The idea of 'learning' is central to our understanding of advanced cognition. But what does it entail?
At the very least, learning involves keeping track of events. Beyond this, it involves finding regularities among those events that repeat over time and across situations. In a fuller sense, learning involves the discovery of statistical redundancies that may be the result of some governing causal processes in the environment. If the environment were completely random, there would only be each moment to 'learn', perhaps never to be repeated. There would be no basis for prediction, mastery, or control, all functions that are critical for behaving, surviving organisms.
The anthropologist and cyberneticist Gregory Bateson developed a framework for analyzing the structure of learning, involving a hierarchy of levels.
Bateson identified a perhaps trivial form of learning, termed 'zero-order' learning, equivalent to sensing or transduction. I see the traffic light turn from red to green, thus I have 'learned' about its new state.
'Protolearning' can be regarded as simple association. I learn that when I see green, I go, and when I see red, I stop. Planaria flatworms show this kind of conditioning in water mazes--if you poke them when they turn right, they will learn after a while to turn left most of the time.
'Deuterolearning' is a learning of context. If you reverse the association, how long does it take for the organism to adapt? Planaria show the same slow learning curve regardless of how many times you switch your poking to reinforce left-turning versus right-turning. Canines, however, get the jist of reversals and speed up their learning curves each time. They learn that reversal is a context.
Dolphins have shown something more extraordinary. If each day, a new 'trick' behavior is recruited and reinforced, after several such days, they will learn to show something new and reward-worthy at the beginning of a session. They understand what 'new' means in a deeply contextual sense.
How deep can humans go?
The author's affiliation with The MITRE Corporation is provided for identification purposes only, and is not intended to convey or imply MITRE's concurrence with, or support for, the positions, opinions or viewpoints expressed by the author.