Gymnasion

Gymnasion reimagines rhetorical education by using machine learning to train and test humans. In the first mini-game, “Speed of Breath,” the user motivates a character to run by sending him inspirational messages. These are assessed by a statistical model trained to distinguish inspiring from non-inspiring quotations. The more inspiring and rapid-fire the messages, the faster the character sprints. “Manual Style Transfer” tasks the user with rewriting a sentence so that it sounds as if it were written by Whitman (according to a model trained for authorship identification) while retaining its meaning (according to Word Mover’s Distance, a metric of semantic similarity [Kusner et al., 2015]). A data visualization guides the user’s progress.