Huygens Colloqium on Neuromorphic Interaction Design

HG00.304 at 12:30

As part of the Radboud Huygens Colloqia series, Roel Vertegaal will present a new theory for the design of interactive systems based on cognitive neuroscience: Neuromorphic Interaction Design. Neuromorphic Interaction Design elegantly explains how users process tasks by first predicting the world around them, then choosing to learn or act when their predictions do not correspond with their observations. Roel will discuss how this yields elegant laws that govern user performance and error by quantifying the amount of 'Surprise' in a task. 'Surprise', and thus the amount of information to be processed when performing a task, is a square function of the ratio of the required precision of a goal and the proximity to that goal (the accuracy). It thus rises very fast with this ratio, making tasks harder to perform. However, human capacity to process these bits is a logarithmic function of the precision/accuracy ratio. Capacity thus only rises slowly with the precision/accuracy ratio. Errors arise when 'Surprise' outstrips the user's capacity to process. Errors can be reduced by training the user to improve their precision/accuracy ratio for a particular task. Roel will conclude with some examples of how to easily measure bits of 'Surprise' in real time in order to study or adapt interactive systems according to the way the human brain processes information. For further reading: Interactive Inference: A Neuromorphic Theory of Human-Computer Interaction