Like me, ants like dark houses/nests with small entrances. Facing a choice between a dark nest with a large entrance (option A) and a light nest with a small entrance (option B), an ant colony faces a trade-off. Some go this way to A and some go that way to B. Suppose we add a third decoy nest option D. Option D is as dark as A but has an even larger entrance. It is thus dominated by A but not by B. How will the ant colony’s behavior change when they face the three options together versus just A and B?
Rational choice theory says that the fractions choosing A and B should not change. Option D is dominated and should never chosen and hence is an irrelevant alternative. Its presence or absence should not affect the choice between A and B.
One psychological theory suggests that the proportion choosing A should go up. Option D helps to crystallize the advantages of option A (the smaller entrance). This may increase the perception of the advantages of A over B as well leading to a change in the proportion of ants choosing A over B.
So what actually happens?
A controlled experiment by Edwards and Pratt answers this question. Edwards and Pratt built nests with the properties above and made ant colonies make repeated binary and ternary choices. They randomized the order of choices, where the nests were located etc. And because they were experimenting with ants, they could cruelly force the choice of nest upon the ants by destroying the old nest the ants lived in by removing it’s roof.
They find no significant change in the proportions choosing A vs B when the decoy D is present. Ant colonies are rational and do not violate the axiom of independence of irrelevant alternatives (IIA).
In other work, Pratt shows that ant colonies obey transitivity (i.e. if a colony prefers A to B and B to C, it prefers A to C).
Why are ant colonies more rational than individual humans? The authors offer a cool hypothesis: choice between colonies is typically made by sending independent scouts sent to the different options. No scout visits different locations. The scouts reports are simply compared and the best option is chosen. A human being contemplates all the choices by herself and has a harder time comparing the attributes independently leading to a violation of IIA.
An ant colony is like a well performing and coordinated decentralized firm with employees passing information up the hierarchy and efficient decisions coming down from the center Can we import lessons into designing firms? Alas, I believe not. A human scout evaluating a decision/option will not be as impartial an ant scout. He will exagerrate its qualities, hoping his option “wins”. He hopes to get the credit for finding the implemented option, get promoted, receive stock options and retire young to the Bay Area. In other words, career concerns ruin a simple transfer of ant colony principles to firms. If we eliminate career concerns within the firm, we will induce moral hazard as there is no incentive to exert costly effort to find the best decisions for the firm. Ants in the same colony do not face the same issue as they are genetically related and have “common values”.
Still, a thought-provoking paper and it has many references to other papers that it builds on. I am going to read more of them.
(Hat tip to Christophe Chamley for the reference)