Pose this question to a person and you’ll learn a lot about them, I’d wager. It’s quite a simple question that I have, more than once, set out to answer and convince others of my answer: yes, they are. To do this the first time, I asked a close yet sceptical friend what they would accept as evidence of a “rational” actor. The discussion was quite long but my friend eventually settled on an answer I will paraphrase:
“I will accept a dog is a rational actor if it can be shown to be enacting Game Theory”
That’s almost tautologically true, as Game Theory concerns itself with strategies employed by rational actors. And so I explained what experiment I was going to set up to show a dog, in fact, can make decisions in accordance with Game Theory. I have to admit to two things: this is not a quantitative comparison of mathematical models derived by Game Theory against behaviours shown by a dog, it is a qualitative analysis of whether it appears a dog is following basic Game Theory behaviours as we understand them; we’re using the term ‘rational’ at a lower benchmark than ‘can ponder evidence and reach defensible claims’.
However, I told my friend I was going to devise an experiment to test the claim that dogs are rational actors. The experiment is this: over the next week I will place post-meal plates, bowls and saucepans on the floor as an offer to my labrador and then take them away randomly. Each item has different amounts of food on them. This will teach the condition that food is on offer for an undefined length of time, which is fundamentally different to how Noodle, this dog, receives other meals. Noodle is motivated by food, so if Noodle is a rational actor she will devise a method or strategy to optimise her food intake.
After I explained this experiment to my friend, I asked them what they would accept as evidence of Game Theory behaviour. My answer, which ultimately formed by hypothesis, is that:
H1 Noodle will switch between food-covered items, before their supply is exhausted, to maximise food flow into her face.
H0 Noodle will not switch between food-covered items, before their supply is exhausted, to maximise food flow into her face.
He discussed that briefly and agreed that transitioning between items takes time, so the “optimal” time to switch is not necessarily when the food supply on that plate is least. But we agreed:
- Modifying behaviour to optimise gains is ‘Game Theory’ behaviour.
- Demonstrating ‘Game Theory’ behaviour is ‘rational’
- Therefore, actors who modify behaviour to maximise gains are rational actors.
So now it was time to see whether Noodle is a rational actor. My experience with my partner’s dog, Lana, is that dogs are rational actors. I know anecdotal evidence is prone to post facto cognitive errors and a host of other problems, but Lana is my inspiration for this experiment. If nothing else, I am exploring whether Lana’s behaviour is repeatable in other labradors.
This is, in fact, what we saw. Noodle initially went for the wok: the item with the most food. But Noodle moved on to the saucepan with peas left in it well before the wok was finished. Then a bowl with large amounts of food left, then back to the wok, then to cleaning plates. The items were then removed.
I can now reject my null hypothesis and consider this support for my claim that dogs are rational actors. But my friend disagreed:
Yeah, but that’s not really “rational” behaviour, is it? It’s evolved.
Which just goes to show you can always offer an unreasonable objection. Even if you have agreed all the terms ahead of time.