purchase Fuzzies and Utilons Separately.)

  14. ^

    if you have the resources to work for free; you could alternatively apply for funding.

  15. ^

    Another point from Jonathan. 

Comments23
Sorted by Click to highlight new comments since:

This is an interesting idea and I appreciate how you've sketched out which kinds of organizations it applies to. I think I agree with your major point: certain services that EAs can buy on the free market will be more efficiently provided by, well, the free market. Such services might include things like: therapy for depression or anxiety; general advertising services; video editing services; immigration law services. 

I would like to propose another set of services that should be free: products you cannot buy on the free market. 

To give an incredibly non-random example: you cannot get EA value aligned career coaching outside of the EA community. (Disclaimer: I work at 80,000 Hours and all of the opinions here are my own, not my employer's.) 

There are other reasons certain services should be free, and I don't know how these fit into your framework. For example: one of 80k's paths to impact is changing the career trajectories of highly skilled people who don't know much about EA; I want to lower the bar for these people to apply for coaching, not raise it by charging money. Another one of our paths to impact is helping boost the careers of undergraduates or people with low incomes; again, I do not want to cut off access to our services by charging money. 

Also, could you please clarify why: The service provider evaluates applications (like a grantmaker does), means the product should be free? I would assume most of the free services you're criticizing do indeed evaluate their potential clients before accepting them? Related, I'm assuming The cost-effectiveness of the service is evaluated (e.g. by funders) in order to keep it running, is also true for many of the free services you're criticizing, but I don't think it means these services should be free. 

Thanks for the comment! 

I'm not sure about "products you cannot buy on the free market." It seems possible that those more often have some of the other qualities I listed, though: they might be undervalued by the market for some reason (so they're probably cheaper, meaning that overhead would be higher if you tried earning a profit and donating it), or they're highly specialized to EA somehow, which would mean that looking for non-EA providers and customers isn't as viable. If those don't hold (and neither do other factors), I'm not sure why they make more sense as free services. 

On your specific example — EA value-aligned career coaching — I think the more relevant point for me is that the benefits of this "service" are not entirely going through straightforward "benefits" to its customers, since part of the point is to make the "customers" more impartially altruistic. This is what I was trying to cover in List item 3 here. I think you also describe this in the paragraph that starts with "there are other reasons..." 

Re: "why: The service provider evaluates applications (like a grantmaker does), means the product should be free?"

Quick clarification: In the section on exceptions, I'm not trying to actively argue that any particular kinds of service should in fact be free (for EA projects). Given that I argue that this is often not an optimal approach, I just wanted to point out that there are types of services for which I think it's more likely to be effective. And these are something like partial signals; if several of the list items apply, I'm less skeptical that offering the service for free is optimal. If none seem to apply, I'm pretty unsure. (I'll try to clarify this after I post this comment.)

Setting that aside: I do think that if you're offering a free service for EA projects (or for "EAs"), you could end up in a situation where you're mostly just going off of whether the person running the project seems to be ~in the community. (I think this happens sometimes, but I'm not sure.) I think that the effectiveness of work in EA varies quite a bit, so I do think that offering services for projects that are filtered for impact makes more sense. 

Relatedly, re: "I'm assuming The cost-effectiveness of the service is evaluated (e.g. by funders) in order to keep it running, is also true for many of the free services you're criticizing, but I don't think it means these services should be free."

This is another factor that makes me less skeptical that a free service is cost-effective (although it's not a slam-dunk for me), vs. a service that's being run based on the providers' own assessment that it seems useful or related situations. 

Larks
18
6
0
1

Thanks very much for writing this, I think it is a great post presenting a case I have made to many people in person.

I think you actually understate your case, because I think this exception you mention is quite weak in practice.

The cost-effectiveness of the service is evaluated (e.g. by funders) in order to keep it running

  1. If grantmakers don’t think a service is valuable enough, they probably won’t fund it. So I’m less worried about services that rely on grants (e.g. larger-scale projects).

When grantmakers evaluate projects, they often do so from a considerable distance. They lack all the facts on the ground that the service users have, and (at least for LTFF/SFF/Lightspeed) are trying to evaluate a large number of applications in a relatively small amount of time. Perhaps they get feedback from some service users; probably most of their information comes from the service provider. It is much harder for them to evaluate the service than it is for the users, because of the classic Hayek problems with centrally planned economies.

Additionally, by applying to grantmakers, you are increasing the amount of work they have to do. Your service users still have to evaluate your service - they need to decide whether to use it or not - but now in addition some very overworked grant evaluators have to struggle through it.

As a result of this, when I'm doing grant evaluation my reaction to many applications is 'this should be a for-profit. Either they haven't thought of this, in which case I should tell them, or people wouldn't pay, in which case they should close down.'

Relatedly, I think many people over-estimate how strong a badge of endorsement getting funding is. People should not treat getting a grant as strong evidence that a service is good; actual happy paying customers is a much stronger sign, even if those customers are much less prestigious than the grantmakers.

Do you [or anyone else] have an opinion about my project for free career coaching for EA developers?

 

I have mixed feelings about it myself

To me, based on what you said, you have provided a lot of value to many people at relatively low cost to yourself. I have the impression that the time was quite counterfactual given you didn't seem to have many burning projects at the time. So, seems pretty good to me on the face of it although for every given detail you know way more than I do!

I generally agree with the reasons given in this article about why you wouldn't want to subsidize your services to EAs or offer them for free, and I agree with the counter-reasons why you might want to do it anyway. I think I weight them differently from you, such that I'm more ready to believe that offering a subsidized or free service can be defensible.

I think I'm less convinced by market signals than you, and more convinced you can notice people making a systematic mistake in how they price something. Obviously there's a huge conflict of interest when you're making this claim about your own work, so you should really try to have some independent cross-check of your value, but I don't find it hard to believe that someone can nevertheless come to reasonably believe something like this is true.

For example, recently-graduated students (or even students still in studies) probably have a hard time justifying high expenses, even when they will eventually pay for themselves. Say you could deliver a salary negotiation class that for 50% of people raised their lifetime earnings by 10%. How much should they value that class, knowing what you know? How much should they value that class, given that they don't know what you know and can't necessarily trust you? How much will they be able to pay for the class, even if they wanted to? In abstract ideal economics land, you can sometimes fix issues like these by offering loans or grants, selling fractions of future earnings or whatever, but in the real world I think things like that will have pretty serious friction and you'll end up being able to deliver to nowhere near the same number of people.

Taking a step back, maybe the idea here is: you need some reason to believe that what you're doing is working. Different feedback signals have different virtues and drawbacks: people paying you is great because it's the prototypical costly signal, so people will tend to give you their honest assessment of your value, but it doesn't particularly force their honest assessment of your value to be correct. Other signals might have more or better information.

I think people considering offering a free or subsidized service to EAs should read this article first, but if they come away thinking they have good reasons to disagree, I'm willing to believe them.

I agree with most of the points in this post and have advocated for some of them within my organisation. I am particularly sympathetic to costly signals (like paying for services) providing a much stronger anchor for a service's value than feedback forms.

That said, the EA marketplace doesn't seem to be particularly efficient. I worry that many projects are under or over-funded, making the willingness to pay signal much more costly for some people than others (independent of the value the service would create through them) and making it hard for sellers to price their goods when they are primarily operating within the EA market.

Also, offering free services helps decrease organisations' dependence on funders, which I think is pretty desirable for grantees on the current margin.
 

I agree that the EA marketplace seems particularly inefficient, and agree that this is an argument against my key point above. My guess is that many people are still undervaluing the importance of willingness-to-pay signals, although there are probably some who don't appreciate the inefficiencies of the market. I'm not really sure what to do with that.

Re: "offering free services helps decrease organisations' dependence on funders, which I think is pretty desirable for grantees on the current margin." 

I think overdependence on specific funders (or on a small number of funders) can be bad, but my preferred solution is "diversify funding more" not "get free services" — to the extent that the first is an option. (I think I might be missing some reason for why it's bad for organizations to depend on varied funders, though, if you in fact meant that.) So I think here I'm more sympathetic to approaches like: if you're considering offering a free service or donating the proceeds, consider asking your potential service recipients whether they'd prefer to just get a donation, or consider donating to a donor lottery or to a smaller fund you appreciate, etc. I tried to outline something like this in point 1 here, but I might have communicated it poorly (or I might be misunderstanding you now). 

Here is a case where I'd appreciate your perspective: I sometimes provide "estimation as a service" as a consulting service for people who are making career decisions. This involves making some estimates about the likelihood of success and the value of different career choices. Some past examples are here (less elaborate) or here (more elaborate).

The problem is that  at free market rates, this costs me, say, $50 to $200 worth of my time. But this is generally too much for students or early career people.

I'd be curious whether you'd think that I should make an exception, or whether you think that the reasons you outline in your post hold.

I think there are good reasons to make an exception here:

  • There's a broader ecosystem value to having different perspectives available on career choice. So some of the indirect value of your services is realized wholly independently of the service user.
  • The risk of someone using your services merely because they are free, when they would be better served by a paid service, seems low.
    • The most obvious "substitute" for your services is arguably 80K, which doesn't to my knowledge charge for its services. 
    • This risk is further reduced because the students or early career people don't have the money to purchase these services at market rate.
  • Moreover, it would be very inefficient to ask a grantmaker to decide which would-be clients should get a ~$50-200 grant for estimation services. This seems different from the cake hypothetical, where the group could (and probably does) seek funding for all its needs and throw a cake line item in the budget.
  • The counterfactual seems to be that some people wouldn't get estimating-as-a-service at all. That seems undesirable on these facts in comparison to potential downsides that seem pretty attenuated.

Thanks Jason

Another great, provocative post. Thank you!

I agree with your rationale because of an experience in a kind of analogous situation. 

A scientist at the company I worked for (a multinational) invented a highly effective and scalable water-purification kit, which was ideal for developing countries with clean-water problems. At first, the company donated these kits. But then we realised that it would be far more effective to sell them (low price, but not free) - because that way, the people who really needed them would get them and use them, and they would not be wasted. Furthermore, we found that local entrepreneurs were able to set up businesses selling and distributing them, which was an added bonus. And, with the same net cost, the company was able to provide far more kits for far more people. 

And yet, this was highly counterintuitive for most people, and there were questions about "why can't we just give them for free?" from people who hadn't thought through the full scenario. 

I basically agree. I price my EA-focused-but-entertainment-only Substack at market rates, with exceptions for students and people who have taken the GWWC Further Pledge.

I think subscription-access is largely an exception to Lizka's reasoning, since the marginal cost to you of another person being able to access your writing is zero!

Her argument was it's harder to evaluate the quality of free services, so the provider might waste their time or people might use services that aren't what they need, and that it might be more effective to charge and donate the profits. I don't see how either of these don't hold in my case?

It's possible Zach's point is already obvious to you, but the one-sentence rephrase is that unless you have perfect price discrimination, charging subscription fees for something with a marginal cost of zero would result in deadweight loss. 

I don't know how clear the one sentence version is, so I asked GPT-4 to explain the idea in more detail.[1]

Alright, let's break this down with a simple example:

Imagine there's a magical factory that produces unlimited cookies at no cost. Once the first cookie is made, every additional cookie after that is free to produce.

Now, let's say the owner of the factory decides to sell unlimited access to these cookies through a subscription model. For a monthly fee of $10, you can get as many cookies as you want.

Here's the catch: not everyone values these cookies the same way.

  1. Alice values the unlimited cookie access at $15 a month.
  2. Bob values it at $8 a month.
  3. Charlie values it at $10 a month.

At the $10 subscription fee:

  • Alice subscribes and is happy.
  • Charlie subscribes and is neutral.
  • Bob doesn't subscribe because he values it at only $8.

Now, remember the cookies cost nothing extra to produce. So, if Bob were able to pay his $8 and get the cookies, both he and the owner would be better off. But because the price is fixed at $10, Bob misses out, and the owner misses out on Bob's $8.

This situation where Bob doesn't get the cookies even though they cost nothing extra to produce is called a "deadweight loss". It's a loss to society because both Bob and the owner could've benefited, but they didn't.

Price discrimination would be like the owner saying, "Alright, Alice, you pay $15; Charlie, you pay $10; and Bob, you pay $8." Everyone gets cookies, the owner maximizes revenue, and there's no deadweight loss.

  1. ^

    Incidentally, it's spooky how good GPT-4 is at explaining concepts in detail from just a one-sentence summary.

Another point worth considering is consumer surplus: in the $10 subscription model, Alice actualizes $5 in benefits over what she paid. Merely counting subscription revenue doesn't capture the surplus to the cookie-eater ecosystem.

So even if the EA service makes somewhat less revenue than the vendor would make on a non-EA gig, it could in some cases still be superior because the EA ecosystem captures the consumer surplus in the former case.

Sure, that makes sense, I was just surprised because it didn't seem relevant to the points raised in this article, which I would argue are more important

(I think knowing whether people still value me spending time writing vs where I should spend the time elsewhere is more useful than a wider audience having access to my EA-themed baby outfit roundup)

You know how many people value your paywalled content at your subscription price, but you don't know how many people would value it at a lower amount.

Thus, the knowledge you get from subscription data almost certainly understates the total net value that would be unlocked for a fixed amount of work if your writing were free. (That's not to suggest that you should feel any obligation to share your baby outfit insights with the world for free.)

That's much less of a drawback where each marginal unit of production incurs a substantial resource cost.

Completely agree with this and have been thinking about mechanisms for doing this lately.

I think there are high context and low context services. A low context service doesn't require knowledge of EA to perform, high context does.

Services that are low-context (i.e. doesn't require knowledge of EA to perform) should be selected based on quality & price. There's no reason to prefer that a member of EA is making money as a web developer (as opposed to any other provider) unless you believe you're getting more than what you pay for based on their membership to the community.

Services that are high-context there's amore nuanced case: there might be a lot of benefits to having more people trained up in that domain specific skill. Essentially, there are positive externalities to having this funded. But, I still believe the signal of how much orgs are willing to pay is very important.

An idea about cost effectiveness evaluation for services: "Primary industry" organisations can approximately model cost effectiveness in terms of their main output: $ per life saved, $ per career change etc. When somebody charges for their services, it requires the primary orgs to evaluate the cost-effectiveness of that service. How many more lives saved if this operational system cost less to run? If we had a better website?

I imagine there could be a way of capturing that trade-off and having service providers estimate their impact through the prices organisations are willing to pay.

Services that are low-context (i.e. doesn't require knowledge of EA to perform) should be selected based on quality & price. There's no reason to prefer that a member of EA is making money as a web developer (as opposed to any other provider) unless you believe you're getting more than what you pay for based on their membership to the community.

All else equal, I think transferring money to EAs rather than non-EAs will transfer some money to effective charities. I don't think this is a large effect, but e.g. it could be 10% of my fee if they're a GWWC pledger, so perhaps I should discount prices offered by pledgers by around 10% before comparing them to other prices.

(In fact, I probably value moving resources to EA-motivated people more than just what they subsequently donate, but the more indirect the model of impact, the less I think it's worth analysing in detail.)

This is an interesting read, thanks so much for putting it together. I can't really talk about things on a larger scope e.g. companies and official consultancy offers, but I suppose on the more micro-level of cakes at conferences or research on a forum...

I think you set out some of the conflicts in a way that helped me structure my thoughts a lot more clearly. I offered an organisation some support in putting together a meta-analysis on this forum in a field I specialise in - this was out of a general sense of altruism and support of the over-arching cause, as well as soothing my conscience a little for otherwise providing critique. At the same time, I did wonder if doing so 'de-valued' the offer i.e. suggested my skills were worth what I was offering (nothing), I had an abundance of time to be doing this (i.e. I'm not in demand), and as an EA-outsider using a pseudonym, I don't come pre-vetted [While my brother is 'known' in EA, I wouldn't feel comfortable using my searchable legal name on a public forum due to my profession putting me at high risk of stalking, and (no doubt) a combination of my age/gender/physicality etc meaning I've had a lot of unwelcome attention online].

I think all of this skews naturally to EA projects selecting in-group (trusted) members for specific tasks, or having to pay for commercial services, which will likely come with a higher price due to any overhead of running a company (unless offering an EA-discount, making it more likely to be EA-aligned and receive promotion through EA - it's perhaps a quid-pro-quo). E.g. I have no idea about accountancy and if I was hiring an accountant, I wouldn't be able to look at their work and know if they were good or not - I'd likely rely on reputation, recommendations, the appearance they are 'legitimate' etc, and if they fit my budget. It makes sense to me for EA projects, this would favour EAs and EA-orgs.

Perhaps there is an issue with a more 'grassroots' concept in EA in the broader sphere of cost-effectiveness where some people likely struggle to properly evaluate their worth, don't charge for their time, and therefore won't get the 'signals' to build on their contributions to realise they should. It's quite a leap from offering ad-hoc services to running a company. On the other hand, with receiving grants/funding or reputation being the markers of 'valuable' work in the community, this is open to corruption. Let's say my family is wealthy, so my Dad is happy to give me some money to start an organisation, there's little/no personal risk if it fails, and I don't need to work for six months so I can focus entirely on getting it off the ground. I could set it up as something official and make it look like a 'proper' company, get a flashy website, devote all of my time to developing content and applying for grants, offer discounts to solicit clients and positive endorsements, advertise, network in EA, etc. Even if my work was terrible, if it is niche enough, it's likely I would seem extremely legitimate within the community and I could enter a positive-feedback loop of increasing revenue, over-confidence and over-recommendation. It seems that there is risk of over- and under-funding effort along the same lines.

This is perhaps a silly question, but do people ever use a sort of Patreon model or have links to individual cash apps like PayPal etc.? [FWIW my offer wasn't taken up so I'm relatively unbiased here] To be basic, I suppose how people might endorse/reward creators of content on social media platforms (a bit like Asimov in your example having a 'Give whatever you like' tin vs pricing the cakes). I've not seen this and it might be against the rules (sorry) or be a gauche idea, but this forum feels relatively unique in the amount of work people put into posts and other community-facing offers from regular contributors which provide a lot of value. It might allow for people to make better cost evaluations of their time/effort and perhaps encourage people who are doing good work to continue and recognise they should be charging for their services, which would allow the organic development to perhaps something official. This might also provide a slight counter-lever to the power dynamics and provide scope for reduced insularity in EA, as mentioned in your post, in encouraging smaller service-providers to not work for free, and allowing some questioning of bigger organisations utilising 'free' services if there was an option to fairly compensate people for their work (even if this was voluntary).

The Patreon idea is interesting! One potential drawback is that it could skew participants' incentives (and feeback mechamisms) a bit in the direction of whatever made the class of people who disproportionately participated in Patreon happy. 

For example, I believe that median salaries for people working in AI safety are considerably higher than those working in global health & wellbeing. If everyone spent (say) 0.03% of their salaries on Patreon subsidies, and people were more inclined to use Patreon to send money to semi-volunteers in their own area of focus, using Patreon as a feedback mechanism would likely skew toward AI safety merely because AI safety folks tend to have more money to give.

More from Lizka
Curated and popular this week
Relevant opportunities