Reinforcement Learning for Machine and Human Performance

Playtime's Over 
Technology Review
Emma Brunskill

" ... Reinforcement-learning techniques that enabled a computer to beat world-class human Go champions last year should be applied to much higher ambitions, according to Stanford University professor Emma Brunskill. She says "data smart" algorithms are an essential ingredient in the creation of artificial tutors via reinforcement learning. "In some cases there's not enough data, or not the right kind of data, which makes it challenging to develop systems that make good decisions," Brunskill notes. 

She says her team at Stanford is developing reinforcement-learning algorithms and statistical methods so computers can devise good suggestions while using less data. Brunskill also stresses humans should participate in reinforcement learning to enable algorithms to "reason" about their own performance and consult with humans for guidance and aid with conducting complex tasks. "Such human-computer collaborations could help students to learn using approaches we can't yet imagine," she says    ... "   Full article in Technology Review.