My cat lies to me. I find this interesting.
My cat – our cat, rather – generally eats tinned food, but occasionally we give him cat biscuits. Not very often, and certainly not often enough as far as he’s concerned. He knows where they’re kept; when hungry will often sit in front of the biscuit cupboard giving it meaningful looks, even if he’s got a bowl full of food.
That’s not the interesting thing, though. What’s interesting is that, on several occasions, he’s sat by the back door and mewed to be let out, only to turn back and head for the biscuit cupboard when I open the door for him. The thinking is fairly straightforward, if you think of it as thinking – it goes roughly like this:
This‘ll get his attention!
But there’s an awful lot going on under the surface, particularly when you think that we’re dealing with a cat. How do you get to that thought? Or, if ascribing thoughts to a cat is a step too far, how do you get to that action? It seems to me that any creature capable of doing the back-door feint would have to go through something like this series of steps:
- Move (instinctively, or at any rate unreflectively) towards the back door when wanting to go out
- Move (unreflectively) towards the biscuit cupboard when fancying a biscuit or two
- Observe that move 1 is usually successful
- Observe that move 2 is usually unsuccessful
- Analyse events involved in successful outcomes to strategies 1 and 2
- Identify common factor, viz. getting a human’s attention
- Reflect on goals of move 1 and move 2
- Identify common intermediate goal of getting human’s attention
- Redefine move 1 as move which achieves intermediate goal
- Plan to make move 2 more effective by preceding it with move 1, thus getting human’s attention before expressing interest in biscuit cupboard
I don’t know about you, but that strikes me as pretty sophisticated thinking, particularly if we assume (as I think we must) that none of these thought processes are conscious.
Cats: they’re brighter than they look. Or rather, they really are as bright as they look.
6 Comments
Jolly clever animals. Hume was quite right about them.
When Andromache wants her breakfast, and thinks I am slow in getting out of bed to give it to her, she begins ripping up a nearby literary periodical. It never fails.
Fat Freddy’s cat is looking less ridiculous all the time.
Throwing stuff off the bedside cabinets is Ichy’s preferred approach.
first, check out Donald Norman’s “Action Lifecycle” for something that looks quite like your list of items.
Second, think about it this way. Could you design a _machine_ to do the same thing? IOW, a goal-seeking automaton that is programmed to achieve “reward” might record a success w/ the tinned food. It might also record a success w/ getting let outside.
The impressive part is the “leap” that i can duplicate some actions from successful outcomes in order to “fix” seemingly unsuccessful outcomes.
I agree – step 9 is a bit mindboggling when you think about it (redefining action-achieving-X as action-achieving-Y-which-leads-to-X, so that it can be reapplied as an action-achieving-Y to other Xs).
Hadn’t come across Norman, but I think his lifecycle falls down at precisely this point – reusing the action already specified for another purpose. I guess a program can’t produce its own subroutines.
yeah – #9 seems “magical”, but i continue to keep a sharp eye out for machine examples.
if you drop the “scripted” model and lean toward emergent results (neural-net type stuff), most of the magic falls away.
the simplest example is “mindlessly” (LOL) applying the actions from successful results to a new target repeatedly using minor variations; in small-world cases, this works! and my cats are quite patient when it comes to training our household humans into doing new things for them.
now the mystery is how the “new target” is selected. seems tabby has an “i want” wiring which can select new targets. machines? seems not.
but, who knows…