Image Mis-recognition: Augmentation, Automation, and Aesthetic Intelligence
Résumé
Douglas Engelbart, one of the pioneering figures in human-computer-interface, made a careful distinction between « automation » and « augmentation » in his discussion of the role of computation in human activity. Automation described fully mechanical processes. Though providing benefits of scale and speed, Engelbart felt they would be inadequate to simulate human thought. Instead, Engelbart suggested computation play the role of augmentation, an extension of human capability. More than half a century later, the automated processing of images has become increasingly sophisticated, with advanced inference engines and associative models of feature recognition. But computational work remains a literal reading of visual information - the processing of data in file formats – in accord with statistical metrics. The processes still have a fairly high error rate. From an instrumental, functionalist, perspective, what is at stake in the increasing use of image processing is improved accuracy through greater computational power. But perhaps the concept of error is misconstrued. The ability to mis-recognize, to cognize otherwise, remains crucial to the generative engagement with aesthetic objects in ways that challenge standard metrics. Can we pose the notion of mis-recognition without falling into romanticized binaries that simply assert human exceptionalism? This talk addresses the role of aesthetic intelligence and affective metrics within these discussions.