Akrasia is the tendency to act against your own long-term interests, and is a problem doubtless only too familiar to us all. In his book "Breakdown of Will", psychologist George C Ainslie sets out a theory of how akrasia arises and why we do the things we do to fight it. His extraordinary proposal takes insights given us by economics into how conflict is resolved and extends them to conflicts of different agencies within a single person, an approach he terms "picoeconomics". The foundation is a curious discovery from experiments on animals and people: the phenomenon of hyperbolic discounting.

We all instinctively assign a lower weight to a reward further in the future than one close at hand; this is "discounting the future". We don't just account for a slightly lower probability of recieving a more distant award, we value it at inherently less for being further away. It's been an active debate on overcomingbias.com whether such discounting can be rational at all. However, even if we allow that discounting can be rational, the way that we and other animals do it has a structure which is inherently irrational: the weighting we give to a future event is, roughly, inversely proportional to how far away it is. This is hyperbolic discounting, and it is an empirically very well confirmed result.

I say "inherently irrational" because it is inconsistent over time: the relative cost of a day's wait is considered differently whether that day's wait is near or far. Looking at a day a month from now, I'd sooner feel awake and alive in the morning than stay up all night reading comments on lesswrong.com. But when that evening comes, it's likely my preferences will reverse; the distance to the morning will be relatively greater, and so my happiness then will be discounted more strongly compared to my present enjoyment, and another groggy morning will await me. To my horror, my future self has different interests to my present self, as surely as if I knew the day a murder pill would be forced upon me.

If I knew that a murder pill really would be forced upon me on a certain date, after which I would want nothing more than to kill as many people as possible as gruesomly as possible, I could not sit idly by waiting for that day to come; I would want to do something now to prevent future carnage, because it is not what the me of today desires. I might attempt to frame myself for a crime, hoping that in prison my ability to go on a killing spree would be contained. And this is exactly the behavour we see in people fighting akrasia: consider the alcoholic who moves to a town in which alcohol is not sold, anticipating a change in desires and deliberately constraining their own future self. Ainslie describes this as "a relationship of limited warfare among successive selves".

And it is this warfare which Ainslie analyses with the tools of behavioural economics. His analysis accounts for the importance of making resolutions in defeating akrasia, and the reasons why a resolution is easier to keep when it represents a "bright clear line" that we cannot fool ourselves into thinking we haven't crossed when we have. It also discusses the dangers of willpower, and the ways in which our intertemporal bargaining can leave us acting against both our short-term and our long-term interests.

I can't really do more than scratch the surface on how this analysis works in this short article; you can read more about the analysis and the book on Ainslie's website, picoeconomics.org. I have the impression that defeating akrasia is the number one priority for many lesswrong.com readers, and this work is the first I've read that really sets out a mechanism that underlies the strange battles that go on between our shorter and longer term interests.

New Comment
86 comments, sorted by Click to highlight new comments since:
Some comments are truncated due to high volume. (⌘F to expand all)Change truncation settings

This reminds me of a webcomic, where the author justifies his lack of self improvement, and his continual sucking at life:

"Pfft. I'll let Future Scott deal with it. That guy's a dick!"

http://kol.coldfront.net/comic/ (No perma-link; it's comic 192, if new one's been posted since I wrote this.)

When dealing with your future self there's an economic balancing act at play, because Future Self's values will inevitable shift. On the extreme side, if Omega had told Aurini'1989 that if he saves his $10 for ten years, it will grow to the point where he can buy every Ninja Turtle action figure out there, Aurini'1989 would have said, "Yes, but Aurini'1999 won't want Ninja Turtles anymore - however, he will likely value the memory of having played with Ninja Turtles." To hold the Future Self completely hostage to the desires of the present makes as little sense as holding the Present Self hostage to the desires of the future.

It breaks down to a tactical problem (which units do you build first in Civ 4?); I'm glad I spent money on that beer five years ago, because I still find value in the memory. What makes the problem difficult to solve is our fuzzy perceptions. First... (read more)


Am I allowed to play my own devil's advocate? Autodevil's advocate, if you will (writing down my ideas often helps me criticize them).

Aurini¹'s premise: Short term examples of Akrasia are due primarily to variability of self. Self¹ and Self² are both pursuing their own interests in a rational manner, it's just that their interests are dissonant.

I still think this is largely the case; most instances of regret are either "Knowing what I know now, I wish I hadn't put all my money in Enron," ie "I based my choices on incorrect data,"; or the other possibility, "I wish I hadn't done that last night, but if you press me, I'll admit that I plan to do it again tonight," the second may be foolish, it may be hypocritical, but it's not Akrasia per se, because the regret is temporary, not existential.

There is a third type, however, which is distinctly counter-rational. Well need an example: getting drunk the night before, and failing to show up to traffic court (thus defaulting on an $X00.00 fine which you could have avoided). All Self(x) where (xn) agree that this was a poor choice. While there are substantial differences of Self over time, and this does not... (read more)

While we're linking to webcomic strips, Miscellanea 2007-11-19 is also quite relevant to this.
What's wrong with the link http://kol.coldfront.net/comic/index.php?strip_id=188 or http://kol.coldfront.net/comic/istrip_files/strips/20090127.gif ? (Also: what the heck sort of comic numbering system puts comic #192 at ID #188?)
1. Did you just ninja me? 2. The type of comic written by someone who has no interest in self improvement. :)
If I knew what ninja meant, perhaps I could answer that.
You found a permanlink, while I couldst not.
1Paul Crowley
Wow, looks like their efforts to defeat permalinking were more thorough than we thought. This link now works: http://kol.coldfront.net/comic/index.php?strip_id=192
Insulting my future self like that sure makes me less anxious about providing for my future self.
Discounting your future yourself (e.g. thinking your future self is a dick) can be a strategy to work more efficiently now.

Akrasia is the tendency to act against your own long-term interests

No, akrasia is acting against your better judgment. This comes apart from imprudence in both directions: (i) someone may be non-akratically imprudent, if they whole-heartedly endorse being biased towards the near; (ii) we may be akratic by failing to act according to other norms (besides prudence) that we reflectively endorse, e.g. morality.

2Paul Crowley
In that case Ainslie has somewhat turned the word to his own ends. Is there a better word for what he's talking about? I don't criticise your definition for not cleaving at the joins because it absolutely seems plausible to me that there could be a unified theory of how we are about morality and how we are about our own long-term interests; the distinction between the two is scarcely made in much of what people say.
Inter-temporal conflict? (Part of the problem with misusing language is that it makes it unclear exactly what one has in mind. I assume Ainslie has a broader target than mere imprudence: foreseeable moral failures may provide similar reasons for precommitment, regret, etc. So perhaps he really does mean general akrasia, despite the misleading definition. But does he also take his topic to include 'murder pills' and ordinary cases of [foreseeable] changes to our ultimate values? Or does he restrict himself solely to cases of intertemporal "conflict" involving akrasia -- i.e. whereby both 'selves' share the same ultimate values, and it's simply a matter of helping them "follow through" on these?)
1Paul Crowley
His topic is specifically those changes of mind that we can anticipate because of hyperbolic discounting.
Okay, that sounds like 'imprudence', then.

I just have one question, it's so obvious but I don't remember it being asked anywhere.

Humans and all animals tested use hyperbolic discounting + hacks on top of it to deal with paradoxes. Why hasn't evolution implemented exponential discounting in any animal? Is it technically impossible the way brain works (perhaps by local optimum), or is hyperbolic discounting + hacks better in the real world than exponential discounting?

I think this is a far more fundamental problem than anything else about akrasia.

Reading the Wikipedia article on hyperbolic discounting it seems like there is some evidence for a quasi-hyperbolic discounting. Looking at the formula, the interpretation is exponential discounting for all future times considered but with a special treatment of the present. How to explain this? It is not unlikely that the brain uses one system for thinking about now and another about the future. Considering the usual workings of evolution, the latter is most likely a much later feature than the former. Considering this, one could perhaps even argue that it would be surprising if there wasn't any differences between the systems. There seems to be some literature referenced at the wiki article. I suggest looking into it if you are interested. I sadly don't have the time right now.
0Paul Crowley
I'm curious to know the answer to that one. My guess is that hyperbolic discounting is technically much easier to implement, and the circumstances of animals in the wild provide fewer opportunities for akrasia so it's not worth the cost of fixing. However there is doubtless room for more investigation of the evolutionary psychology of hyperbolic discounting. EDIT also see this comment

I used to have a system of implicit moral contract with myself.

I saw my own situation as an iterated prisoner dilemma; any of my future selves could desist against its other selves, negating the hopes and dreams of past selves, depriving further selves of certain prospects, all of that for a short term benefit that would have negative long term consequences. The first to desist would win something on the moment, the others loose their investment or their potential. So I tried to keep to my word and plan ahead.

Not sure if I'm still strong willed enough to affirm I'm working like that. Actually, probably not in most cases.

3Paul Crowley
Ainslie describes it more as a sequence of one-shot prisoner's dilemmas; because we're talking about agencies that differ as time passes, it doesn't necessarily make sense to think of them as existing continuously.

I found this article both interesting and informative. I definitely plan to spend some time studying picoeconomics.

One interesting effect that I have found in personal productivity efforts is that applying techniques to enforce resolution and overcome passive resistance can change the perceived emotional weighting between alternatives, often quite rapidly.

For example, let's say I'm reading LW instead of writing a term paper. I've made a (probably irrational) decision that the negative emotion of exerting the effort to write the term paper exceeds the n... (read more)


You've actually missed a key distinction here: the negative emotion of the incomplete assignment is almost certainly what makes you procrastinate... and you're mistakenly interpreting that negative emotion as being about the writing.

What happens is this: since you feel the unfinished item pressure every time you think about doing the task, you literally condition yourself to feel bad about doing the task. It becomes a cached thought (actually a cached somatic marker) tagging the task with the same unpleasantness as the unpleasantness of it "hanging over you".

So, it's not that the process of writing really bothers you, it's the unfinishedness of the task that's bothering you. However, your logical brain assumes that it means you don't want to write (because it doesn't have any built-in grasp of how emotional conditioning works), and so it looks for logical explanations why the writing would be hard.

When you're busy writing, however, you're not thinking about that unfinishedness, so it doesn't come up -- the somatic marker isn't being triggered. That's not at all the same thing as "shifting the balance".

The actual way to fix this is to make it so you don't fee... (read more)

6Paul Crowley
In the middle of some useful task, I have more than once said to myself: "I am not hearing any crap about how I never get around to anything, because however true it might be at any other time, I am doing something useful right now. Anything that needs to be said about the need to do things will have to wait for a time when I'm not doing things!" This works very well for me.
Thank you for your thoughtful response. As it happens, I disagree with your premise that the negative emotion of the incomplete assignment is almost certainly what makes me procrastinate. Yes, that's a potential factor, but only one of many. For example, there's the difference between anticipated and actual difficulty of performing a procrastinated task progress. But in the the spirit of rationality, I will give your suggestions a fair trial. You are absolutely correct that the most effective way to figure out what works is to use the scientific approach - design an experiment to test the hypothesis, test, assess the results, and go from there.
In this case, the hypothesis I bet on is: 1. You will identify a specific set of physical behaviors (muscle tension changes, viscera sensations, etc.) that accompany the thought 2. These behaviors are preceded by some mental representation (however brief) of some expected result -- such as being yelled at for not finishing the task, or some other social status-impacting event that could come about as a result of failing to complete it successfully or failing to complete it at all 3. Identifying and changing the thought process that led to creating and caching the expected outcome, will result in the cached thought going away, and possibly taking the somatic marker with it, or at least diminishing it in intensity. If the somatic marker remains or is replaced by a new one, there will be a new cached thought that goes with it. There are exceptions to this pattern; some somatic markers are straight-up conditioning (i.e., there's no cached predictive thought in play - the marker is directly tied to the initial thought), and some are rooted in what I call "holes in the soul" -- a compulsion to fulfill an emotional need that's not being otherwise met. But most chronic procrastination in my experience follows the "main sequence" I've outlined above. I used to waste a LOT of time helping people get over the "effort" they perceived associated with doing things... only to find out that it was 99% anosognosia -- misdirected explanations of the pain. If we don't feel pushed to do something in the first place, then we don't usually experience the time spent as being effortful. So nowadays, I get results a lot faster by focusing on eliminating a handful of feelings associated with NOT doing the task, than the seemingly infinite number of new complaints people can generate about DOING the task.
I really like pbjeby's advice and I think it applies to many types of procrastinating where we have negative feelings associated with the future task (which we can make worse through classical conditioning); the only part of his post I disagree with is his reduction of Mike's problem to negative emotion about the incomplete task. I agree with Mike that pleasure in our current activity can also be a part of procrastination, not just displeasure about the future task. For instance, I often have trouble stopping reading in order to do incomplete work, but I also have trouble stopping reading to make myself go to bed, even when I'm tired. Now, I enjoy sleeping, and I don't feel negative emotion about it: I just take even more pleasure in reading. Yet, my goal was to go to bed on time.
It can be... but rarely is in people who suffer from chronic procrastination. Usually, they don't enjoy the thing they're using as an escape. And I'm not aware of anybody who goes out of their way to do something they REALLY enjoy when they're procrastinating. Usually, they go for mind-numbing distraction rather than true involvement or enjoyment. For the most part, this is one of those areas where trusting your rational mind will lead you astray, because it's just telling you rational lies. It doesn't know what's actually going on, and so just makes up believable stories -- "that terrible LessWrong.com site tempted me and made me avoid my work..." And this is especially likely to be the case if you're also ashamed of the bad feelings you have about the task... Quite so... but that's not something I think of as procrastination. It might be akrasia, but if you told that story to a "real" procrastinator, they might find it insulting. One reason I'm a bit passionate about this, is that while the things you're saying may be true for you, they are not true for chronic procrastination, and not what a procrastinator needs to hear in order to get better. It's akin to telling an alcoholic that lots of people can handle their liquor. It might be possible to teach the alcoholic to handle their liquor, but that's definitely not the first order of business: detoxification is. Negative emotions (and "seriousness" in most forms) are to a procrastinator what alcohol is to an alcoholic: a drug addiction with serious real-life impact. (Really, it's only been in the last few weeks that it's even occurred to me myself that negative emotions and seriousness have some practical uses, once you're no longer addicted to them. And I'm still trying to get used to the idea, because I've spent the last year and a half or so trying to eradicate them from my life.)
Here's my current interpretation of where pleasure in the current activity comes into it for me: I would play a computer game, which I think should be pleasurable, and used to be, but feel guilty about procrastinating about something else, so I didn't enjoy it as much, or perhaps at all. If I think about stopping before I've gotten the expected enjoyment, that is unpleasant, so I would avoid stopping or thinking about stopping. I would stop eventually and feel bad about having wasted so much time. It's not so much that that I picked a game was inherently mind-numbing or unenjoyable, but that I turned it into something mind-numbing because I was avoiding these unpleasant thoughts.

Does hyperbolic discounting mean that the sunk-cost fallacy can be adaptive in certain situations, by "locking in" previous decisions?


I'd like to hear more about akrasia on LW. It seems to be supremely important.

I recommend this

Does evolutionary psychology provide an explanation for hyperbolic discounting? I found one explanation at http://www.daviddfriedman.com/Academic/econ_and_evol_psych/economics_and_evol_psych.html#fnB27 but it doesn't seem to apply to the example of preference reversal between sleeping early and staying up.

5Wei Dai
I asked earlier: There are actually many attempts to answer this question in the economics literature. I'm not sure why I didn't find them earlier. One paper is Uncertainty and Hyperbolic Discounting and others can be found by searching for papers that cite it.
1Bruno Mailly
Basically : In the ancestral environment, future gains were THAT unsure. BTW, I would not be surprised if evolution led to populations enduring bad seasons to become better at planning, especially long-term, and if this played a role in the enlightenment and industrial revolution. Edit : Cold climates demand more intertemporal self-control than warm climates

The problem I have with considering future discounting is that it forces me to formulate a consistent personal identity across time scales longer than a few moments. I've never successfully managed that.

To my horror, my future self has different interests to my present self

Can you describe 'my future self' without any sort of pronoun? If you could do that, the horror might, y'know, go away a bit. Thou art physics, after all.

as surely as if I knew the day a murder pill would be forced upon me.

Not quite as surely, otherwise you'd be taking steps to s... (read more)

Good points. I share your concern. But it's not clear which direction rationality cuts in this case. If I have no special attachment to the "me" of one year from now, why should I sacrifice present interests for his? On the other hand, I've been wondering recently if it's possible to salvage our folk concept of identity by positing that, while "me" at T2 might not be "me" in any robust sense, 1). there will be a person (or locus of consciousness, if you will) at T2 who thinks he's me, and shares many of my memories and behavioral predispositions, and 2). that person will be disproportionately influenced by my actions today. I think it follows from ethical considerations, then, if not prudential ones, that I should act today in a way that is in keeping with my best interests, so as not to unduly harm that future person. Now, what would really be interesting would be if we discovered that the "rational" thing to do would be some averaging of the two extremes -- i.e., I continue to act generally in my future best interests, but also prioritize present and near-term happiness to a much greater degree than seems naively appropriate.

I wonder if hyperbolic discounting uses the visual processing system? It certainly works like perspective foreshortening.

Richard Bandler's concept of "submodalities" strongly suggests a connection, since it uses image properties like distance, size, brightness, etc. to manipulate emotional response to imagined goals and behaviors. (An example of "front seat" driving -- i.e., directly manipulating the drivers of our behavior, rather than attempting to work around them.)

Excellent article and topic. I suffer from this. My main problem (which is merely an excuse) is that there is a difference between what I think I want to do and what my body and mind actually wants to do when it's doing the things I tell it to do. Multiple selves become evident when this happens. The self that has planned the actions, and the self that - in doing those actions - gives up to do other (more fun) things. My approach is to make successive changes to the self who does those actions that 'I' plan, by trying to implement rules for him to follow. But I find it a constant uphill struggle, because he always outsmarts me.

Would you mind expanding on this?

His extraordinary proposal takes insights given us by economics into how conflict is resolved and extends them to conflicts of different agencies within a single person, an approach he terms "picoeconomics".

I haven't read Breakdown of Will, but Thomas Schelling makes a similar proposal in his articles on "egonomics".

On the broader question of how to respond to irrationality, I strongly recommend Jon Elster's chapter (including the references) in Explaining Social Behavior.

I haven't worked out the mathematical details, but qualitatively it seems to me that we discount the future more than expected because we don't know what our desires will be like in the future (but nevertheless want to maximize the happiness of our future selves, whatever that may consist in). This means whatever action I take now to benefit my future self has an extra decrement in (present) utility because of my uncertainty in how much the future self will be benefited. And then there's the higher-order effect that the future self may have turned into ... (read more)

There are many different ways in which we could discount the future. The problem with almost all of them -- including the "hyperbolic" discounting Ainslie describes -- is not (necessarily) the mere fact that they discount, nor that they discount too much, but that they discount inconsistently: given times t1,t2,t3,t4, the relative importance of times t3 and t4 as seen from t1 is not the same as their relative importance as seen from t4. Or, to put it differently: if I apply t2-as-seen-from-t1 discounting together with t3-as-seen-from-t2 discounting, I don't get the same as if I apply t3-as-seen-from-t1 discounting. It is possible to discount the future consistently, but there's basically only one degree of freedom when you choose how to do so. If you give events a time t in the future weight proportional to (constant)^t then that's consistent. It doesn't open you up to the bug ciphergoth describes, where your judgement now is that times t1 and t2 are almost equally important, whereas when t1 comes along you regard it as much more important than t2. (If you don't discount at all, that's the special case where the constant is 1.) Ooo, no, actually you have more degrees of freedom than that: the most general scheme is that you choose a function F(t) and weight things according to that function. (Important note: one function, and its argument is absolute time, not time difference.) But the exponential case is the only possibility if you want your discounting function to be invariant if your whole life is shifted in time. (Which you might not -- if, e.g., there are external events that make a big difference.) Anyway, the point is: it's not discounting "more than expected" that's the issue, it's having a pattern of discounting that's not internally consistent.
Okay, I see how my comment was off-target. To explain the pattern described would require something more along the lines of "People know that the state of things (external or internal) can change quickly, yet over the long term tend to regress to the mean. Therefore they privilege the present over the immediate future, but regard two points in the far future as the same, having no way to distinguish between them." But that's both speculative and fairly empty of content.
1Paul Crowley
You mean F(t_1, t_2) where t_1 is the decision time and t_2 is the time of the event whose utility is weighed. Yes, that's the general form, but we assume that discounting is roughly constant across time (ie depends only on t_2 - t_1). I guess it would mesh with our instincts if discounting varied with age, but in the simpler special case where we consider only timespans that are short relative to our whole lives the theory works well; there's room to consider how this extends to a more general theorem.
No, that's too general; for instance, hyperbolic discounting is F(a,b) = b-a, but hyperbolic discounting is inconsistent in the relevant sense. For consistency we need F(a,b) F(b,c) = F(a,c), or equivalently F(b,c) = F(a,c) / F(a,b) = G(c)/G(b) where G(t) = F(a,t). (Note that the dependence on a has gone away.) This is equivalent to discounting things at time t by a factor G(t), which is the general form I described. Depending only on time differences is the same thing as being invariant under time-shifting your whole life.
1Paul Crowley
I started getting into this, but there's not really much point - the important thing is that we agree that if we require that preferences be invariant under time-shifting and not reverse as the choices approach, then only exponential discounting meets these criteria (treating not discounting at all as a special case of exponential discounting)

If we assume that (a) future discounting is potentially rational, and that (b) to be rational, the relative weightings we give to March 30 and March 31 should be the same whether it's March 29 or Jan 1, does it follow that rational future discounting would involve exponential decay? Like, a half-life?

For example, assuming the half life is a month, a day a month from now has half the weighting of today, and a month from that has half the weighting of that, and so on?

I was reminded of this post by a blog article I've just read: http://youarenotsosmart.com/2010/10/27/procrastination/ - it covers the same topic, but I think it presents it in an easier-to-grasp way for folks who aren't actively trying to be more rational.

Thanks for linking to that. It was helpful for me.

Excellent food for thought. I especially loved the point relating the distance of the reward to the actual rewarding process itself, and yes defeating Akrasia is the one thing that is (probably) most relevant to Lesswrong readers. This is because most LW-ers are by nature (probably) smarter than their immediate surroundings and it is not understanding of situations that is holding them back. It then (yes, probably!) boil down to either interpersonal skills and/or Akrasia. And the two are not completely mutually exclusive.

This is , at least at first glance, an important step in the right direction.

Ainslie has written quite a few interesting papers in the meantime: http://picoeconomics.org/articles2.html


I generally object to use of the term rational as a moral pejorative as the way its used in the article. We are all dealing with imperfect information, quantum uncertainty and human identity issues. We may suck at it, but the advice to just be more "rational" is insulting to people who are trying their best in an imperfect world. If you think discounting or valuing the future is so easy that you can bandy about words like rational, then I can show you how to make a killing in the mortgage backed securities market.

3Paul Crowley
Rereading, I do not think I could have done more to make it clear that I reject the moralism you accuse me of. As I say above, in large part I'm interested in this subject because I suffer from the failings it describes, and I think Ainslie goes a long way towards showing why it's not just a question of "trying to be more rational".