In­side/​Out­side View

TagLast edit: 1 Oct 2020 18:25 UTC by Ruby

An In­side View on a topic in­volves mak­ing pre­dic­tions based on your un­der­stand­ing of the de­tails of the pro­cess. An Out­side View in­volves ig­nor­ing these de­tails and us­ing an es­ti­mate based on a class of roughly similar pre­vi­ous cases (al­ter­na­tively, this is called refer­ence class fore­cast­ing).

For ex­am­ple, some­one work­ing on a pro­ject may es­ti­mate that they can rea­son­ably get 20% of it done per day, so they will get it done in five days (in­side view). Or they might con­sider that all of their pre­vi­ous pro­jects were com­pleted just be­fore the dead­line, so since the dead­line for this pro­ject is in 30 days, that’s when it will get done (out­side view). Another term for tak­ing an out­side view is refer­ence class fore­cast­ing.

The plan­ning ex­am­ple is dis­cussed in The Plan­ning Fal­lacy. Pos­si­ble limi­ta­tions and prob­lems with us­ing the out­side view are dis­cussed in The Out­side View’s Do­main and “Out­side View” as Con­ver­sa­tion-Halter. Model Com­bi­na­tion and Ad­just­ment dis­cusses the im­pli­ca­tions of there usu­ally ex­ist­ing mul­ti­ple differ­ent out­side views.

I did 1500 piece jig­saw puz­zle of fire­works, my first jig­saw in at least ten years. Sev­eral times I had the strong im­pres­sion that I had care­fully elimi­nated ev­ery pos­si­ble place a piece could go, or ev­ery pos­si­ble piece that could go in a place. I was very tempted to con­clude that many pieces were miss­ing, or that the box had ex­tra pieces from an­other puz­zle. This wasn’t im­pos­si­ble – the puz­zle was an open box a rel­a­tive had done be­fore. And the al­ter­na­tive seemed hu­mil­i­at­ing.

But I al­lowed a very differ­ent part of my mind, us­ing differ­ent con­sid­er­a­tions, to over­rule this judg­ment; so many ex­tra or miss­ing pieces seemed un­likely. And in the end there was only one miss­ing and no ex­tra pieces. I re­call a similar ex­pe­rience when I was learn­ing to pro­gram. I would care­fully check my pro­gram and find no er­rors, and then when my pro­gram wouldn’t run I was tempted to sus­pect com­piler or hard­ware er­rors. Of course the prob­lem was al­most always my fault.

– Robin Han­son, Be­ware the In­side View

Fur­ther Examples

Th­ese illus­trate failure to use the Out­side View.

Ex­am­ple 1: Ja­panese stu­dents ex­pected to finish their es­says an av­er­age of 10 days be­fore dead­line. The av­er­age com­ple­tion time was ac­tu­ally 1 day be­fore dead­line. When asked when they’d com­pleted similar, pre­vi­ous tasks, the av­er­age re­ply was 1 day be­fore dead­line. (Buehler, R., Griffin, D., & Ross, M. 2002. In­side the plan­ning fal­lacy: The causes and con­se­quences of op­ti­mistic time pre­dic­tions. Heuris­tics and bi­ases: The psy­chol­ogy of in­tu­itive judg­ment, 250-270. Cam­bridge, UK: Cam­bridge Univer­sity Press.)

Ex­am­ple 2: Stu­dents in­structed to vi­su­al­ize how, where, and when they would perform their Christ­mas shop­ping, ex­pected to finish shop­ping more than a week be­fore Christ­mas. A con­trol group asked when they ex­pected their Christ­mas shop­ping to be finished, ex­pected it to be done 4 days be­fore Christ­mas. Both groups finished 3 days be­fore Christ­mas. (Buehler, R., Griffin, D. and Ross, M. 1995. It’s about time: Op­ti­mistic pre­dic­tions in work and love. Euro­pean Re­view of So­cial Psy­chol­ogy, Vol­ume 6, eds. W. Stroebe and M. Hew­stone. Chich­ester: John Wiley & Sons.)

It is con­tro­ver­sial how far the les­son of these ex­per­i­ments can be ex­tended. Robin Han­son ar­gues that this im­plies that, in fu­tur­ism, fore­casts should be made by try­ing to find a refer­ence class of similar cases, rather than by try­ing to vi­su­al­ize out­comes. Eliezer Yud­kowsky re­sponds that this leads to “refer­ence class ten­nis” wherein peo­ple feel that the same event ‘ob­vi­ously’ be­longs to two differ­ent refer­ence classes, and that the above ex­per­i­ments were performed in cases where the new ex­am­ple was highly similar to past ex­am­ples. I.e., this year’s Christ­mas shop­ping op­ti­mism and last year’s Christ­mas shop­ping op­ti­mism are much more similar to one an­other, than the in­ven­tion of the In­ter­net is to the in­ven­tion of agri­cul­ture. If some­one else then feels that the in­ven­tion of the In­ter­net is more like the cat­e­gory ‘re­cent com­mu­ni­ca­tions in­no­va­tions’ and should be fore­cast by refer­ence to tele­vi­sion in­stead of agri­cul­ture, both sides plead­ing the out­side view has no re­s­olu­tion ex­cept “I’m tak­ing my refer­ence class and go­ing home!”

Notable Posts

See Also

Model Com­bi­na­tion and Adjustment

lukeprog17 Jul 2013 20:31 UTC
75 points
41 comments5 min readLW link

Hero Licensing

Eliezer Yudkowsky21 Nov 2017 21:13 UTC
251 points
83 comments52 min readLW link

Con­fi­dence lev­els in­side and out­side an argument

Scott Alexander16 Dec 2010 3:06 UTC
165 points
190 comments6 min readLW link

Mul­ti­tudi­nous out­side views

Davidmanheim18 Aug 2020 6:21 UTC
47 points
13 comments3 min readLW link

What cog­ni­tive bi­ases feel like from the inside

chaosmage3 Jan 2020 14:24 UTC
201 points
23 comments4 min readLW link

Refer­ence class of the unclassreferenceable

taw8 Jan 2010 4:13 UTC
25 points
154 comments1 min readLW link

“Out­side View!” as Con­ver­sa­tion-Halter

Eliezer Yudkowsky24 Feb 2010 5:53 UTC
63 points
103 comments7 min readLW link

The Out­side View’s Domain

Eliezer Yudkowsky21 Jun 2008 3:17 UTC
17 points
14 comments10 min readLW link

Plan­ning Fallacy

Eliezer Yudkowsky17 Sep 2007 7:06 UTC
94 points
45 comments3 min readLW link

The Weak In­side View

Eliezer Yudkowsky18 Nov 2008 18:37 UTC
21 points
22 comments5 min readLW link

The Out­side View isn’t magic

Stuart_Armstrong27 Sep 2017 14:33 UTC
39 points
4 comments6 min readLW link

Cor­rigi­bil­ity as out­side view

TurnTrout8 May 2020 21:56 UTC
37 points
11 comments4 min readLW link

Out­side View(s) and MIRI’s FAI Endgame

Wei_Dai28 Aug 2013 23:27 UTC
16 points
60 comments2 min readLW link

Out­side View as the Main De­bi­as­ing Technique

abramdemski16 Oct 2017 21:53 UTC
16 points
4 comments2 min readLW link

In defense of the out­side view

cousin_it15 Jan 2010 11:01 UTC
14 points
29 comments2 min readLW link

Dis­cus­sion: weight­ing in­side view ver­sus out­side view on ex­tinc­tion events

Ilverin25 Feb 2016 5:18 UTC
3 points
4 comments1 min readLW link

Avoid mis­in­ter­pret­ing your emotions

Kaj_Sotala14 Feb 2012 23:51 UTC
71 points
32 comments7 min readLW link

Toward a New Tech­ni­cal Ex­pla­na­tion of Tech­ni­cal Explanation

abramdemski16 Feb 2018 0:44 UTC
144 points
36 comments18 min readLW link2 nominations1 review

The Prob­le­matic Third Per­son Perspective

abramdemski5 Oct 2017 20:44 UTC
48 points
3 comments4 min readLW link

Sus­pi­ciously bal­anced evidence

gjm12 Feb 2020 17:04 UTC
50 points
24 comments4 min readLW link

Nav­i­gat­ing dis­agree­ment: How to keep your eye on the ev­i­dence

AnnaSalamon24 Apr 2010 22:47 UTC
37 points
73 comments6 min readLW link

Com­ment on SSC’s Re­view of Inad­e­quate Equilibria

Ben Pace1 Dec 2017 11:46 UTC
30 points
5 comments2 min readLW link

Mis­takes with Con­ser­va­tion of Ex­pected Evidence

abramdemski8 Jun 2019 23:07 UTC
148 points
15 comments12 min readLW link

Time­less Modesty?

abramdemski24 Nov 2017 11:12 UTC
29 points
2 comments3 min readLW link

In­duc­tion; or, the rules and eti­quette of refer­ence class tennis

paulfchristiano3 Mar 2013 23:27 UTC
6 points
8 comments9 min readLW link

[Question] What is the right phrase for “the­o­ret­i­cal ev­i­dence”?

adamzerner1 Nov 2020 20:43 UTC
25 points
41 comments2 min readLW link

Gears Level & Policy Level

abramdemski24 Nov 2017 7:17 UTC
89 points
8 comments7 min readLW link

Against Modest Epistemology

Eliezer Yudkowsky14 Nov 2017 20:40 UTC
84 points
49 comments15 min readLW link

Plac­ing Your­self as an In­stance of a Class

abramdemski3 Oct 2017 19:10 UTC
51 points
5 comments3 min readLW link

What Makes My At­tempt Spe­cial?

Andy_McKenzie26 Sep 2010 6:55 UTC
35 points
22 comments2 min readLW link

[Question] What is a rea­son­able out­side view for the fate of so­cial move­ments?

jacobjacob4 Jan 2019 0:21 UTC
36 points
27 comments1 min readLW link

Kah­ne­man’s Plan­ning Anecdote

Eliezer Yudkowsky17 Sep 2007 16:39 UTC
27 points
8 comments2 min readLW link

Trust­ing Ex­pert Consensus

ChrisHallquist16 Oct 2013 20:22 UTC
28 points
81 comments17 min readLW link

How To Be More Con­fi­dent… That You’re Wrong

Wei_Dai22 May 2011 23:30 UTC
26 points
25 comments1 min readLW link

In­stru­men­tal Ra­tion­al­ity 2: Plan­ning 101

lifelonglearner6 Oct 2017 14:23 UTC
26 points
4 comments14 min readLW link

Are You Anosog­nosic?

Eliezer Yudkowsky19 Jul 2009 4:35 UTC
17 points
67 comments1 min readLW link
No comments.