Epistemological Implications of a Reduction of Theoretical Implausibility to Cognitive Dissonance

The Aron­so­nian re-in­ter­pre­ta­tion of cog­ni­tive dis­so­nance as caused by ideas in con­flict with self-image fore­stalled some ob­vi­ous ap­pli­ca­tions to philo­soph­i­cal is­sues ly­ing at the bor­der with psy­chol­ogy. As the ac­tion-ori­ented ap­proach sug­gests, when Fest­inger’s the­ory is deep­ened to per­tain to the re­la­tions be­tween far-mode and near-mode rep­re­sen­ta­tions, the similar­ity be­tween cog­ni­tive dis­so­nance and the­o­ret­i­cal plau­si­bil­ity be­comes al­most ob­vi­ous. Im­plau­si­bil­ity has the same prop­er­ties and role as cog­ni­tive dis­so­nance. It is an aver­sive state that that can mo­ti­vate a change in far-mode be­liefs, and the change is to­ward a form of co­her­ence among be­liefs. Ri­val the­o­ries can be rated on a sin­gle di­men­sion of plau­si­bil­ity in the same way that they evoke differ­ent de­grees of dis­so­nance.

The re­duc­tion of im­plau­si­bil­ity to cog­ni­tive dis­so­nance bears sig­nifi­cant philo­soph­i­cal weight. It de­nies both Bayesian and co­her­en­tist the­o­ries of knowl­edge. The fash­ion­able Bayesian in­ter­pre­ta­tion of im­plau­si­bil­ity is in terms of de­grees of ra­tio­nal be­lief. A the­ory is im­plau­si­ble ac­cord­ing to the Bayesian School when it pos­sesses a low a pri­ori prob­a­bil­ity. But we don’t thereby scale our be­liefs for ra­tio­nal­ity if we scale be­liefs by how much dis­so­nance they cause,. More­over, to scale them by ra­tio­nal­ity would re­quire that we have some in­de­pen­dent rea­son for think­ing, apart from the com­par­a­tive cog­ni­tive dis­so­nance they arouse, that one the­ory is more ra­tio­nal than is the other. Scal­ing our be­liefs by the cog­ni­tive dis­so­nance they arouse can­not it­self be jus­tified on a pri­ori grounds, since dis­so­nance re­duc­tion of­ten takes us sys­tem­at­i­cally away from the truth, as in fact is the case in most ex­per­i­men­tal stud­ies of dis­so­nance. (This helps ex­plain why the iden­tity of dis­so­nance and im­plau­si­bil­ity hasn’t pre­vi­ously been noted.)

Re­gard­ing the other nega­tive im­pli­ca­tion of the dis­so­nance ac­count of im­plau­si­bil­ity—co­her­en­tist the­o­ries of knowl­edge haven’t ar­rived at a clear mean­ing of “co­her­ence,” but co­her­en­tist the­o­ries em­pha­size log­i­cal and ex­plana­tory re­la­tions among far-mode ideas (al­though re­cent ver­sions have in­cluded the role of ideas de­rived di­rectly from per­cep­tion). The dis­so­nance the­ory of im­plau­si­bil­ity holds that dis­so­nance is aroused by prag­matic in­com­pat­i­bil­ity be­tween near-mode and far-mode cog­ni­tion. How­ever, we don’t seem to be im­mune to con­flicts be­tween our far-mode ideas, al­though the ex­tent to which we are—the ex­tent of the im­mu­nity to far-mode hypocrisy—tends to sur­prise many of us. The re­s­olu­tion of this prob­lem is that the log­i­cal anal­y­sis of the re­la­tions be­tween far-mode ideas is it­self a near-mode ac­tivity. (Con­sider that the prac­tice of math­e­mat­ics is near-mode, as much as its con­tent is ab­stract.) We are sen­si­tive to in­con­sis­ten­cies in far-mode ideas only to the ex­tent that we draw upon them in our an­a­lyt­i­cal prac­tices—and to the ex­tent that our own ac­tivity in­volves such prac­tices. Those in­volved in ex­pound­ing a doc­trine or act­ing in its terms will be sub­ject to dis­so­nance to the ex­tent that far-mode ideas prag­mat­i­cally con­flict with the perfor­mance of their an­a­lyt­i­cal perfor­mances.

The dis­so­nance the­ory of plau­si­bil­ity also bears no the mys­tery of the con­jur­ing up of the­o­ret­i­cal terms. We know that sci­en­tific the­o­ries go be­yond the em­piri­cal ev­i­dence, as in prin­ci­ple there are in­finitely many the­o­ries con­sis­tent with any set of em­piri­cal facts. On the dis­so­nance ac­count of plau­si­bil­ity, the­ory cre­ation and ac­cep­tance is driven by dis­so­nance re­duc­tion. Far-more the­o­ries pro­mote sci­en­tific prac­tices by en­er­giz­ing them. They do this by pro­vid­ing the frame­work in which sci­en­tists work. If work is to be sys­tem­atic, a frame­work is nec­es­sary, but are the declar­a­tive propo­si­tions the frame­work ex­presses true? Do they have a prob­a­bil­ity of truth?

Scien­tific Real­ist philoso­phers of sci­ence have ar­gued con­vinc­ingly that the­o­ret­i­cal propo­si­tions in sci­ence of­ten pur­port to be true, but no­body has come close to pro­vid­ing an ac­count of what it means for an ab­stract the­ory to be prob­a­ble, such that we can in­quire re­gard­ing the epistemic prob­a­bil­ity that New­to­nian physics was true? The no­tion that we have ra­tio­nal de­grees of be­lief in the­o­ries does ac­cord with some in­tu­itions. Plau­si­bil­ity must al­low at least or­di­nal rank­ing, since dis­so­nance in­volves choice be­tween differ­ent cog­ni­tive states ac­cord­ing to their plau­si­bil­ity. This in turn means that the laws of prob­a­bil­ity ap­ply to or­di­nal re­la­tions. For ex­am­ple, the plau­si­bil­ity of The­ory A and The­ory B will never be greater than the plau­si­bil­ity of The­ory A. But let me sug­gest that even this is a product of dis­so­nance as shaped by the­o­ret­i­cal de­vel­op­ment, as is shown in stud­ies show­ing that in many situ­a­tions we em­piri­cally find the con­junc­tion fal­lacy com­pel­ling—that is, plau­si­ble.


Some­times the search for dis­so­nance re­duc­tion leads to truth, and some­times it leads away from truth. Ra­tion­al­ity is a limit­ing case of dis­so­nance re­duc­tion, but it’s one im­pos­si­ble to spec­ify ex­cept from within a psy­che sub­ject to the laws of cog­ni­tive dis­so­nance. Then this prob­lem: how do we even ex­press the ex­pec­ta­tion that sci­en­tific the­o­ries get closer to the truth and re­li­gious the­o­ries do not? We can say that sci­en­tific the­o­ries de­pend on ex­per­i­men­tal and ob­ser­va­tional prac­tices and there­fore have at least the pos­si­bil­ity of rest­ing on ac­tual ev­i­dence. We can say sci­en­tific the­o­ries have greater plau­si­bil­ity than re­li­gious the­o­ries, these both be­ing judg­ments that are a product of the law of dis­so­nance. But, counter-in­tu­itively (at least for me), we can’t say that sci­en­tific the­o­ries are more prob­a­ble than re­li­gious the­o­ries. It isn’t, it’s im­por­tant to no­tice, that we don’t know which is more prob­a­ble. Rather, the whole no­tion of prob­a­bil­ity as ap­plied to the­o­ries is mis­be­got­ten. That a the­ory is im­plau­si­ble or plau­si­ble is a far-mode con­clu­sion, and far-mode doesn’t deal in the rel­a­tive fre­quen­cies mod­eled by the prob­a­bil­ity calcu­lus.

A sim­ple ex­am­ple might be clar­ify­ing in show­ing the limits of the con­cept of prob­a­bil­ity and its close­ness to near-mode ex­pe­rience. Dur­ing the last pres­i­den­tial elec­tion, pol­lsters ar­rived at es­ti­mates of the prob­a­bil­ity of win­ning for each can­di­date. Pol­lsters use mainly near-mode rea­son­ing to en­g­ineer the best pre­dic­tive for­mu­las. The pol­lsters sub­stan­tially agreed, as you would ex­pect when they each ap­plied similar method­olo­gies, all based on sim­ple ex­trap­o­la­tion of the near-mode pro­cess of sam­pling and gen­er­al­iz­ing to a defined pop­u­la­tion.

The ac­cu­racy of these con­clu­sions, how­ever, de­pended on cer­tain far-mode as­sump­tions, such as that peo­ple tak­ing polls re­spond hon­estly. What if this as­sump­tion didn’t hold? Well, it didn’t; Trump won and the main rea­son the polls got it wrong was (or might have been, if you pre­fer) that vot­ers pol­led weren’t hon­est about their prefer­ences. We might ask, what should have been the true prob­a­bil­ity es­ti­mate, given that the pol­lsters didn’t take into ac­count the prob­a­bil­ity that their model was based on false as­sump­tions. How should they have taken this into ac­count? Prob­a­bil­ity es­ti­mates re­sult from the near-mode op­er­a­tion of fit­ting ob­ser­va­tion to a rel­a­tive fre­quency model. We can com­pli­cate the model to take ac­count of more in­for­ma­tion, but what we can’t do is ad­just the prob­a­bil­ity es­ti­mate to take ac­count of the model’s own fal­li­bil­ity. (At a deeper level, Bayesian es­ti­mates can’t ad­just for the prob­a­bil­ity that the Bayesian method­ol­ogy it­self is in­valid—as I con­tend it is in fact.) If it makes sense to as­sign a prob­a­bil­ity to a the­ory be­ing true, how much be­lief should be ac­corded in some ideal­ized ra­tio­nal world, then it should be pos­si­ble to ap­prox­i­mate that prob­a­bil­ity. Some­one can ad­just it “in­tu­itively,” but my point is that there is noth­ing ap­pro­pri­ate for an in­tu­ition to be about. The­o­ret­i­cal plau­si­bil­ity is not prob­a­bil­ity.

At this point arises a skep­ti­cal temp­ta­tion, for not only is our knowl­edge not ab­solute, it isn’t even prob­a­ble. Plau­si­bil­ity can sys­tem­at­i­cally take us away from knowl­edge. We seem to long for a ra­tio­nale for do­ing the ra­tio­nal thing, and such a ra­tio­nale is sup­plied when knowl­edge and ra­tio­nal con­duct is rep­re­sented by Bayesian and de­ci­sion-the­o­retic for­mu­las. We see our­selves as free and availed of (men­tally) un­limited choice. We are ra­tio­nal be­cause we choose to be, and that en­tails that the choice it­self be ra­tio­nal. But our pos­si­bil­ities aren’t un­limited, freely cho­sen. Our ideas will evolve in ac­cor­dance with the de­mands of cog­ni­tive dis­so­nance or else moved by re­cep­tivity to sug­ges­tion. There’s no “free will” to seize the ini­ti­a­tive, and no di­rect ac­cess to ra­tio­nal­ity to guide us.