# Socially Efficient Commitment Devices

Monday, March 4, 2013
By dreeves

StickK popularized the idea of the anti-charity [1] as a commitment device. Another [Update: former] Beeminder competitor, Aherk, offers to publish embarrassing photos of you on Facebook to ensure you don’t fall prey to akrasia. Another clever idea — proposed by Jennifer Hamon on Akratics Anonymous — is to set up a script to automatically delete important files if you don’t keep making progress on a writing project.

I hate all those ideas.

#### “StickK’s anti-charities seem the most egregious, actively harming the world.”

Take my thoughts on this with a huge grain of salt, given my conflict of interest, but I really dislike commitment devices that destroy things — either information or other forms of value. StickK’s anti-charities [2] seem the most egregious, actively harming the world. I’m certainly motivated to not allow the world to become a worse place, so it’s not that it would be ineffective as a commitment contract. Just that I’m also motivated to prevent things that don’t make the world worse in any way, like paying money to a third party (who’s not evil).

At the other extreme, I’m skeptical of commitment contracts where the beneficiary is a (good) charity because what kind of jerk is motivated to not give to charity? If you set an amount that you really can’t afford then that can work. Unfortunately no one seems to ever have the guts to do that.

So, yes, this is all an elaborate rationalization for Beeminder getting all the money! I’m serious though, I think the economics / psychology / behavioral economics are sound. Commitment contracts should be socially efficient and deleting even files on your own computer isn’t!

In fact, if you set up a commitment device where you destroyed your computer if you didn’t stay on your yellow brick road then I actually would call that kind of immoral, even though it was your own computer. Because you could’ve instead given it away and not wasted anything. Burning money, in contrast, doesn’t waste anything (except some paper).

So I’m fine with the hypothetical money-shredding alarm clock, even though no one would ever actually use one of those. (There are some clever phone apps that make you jump through various hoops to make the alarm turn off that might be useful though.) And I guess, if I’m to be true to my principles, Aherk is fine too, from a social efficiency point of view, as long as your friends would get amusement out of your embarrassing photos commensurate with your own embarrassment.

The whole point of a commitment device is to set up an automatic potential penalty that punishes you in some way. Just make sure that the amount of harm caused to yourself is counterbalanced by a greater or equal amount of benefit to someone else, leaving the world no worse than you found it!

## Footnotes

[1] The idea originated, as far as we know, in 1984 in John Bear’s The Blackmail Diet where the author describes his own commitment contract to lose 70 pounds in a year or pay $5k to the American Nazi Party. Thanks to Daniel Akst’s We Have Met the Enemy where we learned that and many other fascinating things about the first world’s ironically biggest problem, self-control. [2] Bethany points out that this is a clever trick for getting more bang for your buck: it increases your utility for that amount of money if it is otherwise going to benefit something horrible such as the Puppy Kicking Society of North America. But at what cost?! Tags: , , , , , , , , • kshuzhongxin Either we assume rationality, or we assume the ability to surpass rationality as our own social planner and make social welfare our benchmark. First, if we assume rationality based on private gain, the negative effect of both commitment devices is exactly equal since a breach results in the same reduction of private utility (even at different amounts of expenditure). Second, taking a welfare economics point of view, mere transfers of wealth (Beeminder) are welfare-neutral, assuming zero transaction costs. It should be noted, however, that when we take social welfare as our benchmark (as opposed to private gain), we may no longer be able to say that anti-charities destroy welfare since the choice of particular anti-charities depends on the private (not society’s) welfare function (here again assuming a basic level of rationality, however). Assuming some probabilistic distribution of welfare effects of anti-charities, I still agree that Beeminder is likely to be the better option from a social welfare point of view. • http://beeminder.com Daniel Reeves What if I commit to pay someone to punch themself in the face? Could we agree that that’s destructive from the welfare economics point of view? As you correctly point out, the monetary transfer is a wash, so the net result is… one punch in the face. Still, you’re right that I was kind of conflating the private vs social welfare points of view. But maybe the face-punching example is an existence proof that anti-charities can be unambiguously bad. I guess your last sentencing is conceding that not only can they be, they probably are in real life, on average. In conclusion, Beeminder should get all the monies! We just proved it, with maths. • kshuzhongxin By the weak version of the Reeves Existence Theorem, there must always exist an anti-charity with a negative expected social welfare effect such that a welfare-neutral* Beeminder is the welfare-maximizing and socially efficient commitment device. By the strong version of the theorem, the expected marginal welfare effect of a payment to a population of anti-charities is less than zero, whereas the expected marginal welfare effect of a payment to Beeminder is greater than or equal to zero. Implications similar. For rigorousness consider thirdly, however, the possibility that anti-charities have a positive *expected* marginal effect on *social* welfare even if the expected marginal effect on *private* welfare is negative. In this case, Beeminder will cease to be the welfare-maximizing solution. This result can be obtained in theory since anti-charities are chosen with respect to private loss whereas the socially relevant criterion is social loss. And it will usually be the case that private loss > social loss for charities like political parties and special interest groups. It follows that the Beeminder solution to the Reeves Existence Theorem is socially efficient only for a subset of the entire conceivable set of anti-charities. * Assume all agents have the same marginal utility of wealth. • http://beeminder.com Daniel Reeves Ha! Agreed. But I bet that Actual Reality is firmly within that subset where Beeminder wins. Like I bet conservatives and liberals would agree that donating equally to both political parties is strictly worse than not donating anything at all. (Which was kind of your original point. we’re probably in violent agreement here. I do also appreciate the point that it’s not as simple as “anti-charities hurt the world! QED!”) • David Pay Sometimes versus Pay Never contracts When I first started with Stickk, I set up a weight loss contract with my nephews education fund as the recipient. (It’s worth noting, Stickk takes a pretty hefty slice as they pass the money along; something like 17%. So, if people commit larger amounts, as I was comfortable doing when the money would go to my nephews, Stickk or Beeminder should still be able to have a working business model) I expected to miss my goal some weeks, make it others, and I was OK with paying sometimes, but not too often, to my nephews.$300? Sure. \$1000 – hey, my brother has a good job!

Sorry Dr. Eeves (am I the only one who misread his name that way?), but I prefer to never, ever pay Beeminder anything. Sadly, hasn’t worked out. I think the money transfer to worthy recipient, with a slice for the processor, does a nice job of adding incentive where you are prepared to miss the goal sometimes. I also think that it is a powerful bit of marketing to say “We’ll take your money and send it to your nephews if you eat that pizza! [Very small print: after taking 17% for us to keep the lights on and buy Raman noodles.]”

• boardtc

I have let one of my beeminder goals skull out this year. Just this morning I was buying a bowl of fruit and found myself wondering why I was doing that when I was not on the yellow brick road!

• Eric Jain

Why not have people who don’t stick to their goals pay people who do? You could still take a cut :-)

• http://beeminder.com Daniel Reeves

Great question! Here’s our answer: http://blog.beeminder.com/gympact/

• http://eric.jain.name/ Eric Jain

Interesting! It’s nice to see that so many different approaches are been tried.

• Amanda

I wrote this as a recommendation on the suggestions board. So here are some of my thoughts there and a few extras to boot:

Pros of anti-charities:

(Super Incentivizing) as you say, this doesn’t incentivize failing in the way that giving to a ‘good’ charity would (I don’t want the warm fuzzies of giving to charity after I fail at a goal).

(Less Likely to Hoard Money from Beeminder) people are probably less likely to lower the amount going to beeminder if the alternative is giving it to a disliked charity and not a good charity that gives them warm fuzzies (assuming they think the good charity does more good than beeminder does)

Cons of anti-charities:

(Anticharities as Charities) one person’s anti-charity is another person’s charity, and if people select a charity they consider good, you’d expect more derailing, and probably less money to
beeminder in the long term as charities will undermine the disincentive
effect of the pledge.

(Less Money for Beeminder) People might pledge less to beeminder than they otherwise would have for the extra disincentive effect of the anti-charity.

(Moral Repugnance) The moral worry: you can expect to fail, and you’re making the world a morally worse place by failing. It’s one thing for you to take the hit, but another for other people to have to do so.

Responses to cons of anti-charities:

(Anticharities as Charities) you could reduce this effect by having some pledge or promise so that people don’t select something they actually consider a charity as their anti-charity.

(Less Money for Beeminder) assuming that people are willing to pay for the added disincentive of an anti-charity, you could mitigate this by having a minimum proportion of beeminder to anti-charity money in the pledge (e.g. beeminder:anti-charity cannot be less than 2:1).

(Moral Repugnance) This is a tough one, and shouldn’t be taken lightly. Here are a couple of possible ways of getting round it: (i) opt for more ‘benign’ anti-charities: e.g. those that couldn’t do too much harm with each additional dollar e.g. anti-charities that don’t have much room for more funding (one could surely work out the more benign anti-charities within a given ’cause’ and giving to a symbolically harmful cause might do a lot of the psychological work anyway), (ii) allow this as an option for goals that involve doing something morally beneficial, such that the expected moral cost of not completing the goal is greater than the expected moral cost of the dollars that go to the anti-charity. (iii) pay a ‘moral offsetting’ fee at the start or end of a pledge that involves an anti-charity component. This money goes to a really effective charity, such that the expected moral cost of an anti-charity pledge is always less than or equal to zero. But once people update on their money having gone to a really effective charity, each payment to the anti-charity is just treated as a loss.

I take the moral harm point seriously though, and it seems bad to choose anti-charities in cases where there will be no moral gains at all and the moral cost won’t be offset in some way.

• eo2

A trick that might work, for principled personalities in particular, would be to pledge to donate a very small amount, e.g. one cent. The actual impact to society is negligible[1] – no noteworthy harm is risked; but it still provides a deterrent effect: by threatening the principle of not supporting malicious causes. “Get off your butt, or you are literally an EVIL NAZI!”

This moral self-image type incentive would go nice with a public image threat. Cue the shaming: This is not attractive to have on your Facebook timeline: “I donated to a hate group just so I could eat some cake. I am a bad person.”

[1] In fact, it even might well be infinitesimally positive, as the anti-charity’s transaction costs (a small amount of time during bookkeeping) probably exceed the received cent. So you steal a bit of their time and piss them off. Yay for you…? ;)