The non-Independence of superficially-Irrelevant Alternatives

post by roystgnr · 2014-01-15T18:46:29.053Z · LW · GW · Legacy · 3 comments

Contents

3 comments

A new article in Biology Letters shows that under some conditions in which animals appear to behave "irrationally" (by apparently failing to conform to the Independence of Irrelevant Alternatives or even the Transitivity axioms of decision theory), the animal behavior may in fact resemble utility-maximizing strategies which also appear to violate those axioms.  The optimal strategies' current preferences are altered in response to the information conveyed by the current presence or absence of various alternatives.

A press release about the article is available from the lead author's university, U. Bristol. A news piece summarizing it is at Nature's website.

These results probably shouldn't surprise the most careful rational thinkers. For instance, "I prefer A to B when I believe more B is likely to be available later, and B to C when I believe more B is likely to be available later" clearly does not necessarily entail "I prefer A to B under all circumstances, and B to C under all circumstances". But there is clearly a pitfall of tempting model oversimplification here. Although this new paper and most discussion of it is carefully putting scare quotes around "irrational", the papers being cited purport to show "intransitive preferences", "violations of rational choice", "irrational decision-making", and "irrational choices", without obvious irony.

3 comments

Comments sorted by top scores.

comment by ThisSpaceAvailable · 2014-01-21T01:20:35.399Z · LW(p) · GW(p)

I'm having trouble parsing the first paragraph. Behavior can't resemble a strategy; those are separate categories. The first paragraph seems to be saying "When animals violate certain axioms, their behavior is consistent with strategies that violate those axioms", which is so trivial I can't see the point of saying it. Or is the claim that the strategies that the behavior is consistent with, while appearing to violate the axioms, does in fact optimize for utility? Can you explain this more explicitly?

Replies from: roystgnr
comment by roystgnr · 2014-01-21T14:54:31.297Z · LW(p) · GW(p)

Does "animals of species S typically choose food A over food B in context X" fall into your "behavior" category or into your "strategy" category? It falls (loosely) into both for me: it is a behavior exhibited by the animals, but a priori one would expect a good possibility that this behavior is one part of an evolutionarily tuned strategy for survival.

Your final interpretation of the claim here was the correct one. It's not just that this set of apparently-irrational animal behaviors is part of the large class "apparently-irrational strategies", right next to "stare motionless at oncoming headlights"; it seems to be part of the narrower class "apparently-irrational strategies which turn out to be rational upon closer examination".

Sorry for my confusing wording.

comment by astrologyloveback · 2014-01-16T12:24:59.694Z · LW(p) · GW(p)

nice touch