Cognitive development in Zurich
The second members’ meeting of the euCognition II research network took place in Zurich on Friday. The thesis for discussion was:
“Cognition emerges during development in a close interplay of experience, of the social and physical environment and of the neuronal mechanisms of growth. An understanding of cognition cannot be achieved without an understanding of the development of cognition. It is thus a necessity for artificial cognitive systems to take development on board.”
Four plenary talks set up the issues and four parallel workshops followed up on each. As is to be expected, there wasn’t much by way of conclusive resolution, though amongst the engineers and roboticists there remained a tendency to believe that while development might be interesting, we’re probably better off just trying to get the specifications right and building to them, rather than trying to grow cognitive artificial systems. At present, it seems a fair point, as no one at the meeting seemed to be able to come up with a single example of where a developmental process has produced a better model or robot than a more traditionally engineered solution.
That said, there wasn’t much agreement on whether traditionally engineered robots were doing a very good job of being cognitive either, but at least progress is being made.
I often find that conferences or meetings like this one are made worth the effort not for the masses of new information or avalanche of relevant research that comes from them, but by one or two things said that act as a stark reminder or a sudden wake up to an ill-seen problem or way of thinking. So it was with the euCog2 meeting.
Linda Smith gave a talk on toddlers’ visual attention and learning which outlined some lovely work examining the role of hands in vision. In relatively natural table-top play with a parent, children things children pay attention to tend to be grabbed and handled. Their short arms tend to mean that handling something will bring it close to the face and also block out other objects that might distract them. Smith suggested that the hands effectively help stabilise the child’s otherwise very dynamic (read: distractable) attention in a way that supports learning about the objects.
One of the points that really struck me though, and is most relevant to the enactive approach, was something she pointed out in the discussion afterward – babies spend the first four months or so of their lives just sitting there. Reaching things doesn’t really start until the fifth month. Now, there are no doubt all kinds of other motor interactions with the world that matter, but it’s an interesting challenge for (rather than to) dynamic sensorimotor theories. This passivity seems to have been selected for, evolutionarily – if so, what’s it doing?
I also thought David Vernon‘s talk a good one. A genuinely enactive roboticist, I had heard of him (he ran the first incarnation of the euCognition network), but hadn’t actually read his stuff before. He was instrumental in the development of the iCub robot, a robotics platform specifically designed to study cognitive development. In particular, some of the issues he is focused on are those that are normally thrown at the enactive approach as counter-examples – things like imagination, anticipation and hypothetical thinking (counterfactual thinking in general, I suppose). Looking forward to hunting out some of his papers and reading a bit more.