The irony is not lost on us that this post has been languishing as a draft + pile of notes for so long that it has a title image from the era when a human illustrated Beeminder’s blog.
The Principle of Delayed Commitment says to consider whether a decision or task will be worse or harder if delayed. Often it won’t be, so delaying is a no-brainer. Procrastination isn’t inherently bad. It’s bad when it causes actual badness. Consequentialism! People commonly overgeneralize the virtue of prudence/preparedness to cases where it’s blatantly irrational.
Yes, we know, this is Decision Theory 101 and intuitively obvious for some people. Optionality has value, the end. But writing it out for a lay audience is the best way we know of to establish it as a concept handle, and we’ve found it to be a useful one. In particular, principlifying it helps catch cases you might otherwise miss. “By the Principle of Delayed Commitment, let’s not bother to do XYZ yet.” Saying it that way clarifies that you’re not (just) procrastinating. That we might do XYZ differently (or not at all) depending how other things play out.
“People commonly overgeneralize the virtue of prudence and preparedness to cases where it’s blatantly irrational”
This can absolutely be taken too far. Some of us (especially me, dreev) may use it to rationalize procrastination beyond what’s actually rational. (“What if a meteor destroys the city before April 15 and I did my taxes early for nothing?”) As Scott Alexander astutely asks, “Should You Reverse Any Advice You Hear?”. Some people need to be more prudent and pre-planny and others do that too much. But there’s a theoretical sense — and for plenty of people a practical sense — in which it’s true and useful.
Toilet seat toggling
Here’s a simple example to start with: Don’t toggle the state of the toilet seat — up vs down — after you use it. You don’t know who will be next to use the toilet, and therefore don’t know what state they’re gonna want it in, so don’t waste effort lowering it when you’re done. What if you’re the next person to use the bathroom? You’d have lowered it for nothing!
Though huge caveat here: actually toilets are gross and you should put a lid on that germ bowl to separate it from the rest of your bathroom.
But in principle, toggling the seat is wasted effort.
The same applies to other shared toggles as well. Should you switch the TV back to cable input after you finish playing games? Nope. Should you switch the seat position and mirrors back after driving if you share the car with someone who is differently sized than you? Nope.
The kids’ carseats
Real world example from Bee, whose parents drive her niece and nephew around town often enough to have their own child carseats in their car. They leave them installed in the backseat most of the time.
They don’t take the carseats out every time they drop the kids off in case they need to give an adult a ride before they see the grandkids again. And when Bee gets a ride from them to go out antiquing or to a choir concert, and needs to use the back seat, she doesn’t reinstall the carseat when she gets out of the car.
On both ends of the transaction they wait until the next time the backseat needs used, and toggle the state — carseat or no carseat — appropriately. Unless it’s certain who is going to be the next passenger in that back seat, they delay commitment and deal with it in the future.
Bridge crossing
Sometimes when you’re riding your double-decker bike to the neighboring town across the river, you think to yourself, “I need to bike over that bridge. In order to get across the river.” And you may be tempted to get that part of your journey out of the way early.
But we propose that it’s more prudent to cross the bridge when you come to it.
(The related aphorism, about enumeration of pre-hatched chickens, says to not depend on things that haven’t happened yet, as opposed to not doing things it may not make sense to do yet.)
You may be wondering at this point if we actually need the “principle of delayed commitment” concept handle when there’s already the perfectly serviceable “cross that bridge when we come to it”. Well, “need” is a strong word but we use both of them. We’d say the bridge version feels more casual — “let’s not worry about that yet” — while the principle version prompts us to think about what’s decision-theoretically optimal.
Premature optimization
Software devs will be familiar with this. Indexing your database before you actually have any data in it. Setting up geographic redundancy before you even have users outside your timezone. Caching too soon. Adding features or building out too much functionality before you’ve had actual users. See especially the Shirk & Turk Principle.
Pre- or over-planning
As a real-world counterpart to the software version above, it can be tempting, especially if you are an anxious person, to expend a lot of effort planning for events that may never come to pass. Applying the principle of delayed commitment can be useful if overplanning is your tendency.
Don’t borrow chairs from your neighbor for your dinner party until your RSVP count actually exceeds your chair capacity. Don’t buy a full kosher set of dishes and build a second kitchen until the Rabbi agrees to come to dinner. Don’t buy plane tickets to attend your grandmother’s funeral until she’s actually dead.
Some of these are more obvious than others. But, again, some people in some situations are prone to blatantly violate the principle and calling it a principle is sometimes helpful.
Lazy evaluation
Here’s another one from computer science.
Suppose I ask you if it’s true that either 248 is an even number or there are infinitely many twin primes.
Lazy evaluation means noticing that the or-statement is true as soon as you figure out that 248 is even, so don’t bother solving the twin prime conjecture.
That example is contrived and ridiculous but this comes up constantly in programming.
Statments like if(x > 0 and long_computation() == x) can be short-circuited and the long_computation() avoided whenever x is not greater than zero.
Delaying those ticket purchases
There are lots of reasons to buy tickets early — for concerts or flights or whatnot. The price might go up, for one thing. And you get more choice in seating or departure time.
But the principle of delayed commitment tells us to trade that off against the advantages to delaying: Something might come up to keep you from going to that wedding. Is there only a 50% chance your cousin will actually go through with it? Are your mother and her brother constantly fighting and your whole family might get uninvited? Are you getting a new boss next month and you’re not sure if you’ll be able to get time off work?
If there’s additional information that might decrease your excitement about attending (maybe the concert set list, or opener), or that might increase your excitement or otherwise change your decisions (you find a friend who wants to go too, but only if you can get seats together), then there’s value in delay.
Ok, the principle of delayed commitment says it can be rational to delay in order to preserve optionality. Ironically, and especially relevant to Beeminder, there’s also a meta-rational reason to delay, in order to remove optionality. I (dreev) blogged about this the other day on LessWrong as part of my stay at Inkhaven: “Strategically Procrastinate as an Anti-Rabbit-Hole Strategy”. It points out that delaying starting a task can serve as a commitment device to cap the amount of time you spend on it. Just-in-timeboxing, we’re tempted to call it.
Final question: Does Beeminder not blatantly violate the principle of delayed commitment, locking you in to a course of action that might turn out to be suboptimal? It sure does! We’ve thought about this forever and our best answer is still Flexible Self-Control. It’s worth “irrationally” pre-committing to thwart your impetuous future self. But retain as much flexibility and optionality as possible. Beeminder does that with an akrasia horizon: You’re only ever on the hook for the upcoming week. So you can change what you’re committed to at any time and that change takes effect with a one-week delay.
Coauthored with Bethany Soule. Thanks to Mary Renaud for helpful discussion. Image credit: Faire Soule-Reeves.