For AI, the “training” technique, in part, seeks to expand the knowledge base artificial intelligence to produce more accurate results in machine learning decision making. A recent heuristics study in humans, however, showed that adolescents arrived at more accurate conclusions due to a larger knowledge base as well as more experience on how to use it.
Rules of Thumb and Heuristics
Heuristics are fundamental in both human pychology and computer programming. Subsequently, experience is the basis of decision-making but, there are different types of experience; Doing helps us acquire direct, “hands-on” experience, while observing or studying are forms of more indirect experience.
Rules of thumb are arguably another form of indirect experience. One example is the recognition heuristic, which is a powerful yet relatively simple way of making decisions quickly.
For instance, if you were asked if San Antonio were bigger than Houston, and you had never heard of San Antonio, how would you answer the question? It might be easier to deduce that the city you know is probably bigger than the one you’ve never heard of, giving you at least some indication of how to more answer the question.
It is important to note that rules of thumb are not a guarantee but they are a guide. There is a fine line between rational decision making like rules of thumb and formulating biases and stereotypes.
Therefore, how children acquire these rules of thumb and how they use them can have important indications for fine tuning machine learning techniques for more accurate (or desirable) results.
While rule of thumb decision making has been widely studied in adults, the same cannot be said for children.
The question, therefore, is whether children and adolescents use the recognition heuristic and if so, how? In a recent study, Researchers at the Max Planck Institute for Human Development in Berlin and the University of California, Berkeley sought to answer that question.
Using What you Know
Researchers in the study used a sample of over one hundred Italian schoolchildren aged nine, twelve, and seventeen, respectively.
Next, the children were asked about two diseases and two cities. The children were required to judge which of the two cities had a larger population, and which of the two diseases occurred more frequently in their country.
The final question for the children was to say which of the two diseases and which of the cities they had heard of before.
Adolescents quickly understood that the recognition heuristic alone is not as useful for estimating the occurrence of diseases as it is in determining the size of a city.
Better With age
Children between eight to twelve years old were using the recognition heuristic, but adolescents were adapting the system into strategy by using it in some situations and abandoning it in others.
The younger children in the sample made more mistakes because they were equating the number of times they had heard of a disease to its prevalence.
Adolescents, however, quickly understood that the recognition heuristic is not as useful for estimating the occurrence of diseases as it is in determining the size of a city.
Therefore, the older children were able to break down their thought process into a staged approach that refines information- an important indication for machine learning that access to vast amounts information is equally important as knowing refine its use.