26 June 2017

Inhuman instruction - When learners aren’t human

One criticism of learning methods rooted in behaviorism is the related methods don't adequately account for human psychology. Notable researcher of operant conditioning and reinforcement learning B.F. Skinner believed that since we can’t perfectly explain human psychology and motivation, then it's pointless to account for it in the design of instruction. But today's learners are different- some of them are computers. Programmers use machine learning methods to write computer programs. Humans program machines to learn. And when learners are no longer human, design constraints suggested by human capabilities no longer apply to the design of instruction. Computers don’t forget and they don’t have attitudes (e.g. motivation, self-efficacy, mindset.) Machines tirelessly work out problems without complaining, forgetting, or working through confidence issues. As a result, many of the ideas of behaviorism and learning are newly applicable to a new generation of learners: the machines. Case in point, check out the new Technology Review list of top technologies and the article about reinforcement learning. Put that algorithm in a skinner box, stand back, and see what it learns.