Image credit: Ractapopulous (CC0)
Many things in the world have a certain structure to them, which we can use to organize our thinking. To mentally represent your family, for example, you could group your family members into men and women, or group them based on where they live. But a more intuitive approach for most people is to organize family members by generation: child, sibling, parent, grandparent. It is as though we instinctively place each family member along a mental line, from young to old.
We use mental lines to organize other types of information too, most notably numbers. But can we also use them to represent new information? Luyckx et al. trained healthy volunteers to associate pictures of six different colored donkeys with six different reward probabilities. One donkey was followed by reward 5% of the time, another was followed by reward 95% of the time, and so on. Through trial and error, the volunteers learned to rank the donkeys in terms of how likely they were to precede a reward. Luyckx et al. then compared the volunteers’ brain activity while viewing the donkeys to their brain activity while viewing the numbers 1 to 6.
The donkeys evoked patterns of electrical brain activity corresponding to the number that signaled their place on a mental line. Thus, donkey 1, with the lowest reward probability, produced a pattern of brain activity similar to that of the number 1, and so on for the others. This suggests that rather than learning in an unstructured way, we use past knowledge of relations among stimuli to organize new information. This phenomenon is called structural alignment.
The results of Luyckx et al. provide the first evidence from brain activity to support structural alignment. They suggest that we use a general understanding of how the world is structured to learn new things. This could be relevant to both education and artificial intelligence. People, and computers, may learn more effectively if taught about the relations between items, rather than just the items in isolation.
