Childrens' thought processes could inspire better computers
By Ben Coxworth
March 16, 2012
Children are sometimes referred to as “sponges,” not because they live off our earnings, but because of their remarkable ability to learn things quickly. Psychologists believe this is because their brains are still wired for learning and exploration – essential qualities for building neural connections – whereas adult minds tend to focus on specific goals, at the expense of imagination and curiosity. Now, scientists from the University of California, Berkeley are studying the cognitive functions of babies, toddlers and preschoolers, in hopes of using their findings to make computers think more like humans.
Through a number of experiments, the research team has discovered that children are quite adept at testing hypotheses, detecting statistical patterns, and drawing conclusions while at the same time adapting to changes. The researchers are particularly interested in children’s understandings of causal relationships, or in other words, how they determine that one thing is caused by another.
As an example, in one experiment, preschoolers got to sing Happy Birthday to a stuffed monkey, whenever it was brought out and a music player was switched on. When the music player was removed, however, the children responded by replacing it with a wooden block, so the singing could continue.
In another experiment, children were shown a toy that lit up and spun around, and were told that red blocks made it light up, green blocks made it spin, and blue ones made it do both. When the children realized that the red and green blocks were much more plentiful than the blue ones, however, they decided that it was in fact a combination of red and green that made the toy do both, and that the blue blocks were duds – a hypothesis that turned out to be accurate.
“Children went with simplicity when there wasn’t strong evidence for an alternative, but as evidence accumulated, they followed its lead,” explained UC Berkeley psychologist Tania Lombrozo.
She and her colleagues hope that their findings could ultimately help computers to consider new possibilities for cause-and-effect relationships, in light of changing odds. This ability could conceivably do everything from allowing them to interact more easily with humans, to identifying which genes cause greater susceptibility to medical problems.
More information is available in the video below.