How To Do What You Want: Akrasia and Self-Binding

Monday, November 15, 2010

By dreeves

UPDATE: A revised version of this article now appears at blog.beeminder.com/akrasia.

Preface: I love dog food, metaphorically speaking, of course. This article is about forcing oneself to do what one wants to do and I, among other things, want a new Messy Matters article to go out no less often than the first of every month. You might point out that I must not want monthly articles as badly as I think I do, given that I didn’t in fact manage to publish this on November 1st. [1] But you would be wrong, as this article will argue. It’s true there’s a version of me that apparently didn’t want it bad enough (the me that, you know, failed to do it) but the version of me writing this now, thinking ahead to December 1st, is fully convinced that it’s absolutely worth meeting that (self-imposed) deadline. Yet the historical evidence is clear: when the time comes, there’s a pretty good chance I won’t actually do it.

So here’s where the dog food comes in: If you, dear reader, come to Messy Matters on or after December 1st and this is still the top story, I’ll send you $1000. I’m perfectly serious. [2] If that sounds nutty, read on.

UPDATE (2011 January): I managed to hold on to my money, but I put up another $5000 — payable to David Yang and David Reiley who bought my contract — on, among other things, continuing to spend at least two hours a week on Messy Matters till May: beeminder.com/d/mm.

UPDATE (2011 June): I held on to the $5000 as well. Phew! But I clearly need to do it again. If you’re intrigued by high-stakes commitment contracts like this, check out this post on the Beeminder blog.

Akrasia: Not doing what you genuinely want to do.

The problem that many of us have in failing to follow through on our intentions is more than just a difficulty in predicting our future desires. It’s not like “Gee, I thought I wanted to get in shape but it turned out there was always something really good on TV!” No, even in hindsight, you regret not doing what you said you wanted to do. It’s not even that you’re merely conflicted about what you want. The trade-off you made—more TV watched, still not in shape—was patently ridiculous. You somehow don’t do what you genuinely want to do. Philosophers back to Plato and Aristotle have a fancy term for this paradoxical failure of the will: akrasia. It encompasses procrastination, lack of self-control, lack of follow-through, and any kind of addictive behavior. [3]

The idea of people doing something other than what they want is a little strange, especially to economists, who often presume that whatever you did was by definition what you wanted to do. [4] But that elegant little theory fails to account for a stubborn fact of human nature: what we want depends on when we’re doing the wanting. This is called time inconsistency and is illustrated nicely in a study on grocery-buying habits: When buying groceries online for delivery tomorrow people buy a lot more ice cream and a lot fewer vegetables than when they’re ordering for delivery next week. In other words, our preferences are inconsistent — in fact, logically contradictory — over time. Your ability to weigh the costs and benefits (yumminess, healthiness) is severely compromised when some of those costs or benefits are immediate, which is why the ice cream vs vegetables decision plays out so differently when you decide it from a distance.

But what really clinches the case that people sometimes act against their own true interests is our use of self-binding, also known as commitment devices.

Commitment device: A way to change one’s own incentives to make an otherwise empty promise credible.

The term commitment device is from game theory and applies to strategic situations. It refers to a way of changing one’s own incentives to make an otherwise empty threat or promise credible. This can be quite rational. A classic example is from Thomas Schelling, who pioneered this aspect of game theory. You’ve been kidnapped and you’d like to promise your kidnapper that if they let you go you won’t rat them out to the police. The promise is useless because the kidnapper knows you’ll have no incentive to keep it once you’re free (and so they won’t set you free). But if you can change your future incentives (implicate yourself in the crime, perhaps?) then suddenly your promise can carry weight. [5] Limiting your future options by voluntarily imposing consequences on your future self can be quite valuable in a strategic negotiation or conflict.

But it’s hard to characterize as rational the use of self-binding with no one but oneself… until you appreciate that there’s in fact more than just one self. [6]

Sapid Sweets and Seductive Sirens

Here’s a ubiquitous example. Why not save money by buying bulk-size candy? It’s technically entirely irrational not to. You should buy the bulk size, save the money, and then ration it to match the pace you’d eat it when buying the individual sizes. Same consumption rate, less cost. The problem is your future self won’t be rational with a pile of candy in the house. Sugar is a bit addictive, after all. So it’s worth spending some extra money to constrain the choices of your gluttonous future self. Similarly, some smokers buy cigarettes by the pack instead of the carton, paying a hefty premium to throttle their future consumption.

The classic, prototypal example of self-binding is Odysseus tying himself to the mast of his ship so he could listen to the Sirens without getting lured onto the rocks. Calling Homer’s Sirens addictive is an understatement since they kind of literally control your mind. But there are a lot of things in modern society (hello, Internet) that are at least a little addictive and getting more so. [7] In “The Acceleration of Addictiveness” Paul Graham argues that this trend is fundamental to the growth of technology itself.

The recalcitrant future-you that thwarts your current intentions is every bit as smart as well-meaning current-you.

So what’s the antidote? There’s an endless stream of advice about how to beat akrasia but I believe that most of it misses a very fundamental point that renders it nearly useless: The myopic future self that thwarts your intentions is every bit as smart as your current, forward-looking self. You can make lists and set rewards and break tasks into small chunks, or plan diets and buy treadmills and establish routines, but mostly your future self will see right through all the tricks and just won’t give a damn. That version of yourself just wants to surf the web and/or eat pie. So that’s what you do, when the chips are down.

The only way to be immune to future-you thumbing his or her nose at your current intentions is self-binding. Which is to say that successful anti-akrasia tricks will involve commitment devices. That’s because, by definition, a commitment device meaningfully constrains your recalcitrant future self’s actions. Here are a slew of commonly used commitment devices (and at the end of this article are additional fascinating ones) in the form of a straw poll so we can see which are most popular among Messy Matters readers:

click through to the post on Messy Matters

The commitment devices people actually use are mostly — with notable exceptions — band-aid solutions to akrasia.

What I find fascinating about that diverse list of commitment devices is the minuscule dent they put in the overall problem of akrasia. The tricks people actually use are mostly band-aid solutions (see the list at the end of this article for some interesting exceptions). The problem is not with the concept of precommitment. (Remember that akrasia is fundamentally a problem of your future self failing to follow through on your current intentions.) The problem is failing to commit hard enough. Take a goal like weighing twenty pounds less in a year. Band-aid commitment devices are things like keeping junk food out of the house and joining a gym. Why not get straight to the heart of the problem and make a contract with a friend that commits you to forfeiting a very painful sum of money if you don’t follow through on your weight-loss plan?

Beeminder

Well, here’s why not: You’re a procrastinator. You might spend the first ten months with your head in the sand and the last two months literally starving yourself, scrambling to meet the deadline. With the goal a year in the future you have a new, recursive commitment problem: getting yourself to get started on the year-long commitment before the last minute. So better than committing to the long-term goal directly is to commit to daily, inexorable progress towards it until you actually reach it. I happen to be working on a tool to facilitate exactly that. Continuing with weight loss as an example, here’s how it works:

Pick your goal weight and goal date.
Let Beeminder create a “Yellow Brick Road” for you to follow:

beeminder example graph

Place a large bet with a friend that you’ll stay on that Yellow Brick Road. If you go off (above) the road even for a single day you lose. [8]
Procrastinate like hell until you’re about to lose the bet.

The change in focus from “weigh 20 pounds less next year” to “be on the yellow brick road tomorrow morning” makes all the difference. If you’re in the wrong lane of your road today then it’s crunch time. You have to be on your road tomorrow morning. Pull an all-nighter on the treadmill if that’s what it takes.

Beeminder’s secret is to break down your long-term goal into day-to-day guidance — the Yellow Brick Road — and to force you to follow it.

In one sense that mentality’s crazy. Whatever you do in any single 24 hour period makes essentially no difference to your weight next year. But that’s the kind of thinking that let you drift away from your ideal weight in the first place. Beeminder’s tagline is “Bring Long-Term Consequences Near!” [9] Akrasia is the failure to adequately consider long-term consequences (like the health problems with remaining overweight) when there are immediate consequences pointing in the opposite direction (this pie is mind-bogglingly delicious!). Beeminder’s secret is to use the Yellow Brick Road to automatically break down your long-term goal into day-to-day guidance. And then, critically, add a commitment device to force yourself to stick to it, to force yourself to actually make that gradual but inexorable progress. In other words, you transform the long-term consequences into near-term consequences (“I’ll forfeit this money if I’m not on the road in the morning!”).

Bright Lines

Many people who agree with the above in principle react viscerally to the idea of involving money. I’m partial to money myself but I don’t claim it’s the only way. Still, many obvious alternatives — like harnessing social disapproval — can fail if implemented naively. To make the long-term consequences of failing at your goal immediate, you need a bright and painful line. Using shame as a motivator with a public Yellow Brick Road doesn’t quite cut it because there’s minimal shame in having one datapoint slightly off your road — you’re still totally going to catch back up, right? As you then gradually drift away from your road the shame grows only gradually as well. You’re back in “one more day won’t matter” akrasia-addled fantasy land. I speak from experience.

To make the long-term consequences of failing at your goal immediate, you need a bright and painful line.

Speaking of bright lines, the next most sticky sticking point with commitment contracts is probably a predilection for leniency and clemency. Or you could call it weaseling when it comes from the self-binders themselves. “Shouldn’t I still win if I get back on my Yellow Brick Road and end up reaching the final goal?” they (preemptively) whine. Absolutely not. Recall Beeminder’s goal: bringing long-term consequences near. In other words, the fact that you lose the game if you’re off the road today is by design. To reprise the core akrasia problem, it’s very hard to, for example, forgo that piece of pie merely because it will make it harder to weigh 20 pounds less 10 months from now. Please! One piece of pie won’t make the difference and there’s plenty of time to catch up! Each individual piece of pie is totally worth it. Same with each workout you really don’t feel like doing right at this moment. Which of course is how you ended up 20 pounds overweight in the first place.

With a Beeminder contract that whole dynamic changes: when you’re in the wrong lane of your road that one piece of pie could very well make the difference tomorrow morning and you’re acutely aware of it. The consequences are immediate. And of course even better is the flip side of that coin: if you are well into the right lane of your road (or below it) then it’s very nice to be able to enjoy your hard-earned safety buffer and eat that piece of pie guilt-free.

Of course the same principles apply for a goal like learning a new technology or exercising or whittling down your inbox. If you’re frustrated with yourself for failing to carve out the time for something that’s important but never time-critical [10], commit to a Yellow Brick Road that keeps you averaging, say, an hour every other day. The width of the road gives you some flexibility but soon enough you’ll be in danger of driving off. If you’ve committed to staying on that road then you’ll be forced to put in the time.

One More Dose of Dog Food — $2k at Risk

You may have noticed that the very problem I’ve decried with long-term goals applies to my dog food contract in the preface. Well, Beeminder to the rescue! I am hereby augmenting that contract with another one. Not only will I cough up $1000 if this is still the top story on Messy Matters on December 1st, I will also cough up $1000 if I don’t stay on the following Yellow Brick Road every single day from today (November 15) through the end of 2010:

If you spot a datapoint below that road then a thousand dollars is yours to claim. Just consult the fine print [11] first. If this works well, I’ll commit to a new, open-ended Yellow Brick Road ensuring Messy Matters goodness for years to come. [See the update in the preface above.]

Image by Kelly Savage; graphs by Beeminder.

Thanks to Sharad Goel, Eva Revesz, Dave Pennock, Melanie Reeves, Dan Goldstein, Jill Renaud, and Bethany Soule for reading drafts of this and to many beta users of beeminder.com (originally known as Kibotzer, the kibitzing robot) for trying out some crazy commitment schemes and putting a total of nearly $20,000 at risk.

More Real-World Commitment Devices

Here are yet more examples of commitment devices in the wild. Or in the laboratory, but that people find real value in regardless.

Putting money in a Christmas Club account — a savings account that you cannot withdraw from until December. [12]
Taking a diet pill that causes diarrhea when you eat too much fat.
Putting your alarm clock across the room. (This is a cheap example since it involves an explicitly compromised future self, in this case a groggy one.)
Taking Antabuse to cause yourself to get sick if you consume alcohol.
Using the kind of piggy bank that you have to smash open to get the money out.
Undergoing bariatric surgery, which works by causing pain/nausea whenever you eat too much in a sitting. [13]
Banning oneself from casinos or online poker sites (they offer this functionality).
As a domestic violence victim, preferring no-drop prosecution. [14]
In an experiment by Dan Ariely on his students, “irrationally” choosing to be bound by spaced-out deadlines for course projects. [15]
Getting married as a way to make it hard to break up? (This may be more about proving your commitment than about imposing a commitment device on yourself.)
Using a device to ration out pain medication. This is a real thing.
Preferring coupons or gift certificates with expiration dates to ensure you’ll eventually get around to using them. (In a study that I cannot for the life of me find a reference for, students were given coupons — some of which had expiration dates — for free ice cream. The students expressed greater value for non-expiring coupons but got more value from the coupons with expiration dates. Without a deadline they would often put it off until they lost the coupon.)
Preferring gift certificates at all instead of cash.
Using a layaway program to force yourself to save up for a cherished item.
Using a so-called kosher phone, actually available in Israel, that stops working on the Sabbath.
Traveling somewhere to do something that you could but won’t do at home, like to escape distractions that you don’t have the willpower to simply ignore.

The ones in bold are especially interesting because they involve voluntarily submitting to very painful consequences if you do the thing you don’t want to do. They seem psychologically very similar to setting up a commitment contract with a large cash penalty.

If you know of any commitment devices I’ve missed (or other anti-akrasia techniques), add them in the comments!

UPDATE: For new ones I’ve learned about since this went to press see the updated version of this article on the Beeminder Blog.

Footnotes

[1] You’ll notice from the Messy Matters archives that we’ve never had a month go by with nothing published here. At some point I got it in my head that such a gap in the archives would be embarrassing — it would look like Messy Matters posts were petering out or something. Or maybe it was just the psychology of Jerry Seinfeld’s “don’t break the chain” productivity trick. Regardless, midnight on the last day of the month became the drop-dead deadline for the next post, and that actually motivated us. Again and again we would try to publish earlier but the end of the month deadline was the only thing I (not speaking for Sharad) could ever stick to. Now we’ve decided that if we’re committed to a once-a-month minimum then the deadline should really be the first of the month instead of the last. But now the deadline isn’t hard anymore: we published on October 1st so as long as we publish sometime in November we haven’t broken the chain. In other words, there’s nothing critical about the first of the month anymore.

Dave Pennock offers an insigtful comment on the nature of self-imposed or otherwise artificial deadlines: “If you miss a deadline then it’s a slippery slope. ‘I already missed it by a day: two days is no worse than one. Four days is no worse than two. I’ve already missed it by four, why not eight?…'” No single day matters yet they add up to unacceptably late. It’s a pernicious form of irrationality, but read on for the solution!

[2] Fine print: You can claim the thousand dollars starting at exactly 1am New York time the night of November 30 (the first hour of December is breathing room in case of technical glitches). Also, to claim it you have to leave a comment here starting with “CLAIMED” in all caps. I won’t pay more than $1000 total so if more than one person claims the money then we’ll either divide it or give it to the first person to claim it. I designate Sharad as arbitrator to resolve that and any other ambiguities. In any case, the spirit of this should be clear and I promise not to do any weaseling. Speaking of the spirit of this contract, if you know me in real life and couldn’t bring yourself to actually take my money because you like me too much or something (in which case you really don’t know me too well) then please don’t claim the money!

Finally, I’ve put another $1000 at stake, requiring me to work on my next Messy Matters post every day (technically: to never let my average time spent per day fall below my target; and in fact any work on Messy Matters counts, not just the writing of the next post) through the end of 2010, but you’ll need to read the rest of this article for the details and your chance to claim it…

[3] Another way to define akrasia is by generalizing from procrastination to include preproperation as well. Procrastination is the irrational delay of tasks with immediate cost and delayed benefit. Preproperation is the irrational not delaying of (overindulgence in) activities with immediate benefit and delayed cost.

[4] Of course if you’re, say, being kept down by The Man or even just, say, flossing then you’re doing something other than what you want. But what we mean is doing something other than what you want given the constraints of the environment. In other words, “what you want” doesn’t mean what you’d hypothetically want in some other set of circumstances. In this sense, even if I literally tie you to a chair and put on a video of “Lawnmower Man 2: Beyond Cyberspace”, you’ll still be doing what you want (squeezing your eyes shut and moaning piteously). But then how exactly can anyone ever do anything other than what they want, in this sense? Read on!

[5] Or take the example of Cortez burning/sinking his ships to remove the possibility of retreat and thus increasing his chances of defeating the Mayans. More abstractly, in a game of Chicken against a rational opponent, if you can manage to remove your own option to yield (like ripping the steering wheel out of your car so you can’t swerve) then you win.

[6] Although I’m using the multiple selves terminology, I view it as no more than a rhetorical device. There are not multiple selves in any real sense. Rather, as I will argue, there’s in fact just one true you whose decision-making is sometimes distorted in the presence of immediate consequences, which act like a drug. Which is to say (as an empirical fact) that when a decision involves some consequences that are immediate and some that are distant, humans irrationally (no amount of future discounting can account for it) over-weight the immediate consequences. So I refer to multiple selves but I believe there is one true self: the one not under the influence of immediate consequences.

[7] James Surowiecki, in a recent New Yorker book review, “Later: What does procrastination tell us about ourselves?”, cites a survey in which the percentage of people who admit to difficulties with procrastination quadrupled between 1978 and 2002.

[8] You might object that this is far too harsh since daily weight fluctuates randomly and with menstrual cycles and whatnot. Quite true, but Beeminder tries to get the best of both worlds. On the one hand, we want unbending rules about what it means to stay on track toward your goal, because that’s what’s motivating (no “I’ll catch up later” where you dig yourself in a hole). On the other hand, you should never lose on a technicality. Beeminder achieves this by auto-adjusting the width of the Yellow Brick Road to account for daily weight fluctuations.

[9] This differs from stickK.com which adds consequences but can’t bring them quite so near. And I should add that “bring long-term consequences near” is one of many possible taglines for Beeminder which is only just now transitioning from a pet side project to help friends and family into a real startup.

[10] I like the analogy of rocks and pebbles in a jar. The jar is your life and the rocks and pebbles are things that fill your life, things you can spend time on. The pebbles are the little things like email and errands, often “urgent” (time-sensitive) but not that important. The rocks are the opposite: things like spending time with family, playing music, learning new skills — whatever matters to your quality of life. What makes the analogy memorable is to watch what happens when you let the pebbles vs the rocks take precedence. If the pebbles go in first then the whole jar ends up full of pebbles, with no room for any rocks. But if you put the rocks in first, plenty of pebbles (the truly urgent ones) will still fit in around them. The lesson: Stop procrastinating and get those rocks in that jar!

[11] So far in 2010 I’ve averaged 3 hours per week working on Messy Matters. This Yellow Brick Road is 2 hours per week, so it feels safe enough. As with the first $1000 I’m putting at risk, leave a comment here starting with “OFF THE ROAD” in all caps to claim my forfeiture if you spot a datapoint below this Yellow Brick Road. Just note that I have until midnight eastern time each day to reach the bottom edge. It’s at my own discretion what to count as legitimate Messy Matters time, but I promise not to abuse that.

On a highly technical note, I track my time stochastically using a somewhat elaborate system Bethany Soule and I concocted, so 2 hours per week actually only means 2 hours per week in expectation. But the system is quite ungameable and this in fact only makes the goal harder. For example, if I’m a half hour below my Yellow Brick Road for the day and wait till 11:30pm to get started, the stochastic self-sampler could fail to sample me during that half hour and I’d be out of luck. I use this system, by the way, because I’m too scatterbrained to faithfully start and stop a timer. Some types of interruptions make that hard, but also you kind of can’t, by definition, stop your timer because you stopped working to space out (if you had the presence of mind to stop the timer then it wasn’t really spacing out).

[12] An excerpt from an article in Forbes by James Surowiecki:

Christmas Club accounts [were] a staple of American banking in the 1970s. The Christmas Club account was, on the surface, a bizarre idea: You gave the bank money every month. The bank paid you no interest, and it would not let you take the money out until Dec 1. This was not much of a bargain. And yet Americans happily put their money into Christmas Club accounts, even though that same money in a savings account would have earned them interest.

(Thanks to Steven Reeves for pointing out that these are now sometimes called Santa Saver accounts.)

[13] Before weight-loss surgery existed people would sometimes have their jaws wired shut to lose weight. Either way, the mechanism is no more or less than making it physically difficult or physically painful to consume too many calories in a sitting.

I have a crazy theory that the same effect could be achieved with a purely monetary commitment device. Imagine putting $20,000 (my guess as to the cost of bariatric surgery) in escrow and having a Yellow Brick Road to follow such that if you ever went off the road, even for just a day, the $20k would be gone. If you ever got close to the edge of the road then you’d really be nervous. That piece of pie would not look so appetizing. It could possibly be enough to put you over the edge, i.e., it might be a $20k piece of pie. You — even that lazy, rapacious version of yourself — would probably genuinely not want to eat it, given the immediacy and severity of the consequences. (I would love to test this theory if I could find a willing guinea pig!)

[14] No-drop prosecution is an option to press charges against an abusive spouse and stipulate that you don’t want to be allowed to drop them (the charges). Actually I believe you don’t literally have the choice to stipulate that — you’re either in a no-drop jurisdiction or you’re not — but some economists have inferred that many domestic violence victims would choose no-drop prosecution if they could. In the case of an abusive relationship I could imagine strategic reasons for that (so you can’t, and your spouse knows you can’t, cave to threats to get you to drop charges) but the claim is that due to the sad psychology of domestic violence victims, it’s really about forcing your future self to follow through with what you know (when you sit back and reflect) is the right thing to do.

[15] The setup is that he let his students pick their project deadlines for the whole semester. Once they’re picked they fail the course (they paid thousands for this course, we presume) if they don’t meet them. So rationally they should choose all the deadlines to be on the last day, to maximize flexibility. But it was clear to students that that was a terrible idea. And sure enough, the ones who precommitted to spaced-out deadlines did better.

Tags: beeminder, behavioral economics, breakdown of will, commitment devices, dynamic inconsistency, economics, ego depletion, egonomics, fallacies, hyperbolic discounting, multiple selves, picoeconomics, precommitment, preference reversals, productivity porn, psychology, rationality, self-binding, time inconsistency, wagers, willpower

Posted 2010 Nov 15. RSS feed for comments on this post. Please leave a response, or trackback from your own site.