Affective Change Through Affective Artificial Intelligence

Abstract

The chapter proposes a new methodology of critical and agentive interventions in human–artificial intelligence (AI) interaction. If human decision-making relies on the combination of affect and cognition, could we make emotional AI more “affective”, and thus “more effective” and trustworthy? Is a controlled human–AI interaction via emerging media the way to approach this conundrum? I interrogate two subsets of affective AI: facial expression analysis and emotion recognition as apparatuses of computer vision and machine learning. I suggest a critical, interventionist, and new materialist methodology when interacting with these systems. My approach is positioned in the framework of affect theory, critical data and technology studies, and surveillance capitalism. The methodology is based on Karen Barad's “agential realism” (2007) during creative practice with affective AI. I differentiate between affect and emotion by bringing in different aspects of affect theory, neuroscience, and quantum mechanics. I view affect as the capacity of becoming and mattering, whereas emotion is the performative materialization of affect. Moreover, agential realism serves me as a theoretical approach to making an impact when interacting with AI processes: it is upon encountering an obstruction (object – virtual or actual) that diffraction (change of matter) occurs. During the process of human interaction with affective AI, errors and bias (obstructions) cause diffraction, which materializes affect into emotion – emotion morphs into motivation, which pushes agents into action. This in the end materializes as an experience, a memory, and even physical reality.