Stafford, who's a professor, as well as an author and scientist, thought he'd designed the perfect experiment for his students: run the little robots through a series of simple learning experiments and chart how quickly their learning occurred. Stafford assumed that with such simple "subjects" he'd be able to plot a beautiful curve on a graph to demonstrate how you can predict the pace of learning.
Stafford had to eat some humble pie, however, which he acknowledges in good humor on his blog. It turns out that the learning of the robots was so haphazard and all over the place that the results were a messy jagged scribble on the graph and not the elegant curve that Stafford predicted. The relevance here is that if things come out in an unexpected manner with such simple robots, how much more difficult must it be to predict the behavior of humans, whose brains and behavior are infinitely more complex. Stafford concludes:
"Trying a simple experiment with the Lego robots gave me a new respect for the experimental method, and the difficulty psychologists face when trying to discover the rules underlying the wondrous variety in human behavior."
(Photo Source: the original blog host from MindHacks)