Probabilistic Alarm Clock

by Henry on January 24, 2007

Lifehacker links to an invention that I’ve thought for years would be a good idea (I’m sure that plenty of other people have had the same thought). Many people have their clocks running a few minutes fast, to encourage them to leave earlier for appointments to get there on time etc etc. The problem with this is that if you’re half-way rational, you’ll correct for the error, making it useless. So the solution is to have a probabilistic clock, where the clock is fast, but you aren’t sure how fast it is within a given and relatively short time range. Thus, you’re more likely to depart early for your appointments and get there on time (or a few minutes ahead, most probably, in many situations). This is exactly what some bloke has programmed, although it doesn’t appear that it has an alarm feature yet.

{ 30 comments }

1

Maria 01.24.07 at 11:13 am

Last weekend’s FT had a Dear Economist (i.e. Dear Tim Harford letter) from someone who always sets his watch fast and still manages to fool himself into being on time. ‘Mark’ wondered how this was possible, what with him being a rational actor who writes to economists asking for life hacks.

The answer was “you have a split personality, a warped view of time and are too lazy to do simple sums. Now put down this magazine: I suspect you are running late for something.”

2

shreeharsh 01.24.07 at 1:15 pm

This is embarassing to say but I never thought of a probabilistically faster clock, despite my engineering degree and all. I’ve done the same thing this past month — speeded up my wrist-watch, but ended up being late anyway because I know it’s faster!

Still, this sort of brings up another point. Now my wrist-watch is the old variety, with hands, and time-marks arranged in a circle, which means that the lowest time-interval you can accurately measure is 5 minutes. When I set it to run fast, I didn’t want to run too fast, thereby getting there early (smart, huh?), so I set it to run faster by something less than 5 minutes. What this something is, I can’t recall — and really, no one thinks of measuring minutes except in multiples of 5. Therefore my watch does seem to be running probablistically faster, since I don’t know exactly how fast it is. No? Sadly the range of values (1 to 4 minutes) cannot be extended — setting the watch faster by 11 to 14 minutes has the same effect (since the rational actor will cancel out the 10 minutes because it’s a multiple of 5 and thereby easily taken care of).

Still, the rational actor proves hard to overcome. Now he simply takes out his cellphone and checks the time!

3

luci 01.24.07 at 1:50 pm

This has occurred to me too, for years. One solution, have a significant other set the clock for you, some random amount of time between 5 and 15 minutes fast. Or, you can close your eyes, and try to do it yourself.

It’s still difficult because of the number of clocks around you at any time (your computer, television, in your car, etc).

4

abb1 01.24.07 at 2:00 pm

The problem with this is that if you’re half-way rational, you’ll correct for the error, making it useless. So the solution is to have a probabilistic clock…

If you had this probabilistic clock, you would be correcting for the average error, making it worse than useless.

5

tom s. 01.24.07 at 2:10 pm

My reaction was the same as that of abb1. Is there something wrong with that logic?

6

greensmile 01.24.07 at 2:10 pm

Now that is REAL human factors based engineering.

7

Charlie Whitaker 01.24.07 at 3:16 pm

4: But that will sometimes make you late: knowing this is your incentive to always believe the clock, just in case.

8

billy 01.24.07 at 3:32 pm

I think you dont understand how the ‘few minutes fast’ (fmf) thing works.

Mostly, the fmf and correct clocks are the same due to the mental correction you cite.

The key difference is this
-most days, people wake up, snooze, wake up, snooze and eventually get up at the right time, irrespective of whether it’s a fmf/correct clock. The first wakeup – whether fmf or correct – is never a crisis. You go back to snoozing.

-on those problem days, this cycle, for some reason does not happen – the first time you get up, the grace period for snoozing is over/reduced. There needs to be an abrupt mental adjustment to accept this. The fmf provides that shock. Of course, after the shock, the mental arithmetic comes into play, but by that time the shock would have corrected the snooze expectations.

So the fmf is a means to set the snooze expectations on first wakeup

9

Chuchundra 01.24.07 at 3:34 pm

Watches are a confidence trick invented by the Swiss.

10

Walt 01.24.07 at 4:20 pm

Why would you average? You have a job interview at 10am. If your clock is right +- 10 minutes, are you going to show up at the average (which is that your clock is right), or 10 minutes before you clock says?

11

radek 01.24.07 at 4:41 pm

I used to do this self defeating trick and now I use it in the classroom as an illustration of how a monetary expansion in the presence of money illusion can give a boost to the economy in the short run but just leads to more inflation and price signal confusion in the long run.

Say you set your alarm clock ahead 10 minutes right before you go to bed (money supply up). Next morning you wake up and being dozy you fail to remember that the time is wrong – so you get up 10 minutes earlier (output up). Next morning however, having had the experience once before, you wake up, take a look at that clock and say to yourself “that thing is 10 mins too fast” and go back to sleep (inflation up, output back down). What to do? Well, you could set it ahead by another 10 mins (or whatever – just pick a new point on the now shifted Phillips Curve). It works for another morning, but the day after that you wake up thinking “wait, is the clock now 10 mins fast, or 20 or 30? Did I reset it again? What is the freakin’ time????”

Making it probabilistic is just the application of the Lucas principle that in order for monetary policy to be effective the monetary authority has to constantly surprise people in the economy else they will incorporate its behavior into their expectations (“that clock is 10 mins fast”) and the policy will have no real effects. But of course, “randomness” is no way to conduct a good monetary policy, and likewise it probably isn’t a good strategy for waking up in the morning.

Which is why now instead I just have my girlfriend call me a couple of times in the morning, in addition to the alarm clock. You can turn the alarm clock off but if I don’t answer the phone I’ll be in the doghouse. Maybe this is like a credible committment to a monetary rule.

12

abb1 01.24.07 at 5:04 pm

If your clock is right +- 10 minutes, are you going to show up at the average (which is that your clock is right), or 10 minutes before you clock says?

No, your clock is not +- 10 minutes, it’s always ahead, it’s ahead by between 5 and 15 minutes and you know it. So, you figure, if you’re 10 minutes late by this clock, you’re probably OK. So, sometimes you’ll be a few minutes late in real time.

13

radek 01.24.07 at 5:17 pm

And people on here were arguing that the Rational Expectations Hypothesis doesn’t hold!!!

14

dearieme 01.24.07 at 5:25 pm

I run my office clock a little slow. So if I arrive a minute late and someone is already waiting, I walk in with him, point to the clock, and thank him for his being there a minute early.
It seems to work.

15

David Margolies 01.24.07 at 5:25 pm

I always wondered if I was rational, or even half way rational, and now I know: according to the experts, I am not. (I use the trick and do not correct, or, more exactly, do not correct until I am out of bed.)

16

a very public sociologist 01.24.07 at 6:39 pm

Aaargh that invention would be an utter nightmare! I rely on my alarm clock being approx half an hour fast! My spine ashivers at the thought!

17

engels 01.24.07 at 6:53 pm

No, your clock is not +- 10 minutes, it’s always ahead, it’s ahead by between 5 and 15 minutes and you know it.

I don’t think that matters. Let’s say the meeting is at time m and whatever time you try to arrive at (a) you will end up arriving at a+r, where r is in the range -5 to 5 (because your estimate of the time may be out by +/- 5 minutes). You will not (rationally) set a to m, because the penalty for being late is larger than the penalty for being early. Depending on the relative size of these penalties you will set a to some number less than m, possibly to m-5.

18

engels 01.24.07 at 6:54 pm

ooh

No, your clock is not +- 10 minutes, it’s always ahead, it’s ahead by between 5 and 15 minutes and you know it.

I don’t think that matters. Let’s say the meeting is at time m and whatever time you try to arrive at (a) you will end up arriving at a+r, where r is in the range negative 5 to 5 (because your estimate of the time may be out by plus or minus 5 minutes). You will not (rationally) set a to m, because the penalty for being late is larger than the penalty for being early. Depending on the relative size of these penalties you will set a to some number less than m, possibly to m minus 5.

19

engels 01.24.07 at 7:04 pm

Then again, if you were rational, you’d just leave on time in the first place.

20

Slocum 01.24.07 at 7:52 pm

But nobody wears a watch anymore, do they? Don’t we all have wireless, auto-synchronizing pocket watches (AKA cell phones) instead?

21

tom s. 01.24.07 at 8:45 pm

11. “Which is why now instead I just have my girlfriend call me a couple of times in the morning, in addition to the alarm clock.”

– which is fine as long as your girlfriend does not suffer from the same problem. But then, I guess you could call her.

22

radek 01.24.07 at 9:41 pm

Arbitrage over time differences. And the fact that she has a pet that needs to be fed early in the morning.

23

nbrej 01.24.07 at 10:38 pm

I love you, Henry.

24

abb1 01.25.07 at 3:18 am

OK, engels.

Why do you need the trick in the first place? Because when your clock shows exact time you are often late.

Why are you often late when your clock shows exact time? Because you’re irrational.

In what way are you irrational? You are irrational in the sense of being too optimistic; IOW: when you calculate the odds, you make systematic mistake assuming they are too favorable.

I’m saying that this probabilistic clock just gives you an additional opportunity to miscalculate the odds, and, obviously, with the same bias we already know you have (too optimistic), and so you’re probably going to be late more often than usual.

25

bad Jim 01.25.07 at 4:38 am

The proliferation of nearly perfect timepieces (mobile phones being the handiest example) is obsolescing this strategy.

I’ll note my preference at the outset: I have two radio-controlled clocks, a clock radio and a watch; my computer contacts the NIST timeserver daily; the landline phones get the time via caller ID; the cable tuner’s time is set remotely. (The thermostats, ovens, coffee maker and sundry other timepieces are manually adjusted and undependable, and yes, changing to and from daylight savings time is a major festival in my household.)

In my experience, people who try to compensate for running late by fooling themselves continue to run late, probably because it’s hard to forget that you’re lying to yourself.

For a commuter, an alarm connected to the net and tracking traffic delays could conceivably be useful, but the option of an additional aleatory uncertainty would not.

26

Doug 01.25.07 at 5:28 am

The proliferation of nearly perfect timepieces (mobile phones being the handiest example) is obsolescing this strategy.

Clearly what we need is an improbabilistic alarm clock. What would Douglas Adams do?

27

Stuart 01.25.07 at 6:54 am

Surely a rational actor is only late for meetings that they don’t really care much about being late for, at least thats how it works for me, so having unreliable time pieces would just act as an excuse for being late (I certainly know some people who use to deliberately have their watch running several minutes late so they can ‘correct’ it when they arrive a few minutes late for a meeting, this would seem to be able to be used for the same purpose).

28

MikeN 01.28.07 at 6:51 am

Any idea how many people do this? My family did and my wife’s family did, so we do too, but any info on what percentage does? I’ve been unable to come up with a Google to answer this.

29

Alex 01.28.07 at 1:12 pm

Anyway, all users of South West Trains have a similar but opposite system – we get up at the same time in the morning but the trains are delayed by a random number of minutes between zero and 30.

30

Dave Cole 01.29.07 at 6:05 am

What about an alarm clock that tuns away when you press snooze? No, really, it exists.

http://www.nandahome.com/

Comments on this entry are closed.