Paper short abstract:

This paper presents a collaborative interdisciplinary project between anthropology and artificial vision programming. It aims to design alternative interfaces that 'model interpretation' —instead of displaying AI outputs— through non-standard co-presence between both human and artificial agencies.

Paper long abstract:

Despite its relevance and popularity in the cognitive ecology of the current digital age, AI often exercises its agency 'off-stage' (Gell), through technological artifacts that mediate most of our daily actions and, at the same time, offers them intelligence, vision, self-learning and other qualities that were, until now, regarded as human qualities. Such is the case of artificial neural networks for image recognition, classification and generation; a 'secret agent' that co-shapes visual culture, redefining —ideologically or unconsciously— both what it means to see, and what there is to see, since how we see the world, makes it visible.

This paper focuses on an experimental interdisciplinary project developed with artificial vision programming, currently developed in collaboration with deep learning engineers and researchers in computational co-creativity and information visualization. The goal is to use anthropological theories about the 'co-composition' of agency in our technological cognitive ecologies (Latour; Lévy; Sánchez-Criado) in dialogue with the ethnographic findings that arise from working on/using algorithms alongside programmers. This project aims to design alternative human-AI interfaces that are able to activate a non-standard co-presence between both human and artificial agencies. This allows a 'modeling interpretation' (Drucker) of visual and archival resources, through the visualization of (and interaction with) artificial vision 'learning' processes, instead of displaying output information that emerges from its obscured activity.