Heuristics for Deciding What to Work On

If you’re like me, you have way more ideas for things to do than time, en­ergy, and willpower to do them with. (And if you’re not like me, you might very well be­come like me if you just kept track of all the times you or some­one else said “Hey, that might be a worth­while pro­ject.”) To give you an idea of what I’m talk­ing about, here are some en­tries on my things-to-pos­si­bly-do list: give speed read­ing an­other shot; im­prove the Less Wrong code­base and add a fea­ture that helps users find old, good posts they haven’t read; ex­per­i­ment with on­line free­lanc­ing work; try my hand at e-com­merce; work as a sales­per­son to build so­cial skills.

One of the things I’ve learned from keep­ing a things-to-pos­si­bly-do list is that do­ing stuff in­evitably takes longer than I in­tu­itively think it will. For ex­am­ple, the main thing I did dur­ing the past 3-day week­end was write 36 Anki cards and 220 lines of Python to pro­gram my­self and my com­puter to help me keep a re­s­olu­tion. In past years, I might have got­ten de­mor­al­ized halfway through, think­ing things were tak­ing too long, but I’ve grad­u­ally got­ten used to things tak­ing longer than I ex­pect.

Given that things take such a long time to get done, it seems worth­while to spend a de­cent amount of time de­cid­ing what to work on. But the stan­dard ob­jec­tive of do­ing what­ever has the high­est ex­pected util­ity is of­ten com­pu­ta­tion­ally in­tractable in prac­tice. For ex­am­ple, what’s the ex­pected util­ity of build­ing so­cial skills?

Given this, I’m work­ing on a list of heuris­tics for the com­pu­ta­tion­ally in­tractable prob­lem of what to work on. Here’s my cur­rent list; feel free to sug­gest ad­di­tions in the com­ments.

  • Watch for in­vest­ments that pay for them­selves. For ex­am­ple, I have two close-to-iden­ti­cal keys that I use mul­ti­ple times daily. About a week ago I taped some pa­per to one of them. I wouldn’t be sur­prised if this time in­vest­ment will have paid it­self back within a few weeks, and start gen­er­at­ing value from then on.

  • Find things that sep­a­rate those who suc­ceed from those who fail, and fo­cus on get­ting those right. For ex­am­ple, I’ve ob­served that men who have suc­ceeded in busi­ness seem more likely than reg­u­lar peo­ple to dis­play high-sta­tus be­hav­ioral cues.

  • Find things that will give you a lot of new in­for­ma­tion in an im­por­tant do­main even if their im­me­di­ate ex­pected value may be low. (An ex­am­ple might be buy­ing a few web­sites on Flippa to learn what strate­gies work on the web in­stead of go­ing through the rel­a­tively la­bo­ri­ous pro­cess of writ­ing a web­site that might fail your­self.)

  • Find things that will start gen­er­at­ing con­tin­u­ous benefit as soon as you do them, and do them first if they’re worth do­ing at all. For ex­am­ple, I’d guess that for most, in­stal­ling LeechBlock or a similar anti-time-wast­ing tool can po­ten­tially gen­er­ate con­tin­u­ous benefit for quite a while, es­pe­cially if you in­stall it while in a re­s­olu­tion-mak­ing mind­set. (Of course, an ar­gu­ment against do­ing all ac­tivi­ties of this type first is that you might be op­ti­miz­ing pre­ma­turely. Con­sider the case of a self-help reader who judges self-help tech­niques based on the num­ber of hours they help him spend read­ing self-help books.)

  • Work on things you’ve been do­ing re­cently, so you aren’t pe­nal­ized by con­text switches. (Un­less you’ve reached some sort of in­tel­lec­tual dead end where re­vis­it­ing the prob­lem later would ac­tu­ally re­sult in in­creased effec­tive­ness.)

  • Do things at the best pos­si­ble time. For ex­am­ple, maybe you’re a PhD stu­dent with an un­pub­lished book you’ve writ­ten and you’re not sure whether you want to spend your sum­mer try­ing to get your book pub­lished or trav­el­ling. Us­ing this heuris­tic, you might de­cide to put off try­ing to get your book pub­lished un­til you’ve got your PhD, with the ra­tio­nale that it will be eas­ier to find a pub­lisher then. (Of course, it’s pos­si­ble that pub­lish­ing your book is so high-value that you shouldn’t put it off, and given the scale of this ex­am­ple it’s prob­a­bly worth do­ing more com­pu­ta­tion to de­ter­mine this.)

  • Do what you want to do—peo­ple tend to be bet­ter at things they like do­ing.

  • In­tu­itive es­ti­ma­tion of ex­pected value.

And here are some ways you could use these heuris­tics:
  • Rank ev­ery­thing on your to-pos­si­bly-do list ac­cord­ing to each heuris­tic. Do the top item ac­cord­ing to each heuris­tic. (The ra­tio­nale be­ing that even if some heuris­tics turn out to be way bet­ter than oth­ers, this ap­proach will gen­er­ate a de­cently good re­turn.)

  • Deter­mine some weight­ing for each heuris­tic. Score each item on your list ac­cord­ing to how well it satis­fies each heuris­tic, then use the weight­ings to de­ter­mine an item’s over­all score.

  • (Prob­a­bly the best...) In­tu­itively rank the items on your list, keep­ing these heuris­tics in mind. Spend rel­a­tively more time figur­ing out the po­si­tions of the top items on your list, since those are the ones you might ac­tu­ally have time to do.

Fur­ther Reading

  • Tom McCabe’s Levels of Ac­tion post touches on the idea of ac­tivi­ties that provide new in­for­ma­tion be­ing high-value.

  • Cal New­port says you’re best off putting a lot of effort in to a very small num­ber of pro­jects. This im­plies that choos­ing one’s pro­jects is im­por­tant, and is also re­lated to the idea of things tak­ing longer than ex­pected to get done.