Projects‎ > ‎

Concept Learning

Concept Learning is a field of Cognitive Science that explores how concepts are attained in humans. One of the Concept Learning approaches is when a concept is represented by some rule that determines whether a stimulus belongs to a category. The learner tries to find a rule to explain the distinction between the categories and to categorize correctly when being confronted with a new stimulus. In this context, the learner observes a set of items and tries to distinguish the attributes of the set items and generate hypotheses indicating the rule of distinct categories. Then, by presenting a new stimulus, the learner tests the hypothesis by determining whether the rule can be applied on the new stimulus. 

We performed an empirical study with 606 students to investigate their Concept Learning through Machine Learning methods. Machine Learning methods, that measure the learning ability of algorithms, are applied to humans as if they were algorithms that learn from data. The methods of Rademacher Complexity and Algorithmic Stability were applied on data collected from Concept Learning tasks to analyze the hypothesis generation of learners. Our results can help improve the ability of educators to detect when learners are passively memorizing the concepts rather than discovering new knowledge. For more details, please have a look at our publications.