The Case for Impact Purchase | Part 1

In this post I will argue for impact purchase as a complement to (not a replacement for) other funding mechanisms, such as grants.


What is impact purchase?

Impact purchase is an alternative funding mechanism for altruistic work. I think the easiest way to explain what impact purchase is, is to compare it with grants.

For a more thorough explanation, and an actual example, read this excellent blogpost.


About this blogpost and the next one

This is part one of what I intend to be a two part blog post series. The second part will be significantly more work, so I’ll do that if this one gets enough interest.

Part 1 (this blogpost) will focus on the timing aspect of impact purchase, i.e. why it is sometimes better to evaluate and fund a project after it is finished.

Part 2 (next blogpost) will focus on the whys and hows of paying for value rather than cost. Most of my suggestions regarding implementation will be in this post.

Disclaimer
So far, my only investigation into this previous attempt to implement impact purchase, was to look around at their now inactive website. I don’t even know why the project ended. If someone could supply more information about this in the comments, that would be great. Otherwise, I will try to find out more for part 2.

Comment request
I am arguing a lot from my own experience and perspective. I hope that other people who want impact purchase to become a thing, will add their stories in the comments, or even just indicate that this is something you would apply for. I think this is relevant both to see how widespread the interest for this is, and what sort of projects it would support.


Reasons to evaluate a project after it is completed

I think the best part of impact purchase is that projects are evaluated after they are done.

You might think that this would be great for the funder (because it is much easier for them to know what they are paying for) but less great for potential recipients (because they now get to take all the risk). If that is how you think, then this should surprise you: I’m a potential recipient and I love the idea of having the funding application happen after the project. I don’t speak for everyone (and this is why we should still have grants) but I do know several other potential recipients who share my preferences.

More flexibility
For me, the time period between [I have a plan], and [I start work on this project], is typically somewhere in the range negative one month to a few weeks. By negative time, I mean that I start working before having a proper plan.

Having a plan is, in my experience, absolutely necessary to get a grant, but getting a grant takes months. There is typically a few months waiting time until the next suitable grant application opens. But even if I’m lucky with that, I’ll have to wait about two more months for grant evaluation. I’m not waiting around for that, and I would be very surprised if this problem was unique to me.

With impact purchase I can just follow the schedule that works best for me and my projects, and have it evaluated and funded some time in the future.

It is easier to evaluate a project after it happened
I have not done grant evaluation, but my impression is that it is super hard. Evaluating the impact of a project after it has happened is probably not easy either, but it has to be easier. This gain can be used to save time, do better evaluations, give more feedback to applicants, or some combination thereof.

If EA is vetting constrained, the gains from evaluating some projects after rather than before they happen could be hugely beneficial.


Why would you (or any EA funder) pay for something that has already happened?

However, you might think that paying after the project is over, is actually a bad deal for the funder, because you want counterfactual impact. You want your donation to create value that otherwise would not have happened. So why would you agree to pay for something that has already happened? Here are two ways of looking at it

Empower people based on their track record
The impact seller is someone who recently did something valuable with their time (because otherwise they would not have impact to sell), which is strong evidence that they are the kind of person who would do valuable things in the future. So let’s give them some money, no strings attached, and let them do whatever they want. Based on past evidence it will probably be good.

Under this motivation, the agreement is still that you pay people based on their past work (if you do something else it is not an impact purchase). But as a donor you are doing this with the expectation of more good work from them in the future.

Incentivise future altruistic work
The prospect of getting paid may motivate people to do good work. By paying some people for their past work, you build expectations and trust in the system, which will motivate more good work.

The difference in motivation matters for implementation
I think any donor needs to be clear about which of these reasons (or other) is motivating you, because your implementation would probably be somewhat different. If you want to empower great people, you may want to look more at who they are and what you expect them to do next. If you want to incentivise future work, it is important that your payouts are predictable.


My story (skip if you want)

Read this as a case story. Impact purchase should obviously not be implemented just for me. I would not write this post if I did not know others who would benefit too. But maybe using myself as an example will clarify some things? I don’t know.

I’m an AI Safety organiser. Last year I organised the Technical AI Safety Unconference (TAISU) and the Learning-by-doing AI Safety Research workshop (LAISR). Both of them were successful and I’d like to run more of these events.

However, this year I’m only planning to run TAISU and not LAISR and it has to do with money. I think that LAISR is potentially very valuable and I would love to keep developing this concept, but the target participants are mostly students, who are not able to pay much for the event. The only way I can get paid for LAISR is with external funding.

Last year I lived at EA hotel, and also used this place as an event venue, so money was not a problem. But I moved out now so I need to start caring about actually getting paid. I still have a lot of free time, so if impact purchase were a thing, I could experiment around with various projects, get paid for the ones that succeed, and learn from that.

I could try applying for a grant, but I’m not going to do that. I have had enough of getting turned down for grants. Maybe you think I should take the fact that I can’t get grants as a signal of something. But given that I mostly succeed, when I get to work on projects on my own terms, I take it as a signal that the grant application process is a bad match for my working style.

I imagine some reader of this post wanting to tell me that I too can learn the skill of applying for grants, and writing up a proper project plan in advance would be a valuable thing to do anyway. If that is you, please trust me when I say that this would not work for me. See the bit about flexibility earlier in this post. It would hurt both my wellbeing and productivity to try to conform to a grant format.

I need to know that a project will happen in order to have the motivation to engage in planning. One of the first things I do when organising an event is to write up the event invitation. I think about what information needs to be in the announcement (dates, venue, cost, etc) and work from there. Anything that does not need to go in the announcement, I’ll figure out later. This process works for me, because it works with the structure of my motivation, not against it. But starting with a grant application triggers no motivation for me.

The one thing I do like about grants is the invitation to ask for money. I have considered trying to get some sort of peer funding (Patreon or similar), but find it incredibly uncomfortable to ask for money unprompted (though I rather do that then grants at this point). I’m hoping that if impact purchase becomes a thing there will be some formal application system.


Some concerns and my responses

Grants provide good feedback, and help people not waste their time on bad projects.
I’ve heard this type of argument from many people when discussing grants. There seems to be this idea that applying for grants is a great source of feedback.

In my personal experience, the information value from applying for grants has been close to zero. Some reasons I personally have been turned down for grants:

  • I did not have a budget. Because of this my project did not get evaluated. (I could not provide a budget since I did not at that time know what my living costs would be, and therefore could not say what salary I needed.)

  • I applied for too little money. The lower bound was 10k and I only needed 5k. Because of this my project did not get evaluated.

  • My project was related to EA Hotel, and no grant making body ever makes up their mind about EA Hotel. Because of this my project did not get evaluated.

I’m not saying none of these problems were my fault. I am not saying I did not learn anything (I did learn how not to write grant applications). Also, I would not be surprised if other people have had very different experiences. I’m only saying that realistically speaking, in the real world, grants are not a reliable way to get project feedback. But even if it were, this would still not be a very strong argument, because it is possible to just ask for feedback on a project idea.

However, I admit that there is something very special and real about feedback that comes in the form “someone is willing to pay for this”. I want this type of feedback! This is actually one of the reasons I want impact purchase. But I’m fine with trying something out first and having it evaluated after the fact, because that would be less of a cost for me than conforming to whatever is needed to get a grant. This is not just speculation, to this point I have infinitely more success in completing projects than navigating grant applications, and I’d love to have some of my past projects evaluated from a funding perspective.

What about reputation risks and other unilateralist curse type of stuff?
Another thing I have come across is the idea that we need to make people apply for grants from central organisations, in order to police what project people are working on.

Given that people in the EA movment are already doing lots of stuff without central oversight, I’m not convinced that this is something we need to worry about. Or maybe the current situation is already a ticking bomb, and introducing impact purchase will make it worse? I don’t think so, but it is a valid concern.

Either way, I do think it is a good idea to avoid buying impacts from projects which have had a large downside risk. I think the evaluation for deciding to buy impact or not, should be at least partly based on: “Do we approve of this work?”, which include things like downside risk and negative externalities.


When not to go for impact purchase

These are the reasons I’m suggesting impact purchase as a complement, not as a replacement for other funding mechanisms. Removing grants would be bad.

When the money is needed up front
Not everyone can afford to run their project without payment in advance.

When the applicant can’t afford not getting paid
Even if a person has the cash at hand to do their project, it might still be a very bad idea for them to spend that money without any guarantee of reimbursement. Grants are safer and some people will need that safety.


Some initial thoughts about implementation

I will discuss implementation more in Part 2, when I have had more time to sort out my own thoughts regarding paying for value. However here’s one thing I think I can say already:

Consistency is important
Impact purchase will have a much stronger effect on people’s behaviours if potential recipients can either predict or have the possibility to learn what outcomes will get funded.

Imagine someone did a project which you like and you would like, which leads you to give them some money. Will this encourage this person (and maybe others) to do more of the same type of projects? Well maybe. Is a past impact purchase strong evidence that the same project will get rewarded in the future? If “yes” your money will much more strongly encourage similar things in the future.

“If I keep this up, I’ll keep getting funded” creates a form of financial stability that can potentially be very powerful. Impact purchases will never be a perfectly reliable source of income, but what sort of employment is these days? The more uncertain the payments from impact purchases are, the more people living off this will be incentivised to put their rewards into personal savings accounts, rather than use the money for their next cool thing.