Abstract:

Over the last three years, the deep learning version of machine learning has made more inroads into learning systems than all the artificial intelligence ever developed, particularly to understand image, voice and language processing, etc. The defining characteristic and strength of 'supervised' machine learning systems is their ability to 'see' patterns that humans cannot because they are hidden in massive amounts of data (with little signal and lots of noise). This paper presents a range of insights and their implications for commercial or public policy that are directly drawn from a process of confronting machine generated patterns (from classification trees as a frequent form of supervised machine learning and Bayesian networks) with behavioural science principles from cognitive and social psychology, decision neuroscience and bioenergetics/neurophysiology.