Enter Wikipedia:

In chemistry, activation energy is a term introduced in 1889 by the Swedish scientist Svante Arrhenius, that is defined as the energy that must be overcome in order for a chemical reaction to occur.

In this article, I propose that:

  • Every action you take has an activation cost (perhaps zero)
  • These costs vary from person to person
  • These costs can change over time
  • Activation costs explain a lot of akrasia

After proposing that, I'd like to explore:

  • Factors that increase activation costs
  • Factors that decrease activation costs

Every action a person takes has an activation cost. The activation cost of a consistent, deeply embedded habit is zero. It happens almost automatically. The activation cost for most people in the United States to exercising is fairly high, and most people are inconsistent about exercising. However, there are people who - every single day - begin by putting their running shoes on and running. Their activation cost to running is effectively zero.

These costs vary from person to person. In the daily running example above, the activation cost to the runner is low. The runner simply starts running in the morning. For most people, it's higher for a variety of reasons we'll get to in a moment. The running example is fairly obvious, but you'll also see phenomenon like a neat person saying to a sloppy one, "Why don't you clean your desk? ... just f'ing do it, man." Assuming the messy person indeed wants to have a clean desk, then it's likely the messy person has a higher activation cost to cleaning his desk. (He could also have less energy/willpower)

These costs can change over time. If the every-morning-runner suffers from a prolonged illness or injury and ceases to run, restarting the program might have a much higher activation cost for a variety of reasons we'll cover in a moment.  

Finally, I'd like to propose that activation costs explain a lot of akrasia and procrastination. Akrasia is defined as "acting against one's better judgment." I think it's possible that an action a person wishes to take has higher activation costs than they have available energy for activation at the moment. There is emerging literature on limited willpower and "ego depletion," here's Wikipedia on the topic:

Ego depletion refers to the idea that self-control or willpower is an exhaustible resource that can be used up. When that energy is low (rather than high), mental activity that requires self-control is impaired. In other words, using one's self-control impairs the ability to control one's self later on. In this sense, the idea of (limited) willpower is correct.

While this is anecdotal, I believe that starting a desired action is frequently the hardest part, and usually the part that requires the most ego/will/energy. Thus, the activation cost. Continuing in motion is not as difficult as starting - as activating.

This implies that there would be two effective ways to beat akrasia-based procrastination. The first would be to lower the activation cost; the second would be to increase energy/willpower/ego available for activation.

Both are valid approaches, but I think lowering activation costs is more sustainable. I think there's local maximums of energy that can be achieved, and it's likely that even the most successful and industrious people will go through low energy periods. Obviously, by lowering an activation cost to zero or near zero, it becomes trivial to do the action as much as is desired.

Some people have a zero activation cost to go running, and do it every day for the benefit of their health. Some people have zero activation cost to cleaning their desk, and do it whenever they realize its messy. Some people have a zero activation cost to self-promote/self-market, and thus they're frequently talking themselves up, promoting, and otherwise trying to get people to pay attention to their work. Most of us have higher activation costs to go running, clean a desk, or to market/promote something. Thus, it burns a lot more energy and is actually effectively impossible to complete the action sometimes.

The following factors seem to increase activation cost (not a complete list):

The following factors seem to decrease activation cost (not a complete list):

  • Deadline urgency
  • Constraints (and thus, lack of opportunity cost)
  • Momentum
  • Grouping/batching tasks together
  • Structured Procrastination
  • Very clear, straightforward instructions
  • Long term habits
  • Cached-self effects
  • Feeling like something is a game

Additionally, another way to go anti-akrasia is to increase energy levels through good diet, exercise, mental health, breathing, collaboration, good work environment, nature, adequate rest and relaxation. Some of these might additionally lower activation costs in addition to increasing energy.

I believe the most effective way to do activities you want to do is to decrease their activation cost to as close to zero as possible. This implies you should defeat ugh fields, reduce trivial inconveniences and barriers, de-compartmentalize (and get something to protect), untangle your identity from the action you're taking, and find as clear instructions as possible. Also, deadlines, constraints, momentum, grouping and batching tasks, structured procrastination, clear instructions, establishing habits, setting up helpful cached-self effects and reducing negative ones, and treating activities to be done as a game all seem to be of value.

I would be excited for more discussion on this topic. I believe activation costs are a large part of what causes procrastination akrasia, and reducing activation costs will help us get what we want.

New Comment
40 comments, sorted by Click to highlight new comments since:
[-][anonymous]100

I've thought in terms of activation energy for years (note: I'm not claiming priority, I'm sure I got it from somewhere else). It might not be falsifiable, and I strongly doubt that it's a grand unified theory of procrastination, but it seems to be a useful way of looking at things, and in some cases it is obviously perfectly accurate.

I have a couple of examples (filled with specifics; I've redacted nothing except company names).

Scheduling medical appointments isn't difficult, but its activation energy is nonzero (I have to call their office during normal working hours, which takes a few minutes), compared to the activation energy of doing nothing, which is zero. But going years without appointments isn't good. After I realized this, I set up a simple system. As I'm leaving my dentist or ophthalmologist, I'll go to their front desk and schedule my next appointment in 6 months (dentist) or 1 year (ophthalmologist). I'm already there, so the activation energy is epsilon, and this ensures that I schedule these appointments like clockwork. Later on, I'll call their office if I need to reschedule, which happens very rarely (and when it does, activation energy is not an issue; the penalty for not rescheduling a conflict is marked in my mind as "dire"; regardless of whether it actually is, that's a useful thing for me to think).

I don't use this system for my primary care physician or ocularist, though, and predictably I don't see them as regularly as I probably should. (There are reasons for the asymmetry: I don't own a car, only a Segway, so to travel more than a few miles I rely on favors from friends or family. My dentist is within Segway distance, and my ophthalmologist is critically important so I could never think of missing an appointment. But I don't need to see my primary care physician more often than every 2 years, which is too long to schedule an appointment in advance, and I can get away with not seeing my ocularist very frequently.)

My second example is more technical, but even more relevant. When I was at [College Name], I hosted my website on their servers, and later (because I had Ethernet and a static IP) from a spare computer reformatted with Red Hat 9. I was new to Linux at the time, so I barely knew what I was doing, but I had boundless energy and I got it working the way I liked. Of course, I didn't write down what I did.

Then I graduated, and moved 1000 miles to work for [Employer Name]. Despite living just a couple of miles away from their bandwidth-rich buildings, the Verizon DSL available to me was terrible, 80 KB/sec down and 16 KB/sec up (I later got Comcast cable, and finally Verizon FiOS at 25 Mbps up/down became available to me, but only a few months ago). I left my server with a grad student acquaintance for a few months, but I needed a permanent solution, so I went with [Dedicated Server Company]. I was used to having root on a Linux box, and my website was completely custom, so shared hosting was unacceptable to me. [Dedicated Server Company] looked like my best option in 2004, and at 325 USD/mo it was expensive but I could afford it. I set up this server, running Red Hat Enterprise Linux 3, and it was even more complicated this time. I set up my own E-mail, not having my college address anymore, and figuring that out was horribly complicated (although this was in the dark ages of 2004, I wasn't an idiot, and demanded encryption to and from the server; Gmail didn't start doing this for everyone until Jan 2010). And of course, I didn't write down how I set up the server, since I was figuring it out as I went. This was also on top of [Dedicated Server Company]'s extensive and incomprehensible customizations. And I modified lots of stuff in the years that followed.

Fast forward to 2009. I noticed how much money I'd spent on this server (325 12 5 = 19,500 USD over 5 years, eek) and how old it was getting. But setting up a new server and reconfiguring my website and E-mail seemed like an impossibly difficult task (activation energy!) given that messing around with Linux servers was no longer on my list of endlessly fascinating things to do. I'd been meaning to find new hosting, but I was procrastinating despite the financial expense.

At this point, it'd be neat to say that I figured out how to lower my activation energy so I could switch servers. But no. Instead, something happened that pushed me over this huge potential barrier: due to a server misconfiguration, I started using a ton of bandwidth, and [Dedicated Server Company] didn't notify me, and I went about 300 GB over, at a cost of 1.25 USD/GB. INFINITE RAGE. I swore that I'd leave [Dedicated Server Company] if it killed me.

It took me several more months, but I finally found [Virtual Server Company] through my best friend's recommendation. While visiting my place, he made me get out my credit card and set up an account with them. During the next couple of months, I despaired about how much work I'd have to do, and how busy I was with other stuff, while he cajoled me into permanently switching, and convinced me to use Google Apps for my mail (instead of running postfix on my own server). Finally, in December 2009, I broke the work of switching servers into a handful of tasks (transfer bash configuration, transfer Apache configuration, set up Google Apps, set up DNS, etc.) and completed one every couple of days. I was soon finished, with a fresh Ubuntu server, and once I was satisfied that I'd never have to look at my old yucky server again, I sent a wonderful mail to [Dedicated Server Company] canceling my account, and excoriating them for (a) their high cost (probably reasonable for a corporation; not for an individual - I was partially insane in 2004 though), (b) their bandwidth overuse charge (at 1.25 USD/GB, an order of magnitude greater than [Virtual Server Company]'s 0.15 USD/GB - being virtualized doesn't lower their bandwidth costs so a buck per gig in 2010 is definitely highway robbery), and (c) their utter inability to send an automated notification when I went over my bandwidth limit, despite the fact that it'd take a drunken teenager about 10 minutes to code that in Python.

Best of all, [Virtual Server Company] costs me only 18 USD/mo. My friend saved me 3684 USD/yr. (And reduced the necessary size of my retirement fund by about 74k USD, the way that I calculate it. That is a lot of money.) Procrastinating for so many months, if not years, was one of the most irrational things I've ever done (and I pride myself on my rationality), but the activation energy of maybe a day's worth of work in total seemed more expensive to me than thousands of dollars.

I did learn something to reduce my future activation energy, though. This time, I wrote down every single thing that I did while setting up my Ubuntu server, so that I could replicate the process on demand (to upgrade Ubuntu - I don't yet trust the process for doing so in place - or to switch companies, not that I'll be doing the latter, since I love [Virtual Server Company]). That's already come in handy as I've upgraded versions once. Whenever I modify my configuration (like when I discovered the magic trick to get "cls" working in PuTTY as it does in Windows), I'm careful to update my notes, so I don't lose anything.

Great examples. I had something similar happening when I shifted out of a business - I had a lot of services I was paying for, and put it off a few months even though it cost me some cash.

In retrospect, it's totally strange and I couldn't identify why. Activation costs probably explain a lot of that. Good suggestion about documenting what you did to make it easier next time, too.

That was really interesting to read, thanks. It sounds like you are or at least were at one point on the way to becoming a (semi)competent sysadmin. Any advice on how one does that for those who think it'd be nice to have a marketable skill that can be learned to employable levels in 3-6 months?

Any advice on how one does that for those who think it'd be nice to have a marketable skill that can be learned to employable levels in 3-6 months?

I don't think one can. Optimistically, one can learn the Unix environment (shells, scripting, the CLI tools, that sort of thing) in 3-6 months, but that's just a start.

Great Ghu. Okay, I believe. Does the answer change if employable is taken to mean worth hiring to do some low level work under supervision of someone who is genuinely competent?

That makes me hesitant to combine learning linux through Ubuntu and Python. Is that a reasonable hesitancy?

Does the answer change if employable is taken to mean worth hiring to do some low level work under supervision of someone who is genuinely competent?

I don't know. My imagination is impoverished, and I can't think of anything that your basic Linux poweruser is useful for that doesn't require more advanced sysadmin chops.

That makes me hesitant to combine learning linux through Ubuntu and Python. Is that a reasonable hesitancy?

Python (and most FLOSS languages) work best under a Linux, so it's not unreasonable. It's more reasonable than learning it under Windows, for example.

Power user! I don't think I've heard that word in a long time. Google Trends somewhat confirms my intuition, showing a fairly steady decline since they started keeping track. Unfortunately they only have data back to 2004, my intuition suggests that the word had already been in decline for some time at that point.

[-][anonymous]00

I picked up a skill point or two in server wrangling, but I'm nowhere near employable as a sysadmin (I'm a C++ programmer). So the only advice I can offer is generic: mess around with stuff that interests you until you get good at it. It's not very structured advice, to be sure. :->

Why would you be hesitant to learn (one or more flavors of) linux and python? I'd call such learning autotelic, but even if it's not fun for you, there are thousands of things in life (including but not limited to many diverse career paths) that are improved by the ability to get a computer to help you do stuff.

You should include your opportunity cost in the question, though. What's the next-best thing you will be (not "could be", but actually expect with the same likelihood as learning linux and python) doing if NOT learning these things. Learning windows administration details and C# is probably pretty valuable. Learning accounting standards and how to talk to business people is valuable. Actually working for someone and learning their specific business is valuable.

Reading less wrong and learning/discussing/exploring far-mode rationality is valuable too, but probably a lot less than the above options.

This question is offtopic for this post, but it could make a good discussion in an open thread. I'll give my answer anyway because I don't have the willpower to resist (or to phrase in the current metaphor, because it's exothermic to do so, and the activation energy is less than my current ambient heat).

I tend to agree with others that 3-6 months is ludicrous to start from scratch and end up with marketable technical competence (there are exceptions, of course - some truly unusual individuals or those starting with competence in some related endeavor might be able to do so).

If you're genuinely interested in the topic, you likely started years ago, and you really should ask "given where I am now, how should I focus my learning to get started in employment in the field". If you're not interested, it's hard to imagine that you'll ever be good at it, and you might should ask "how do I generate interest (in myself) for this topic so I can get good at it".

It might be possible to do it in a year, however:

(Perhaps with some additional techniques like spaced repetition?)

I would add "writing it down" to the decreasing activation cost list, or at least working it into "Very clear, straightforward instructions". For me, things generally get done much faster when I put them in my planner, or even just in a text file on my desktop, than when they're left to float around in my head. Whether that's because the task is made into clear and straightforward instructions, or because it creates a sense of urgency ("Oh man, it's written down, I gotta do it"), or just because it exists as a thing in the world rather than an idea, I don't know, but it's definitely a result.

Also, this article convinced me to finally stop procrastinating on setting up a LW account after a few months of readership. Hello!

Welcome!

I'm also a big fan of writing things down, just because it frees up that space in my brain for something else. Never mind the obvious reduced likelihood of forgetting about it. :)

I usually refer to this effect by noting that "context switching is expensive." I got it from CPU scheduling but it's also famously true of light bulbs: turning it off and then on again costs more energy than leaving it on for a short time. (I forget what the actual time threshold is offhand.) This is why grouping/batching tasks works--if you can do three things in a row in the same place, you don't have to switch contexts for each one. It's also a case against multitasking.

I would add to your hurts/help lists "having an overwhelming quantity of work" and "setting yourself clear, modest goals," respectively. I've been discussing elsewhere the task system I've been using for myself lately; the relevant part is that most of the time I only see my short "today" list of about 5-6 items. I don't even look at the long list of tasks to-do-ever, except when I'm adding something or picking out a few tasks to go on the next day's list.

Another suggestion for the helpful list: "novelty." That's what pulls your attention to google reader, but it's also what makes you feel more organized and productive when you're trying out a new organization system. I wonder if one could use this effect deliberately--change the format or style of your organization/task list periodically, to keep it fresh and interesting.

I usually refer to this effect by noting that "context switching is expensive." I got it from CPU scheduling but it's also famously true of light bulbs: turning it off and then on again costs more energy than leaving it on for a short time.

Joel Spolsky has a great article on this:
Human Task Switches Considered Harmful

Oh, spiffy. Maybe that's where the person I got the habit from heard it (although it doesn't really seem like his style). Thanks. :)

light bulbs: turning it off and then on again costs more energy than leaving it on for a short time. (I forget what the actual time threshold is offhand.)

IIRC, it's around 30 minutes for an incandescent and 2 hours for standard fluorescent. This is supposed to include the energy costs of manufacture vs the increased wear-and-tear. I don't have a citation handy, and I don't know how modern compact fluorescent bulbs change things.

change the format or style of your organization/task list periodically, to keep it fresh and interesting.

Hah, I actually do this because I'm terrible at holding to my organisational systems for more than a few days at a time, and usually assume that it's because my system wasn't awesome enough. Maybe I'll just start considering my need to tinker with my system as more of a feature than a bug.

Exactly! That's just the realization I had. It's kind of in the spirit of structured procrastination--put your wandering mind to work for you.

Just for the record, 'akrasia' is a stupid word. It fails to carve reality at its joints. It's dangerous to reify the lack of something, mostly because it's hard to catch subtle lack of reflectivity about that lack of something in the midst of more complex lines of reasoning. I'm reflective enough at the moment to know that my previous sentence was not an ironic example, but I can't trust myself to always be reflective enough when e.g. my thoughts are bent by affective bias, so in general I avoid 'stupid' words like 'akrasia' unless I am making conscious effort to be careful about reflectivity over concept clarification. Also, rationalists who are less experienced than me would probably still do it wrong even if they were being consciously reflective about it, so they should never use words like 'akrasia'.

Could you expand on your ideas for better ways to talk about difficulties with taking action?

I'd rather get a really solid framework for reasoning about 'dispositional skill' before I talk about 'akrasia' in particular; this will take a few months of hard work by me and others. We're currently doing a large-scale literature review before we start playing around with our own unifying theories, and the review/learning process should take a month or more. A lot of it will involve talking to domain experts, et cetera.

  • Every action you take has an activation cost (perhaps zero)
  • These costs vary from person to person
  • These costs can change over time

Wow, is there anything your theory can't explain? Seriously, I think "activation costs" are so broad as to be useless.

Wow, is there anything your theory can't explain? Seriously, I think "activation costs" are so broad as to be useless.

Okay, I'd like this to be useful. What do you recommend we do for elaboration, detail, data, processes, or otherwise to explore this in a more useful way?

I can't buy this as anything but a fairly superficial metaphor.

Activation energy for chemical reactions is a useful model because it uses measurable units on the same scale as output energy, and because it's very repeatable from one experiment to the next. You can make very good predictions about how to start (or fail to start) reactions using this theory.

I'm not sure what units of energy are used to measure the activation cost you're talking about, and it seems that the costs vary across time, individual, and situation to the point that this theory can't make any good testable predictions.

It would go a long way to make some testable predictions based on this, or some evidence that this model can be used to do something.

I'm not sure what units of energy are used to measure the activation cost you're talking about, and it seems that the costs vary across time, individual, and situation to the point that this theory can't make any good testable predictions.

Hmm, actually, I wonder if you could do a self-assessed survey of "how hard" something seems to be to start doing. It'd face all the problems of self-assessment, but maybe you'd come out with some interesting data. Then you could track some related variables - how competent a person feels at it, identity-related, how often they've been doing it recently, etc. Might be able to come up with some interesting data.

It would go a long way to make some testable predictions based on this, or some evidence that this model can be used to do something.

Indeed. This was just a starting point. Also, I was going more for practical instrumental value than for epistemic, but I think there'd be value in expanding. Good comment, cheers.

I should note that I do like and use this as a metaphor in casual conversation, mostly as a way to ask my spouse or friend if they'd like to add motivation in some way to get me over the hump to start doing something I think benefits us both.

However, I tend to think of it as a shallow similarity, and I'm hesitant to try to draw deeper inferences from chemistry into motivation theory. I don't know if the anti-akrasia tactics being mentioned by many comments here are reducing the activation energy, adding energy to cause the reaction, or providing an alternate reaction path like a catalyst. I don't even know how to measure it so that I can determine which of these (if any) are in play.

Oh, God, yes. Sometimes it feels like I have to haul myself over a threshold to get myself to do things.

I definitely face this issue, I find it difficult to do anything I'm not in the habit of doing, especially if I'm doing it for the first time.

The single most useful tactic I have for lowering activation costs is playing podcasts and audiobooks on my iPod. It lowers the activation cost for household chores like dishes significantly, and it is particularly effective in lowering activation costs for outdoor exercise. Impending boredom is the highest part of the activation cost for tasks that involve leaving my computer.

[-]knb30

One minor hack to lower activation costs is to just resolve to do the easiest possible thing to start a project when you're procrastinating. So if I'm web-surfing and I know I should start that report, I'll just open up a text file, give it a name, and create the title page, and leave a shortcut on the desktop. It makes it much easier when I decide to go back to the project.

Actually, the same day you posted your article, Jess_Riedel posted this article on Willpower..

As a frequent procrastinator, I can affirm that to start is often the most difficult part (but not always). However, I would like to see your theory stated in less vague fashion. It would be helpful to know whether one can measure the activation costs (in approximation at least) or whether the hypothesis can be formally included into a utilitarian decision theory, or most generally, whether there is a testable prediction that follows from the existence of activation costs.

This implies that there would be two effective ways to beat akrasia-based procrastination. The first would be to lower the activation cost; the second would be to increase energy/willpower/ego available for activation.

Let's say I want to do some work W and I tend to procrastinate. In order to make it easier, I apply the technique T. In retrospect, T proves to be successful. It is either because T lowers the activation cost of W, or because T increases my willpower. How do I recognise which was the case?

Very, very good comment. I like your first suggestion. As for the second question, I'm not sure yet. I'll give some thought to this.

Great feedback, really good feedback. Thank you.

I like this metaphor a lot (though I agree with critical replies that question its value as a theory.)

Speaking of reducing activation costs, breaking big tasks down into atomic physical doable actions and figuring out the 'next action', GTD-style, can greatly reduce activation cost. GTD's concept of "next action" was extremely useful to me -- it was the missing key piece of my own productivity system (which, as it turned out, is pretty similar to Allen's GTD in many areas).

Speaking of reducing activation costs, breaking big tasks down into atomic physical doable actions and figuring out the 'next action', GTD-style, can greatly reduce activation cost.

Technically, GTD only moves the activation cost around, though it does reduce switching costs by batching up the processing.

In other words, the "activation cost" of a vague/unspecified task is that you must first make it clear and specific. GTD doesn't eliminate this cost, just batches it up and advises you get in the habit of paying it early rather than late.

Wow, amazing post! I actually thought of the EXACT same idea too - I'm surprised that someone came up with it independently of me

Anyways, my idea is listed here: http://talk.collegeconfidential.com/college-confidential-cafe/391872-model-activation-energy-applied-motivation-theory.html (not intended to compete, but some may find it interesting). It might have some more analogies that might be interesting. (e.g., as a person with ADD, I'm especially constrained by activation energy, so I frequently try to find ways to lower it). Automating one's behavior is a way to lower it.

In short, anxiety, fatigue, and ADD can easily increase activation energy. Stimulants (like ADD drugs), automation, and stress-reducing tactics decrease activation energy.

Or in Robin Hanson's terms, "near" and "far". Things that are "near" require less activation energy than things that are "far".

I like this analogy. I'd like to add that activation energy is required, but often not spent, and some activities (such as exercise) leave you with more energy than you started with; having a high activation energy is not at all the same as being draining.

What do you intend by the analogy? That the instantaneous effort it takes to sustain a course is less than it takes to commit to and begin it?

I agree that where possible it makes sense to order your life so that what you should do is in fact easy and even pleasurable. I'm not worried that, if I become too comfortable, my willpower or desire for novelty will atrophy, but if I were, I would set up a metaphorical alarm clock to wake me from my too-comfortable routine.

All discussion on this post is very welcome - I hope we can build more on this topic.

Re: my last post, it suffered from three points - it could have used more editing to make my key points more clearly, it should have included practical alternatives, and the examples should've been more generic. I tried to spend more time with this one and not repeat those errors. I appreciate all the feedback and replies on the last post, and I might try to write a follow-on post about deontological ethics, consequentalism, and how context matters for choosing actions. Thanks for all that feedback and discussion, I'll go through all of it very closely when I have a little more time, and might try to write that post on ethics and context if I think there's a way I could do it at a high standard.