Motivated Stopping and Motivated Continuation

While I dis­agree with some views of the Fast and Fru­gal crowd—in my opinion they make a few too many lemons into lemon­ade—it also seems to me that they tend to de­velop the most psy­cholog­i­cally re­al­is­tic mod­els of any school of de­ci­sion the­ory. Most ex­per­i­ments pre­sent the sub­jects with op­tions, and the sub­ject chooses an op­tion, and that’s the ex­per­i­men­tal re­sult. The fru­gal­ists re­al­ized that in real life, you have to gen­er­ate your op­tions, and they stud­ied how sub­jects did that.

Like­wise, al­though many ex­per­i­ments pre­sent ev­i­dence on a silver plat­ter, in real life you have to gather ev­i­dence, which may be costly, and at some point de­cide that you have enough ev­i­dence to stop and choose. When you’re buy­ing a house, you don’t get ex­actly ten houses to choose from, and you aren’t led on a guided tour of all of them be­fore you’re al­lowed to de­cide any­thing. You look at one house, and an­other, and com­pare them to each other; you ad­just your as­pira­tions—re­con­sider how much you re­ally need to be close to your work­place and how much you’re re­ally will­ing to pay; you de­cide which house to look at next; and at some point you de­cide that you’ve seen enough houses, and choose.

Gilovich’s dis­tinc­tion be­tween mo­ti­vated skep­ti­cism and mo­ti­vated cre­dulity high­lights how con­clu­sions a per­son does not want to be­lieve are held to a higher stan­dard than con­clu­sions a per­son wants to be­lieve. A mo­ti­vated skep­tic asks if the ev­i­dence com­pels them to ac­cept the con­clu­sion; a mo­ti­vated cre­dulist asks if the ev­i­dence al­lows them to ac­cept the con­clu­sion.

I sug­gest that an analo­gous bias in psy­cholog­i­cally re­al­is­tic search is mo­ti­vated stop­ping and mo­ti­vated con­tinu­a­tion: when we have a hid­den mo­tive for choos­ing the “best” cur­rent op­tion, we have a hid­den mo­tive to stop, and choose, and re­ject con­sid­er­a­tion of any more op­tions. When we have a hid­den mo­tive to re­ject the cur­rent best op­tion, we have a hid­den mo­tive to sus­pend judg­ment pend­ing ad­di­tional ev­i­dence, to gen­er­ate more op­tions—to find some­thing, any­thing, to do in­stead of com­ing to a con­clu­sion.

A ma­jor his­tor­i­cal scan­dal in statis­tics was R. A. Fisher, an em­i­nent founder of the field, in­sist­ing that no causal link had been es­tab­lished be­tween smok­ing and lung can­cer. “Cor­re­la­tion is not cau­sa­tion,” he tes­tified to Congress. Per­haps smok­ers had a gene which both pre­dis­posed them to smoke and pre­dis­posed them to lung can­cer.

Or maybe Fisher’s be­ing em­ployed as a con­sul­tant for to­bacco firms gave him a hid­den mo­tive to de­cide that the ev­i­dence already gath­ered was in­suffi­cient to come to a con­clu­sion, and it was bet­ter to keep look­ing. Fisher was also a smoker him­self, and died of colon can­cer in 1962.1

Like many other forms of mo­ti­vated skep­ti­cism, mo­ti­vated con­tinu­a­tion can try to dis­guise it­self as vir­tu­ous ra­tio­nal­ity. Who can ar­gue against gath­er­ing more ev­i­dence?2

I can. Ev­i­dence is of­ten costly, and worse, slow, and there is cer­tainly noth­ing vir­tu­ous about re­fus­ing to in­te­grate the ev­i­dence you already have. You can always change your mind later.3

As for mo­ti­vated stop­ping, it ap­pears in ev­ery place a third al­ter­na­tive is feared, and wher­ever you have an ar­gu­ment whose ob­vi­ous coun­ter­ar­gu­ment you would rather not see, and in other places as well. It ap­pears when you pur­sue a course of ac­tion that makes you feel good just for act­ing, and so you’d rather not in­ves­ti­gate how well your plan re­ally worked, for fear of de­stroy­ing the warm glow of moral satis­fac­tion you paid good money to pur­chase.4 It ap­pears wher­ever your be­liefs and an­ti­ci­pa­tions get out of sync, so you have a rea­son to fear any new ev­i­dence gath­ered.5

The moral is that the de­ci­sion to ter­mi­nate a search pro­ce­dure (tem­porar­ily or per­ma­nently) is, like the search pro­ce­dure it­self, sub­ject to bias and hid­den mo­tives. You should sus­pect mo­ti­vated stop­ping when you close off search, af­ter com­ing to a com­fortable con­clu­sion, and yet there’s a lot of fast cheap ev­i­dence you haven’t gath­ered yet—there are web­sites you could visit, there are counter-counter ar­gu­ments you could con­sider, or you haven’t closed your eyes for five min­utes by the clock try­ing to think of a bet­ter op­tion. You should sus­pect mo­ti­vated con­tinu­a­tion when some ev­i­dence is lean­ing in a way you don’t like, but you de­cide that more ev­i­dence is needed—ex­pen­sive ev­i­dence that you know you can’t gather any­time soon, as op­posed to some­thing you’re go­ing to look up on Google in thirty min­utes—be­fore you’ll have to do any­thing un­com­fortable.

1Ad hominem note: Fisher was a fre­quen­tist. Bayesi­ans are more rea­son­able about in­fer­ring prob­a­ble causal­ity; see Judea Pearl’s Causal­ity: Models, Rea­son­ing, and In­fer­ence.

2Com­pare Robin Han­son, “Cut Medicine In Half,” Over­com­ing Bias (blog), Septem­ber 10, 2007, http://​​www.over­com­ing­bias.com/​​2007/​​09/​​cut-medicine-in.html.

3Ap­par­ent con­tra­dic­tion re­solved as fol­lows: Spend­ing one hour dis­cussing the prob­lem, with your mind care­fully cleared of all con­clu­sions, is differ­ent from wait­ing ten years on an­other $20 mil­lion study.

4See “‘Can’t Say No’ Spend­ing.” http://​​less­wrong.com/​​lw/​​kb/​​cant_say_no_spend­ing.

5See “Belief in Belief” in Map and Ter­ri­tory.