New design framework treats AI agents as responsive partners, not prediction machines
Researchers propose reimagining how AI systems interact with users—shifting from rigid transaction logic to dynamic, adaptive relationships. The approach could transform product design across gaming, customer service, and robotics by creating systems that respond to context rather than simply optimize for predetermined goals.
Originaltitel: The Affective Reservoir: From Transactional Rules to Relational Rhythms
<p>This article builds on the argument that design for complex interactive systems should shift from creating transactional interactions to ‘organizing relational complexity’. Grounded in agential realism, we reframe computational agents from black-box predictors to material-discursive apparatuses. We utilize a standard reservoir computing architecture, conceptualized here as the Affective Reservoir, as a diffractive instrument to visualize the co-constitution of gameplay. In doing so, we replace the teleological concept of a fixed ‘goal’ with the agential realist concept of a ‘yearning’: the continuous negotiation of situated tension. By analyzing the reservoir’s dynamics, we show how coherent regimes of interaction emerge within the agent’s internal state space, not from error minimization but from Dynamical Friction; the intense interference pattern generated when the agent’s Re-membered Inertia (habitual momentum) resists the Affective Gradients (situational forcing) of its environment. Ultimately, we argue that by orchestrating an agent’s capacity to be affected via its resistance to and resonance with the environment, designers can move beyond transactional logic to sustain emergent relational phenomena.</p>