Interpretive Variability in Generative AI Feedback: A Structural-Determinist Account of User Experience

Abstract

User experience (UX) research on AI feedback typically treats response variability as noise to be minimized rather than as a phenomenon requiring explanation. Yet this variability is persistent-ranging from integration to dismissal-even when feedback procedure is held constant. This study examines this divergence through a controlled interaction in which 20 expert participants (doctoral/master’s professionals) received AI-generated critiques of their own texts, produced under the same feedback methodology. Treating AI output as linguistic perturbation generated under a standardized procedure, we conducted a four-stage qualitative analysis tracing initial orientation, feedback features, reported stance and interactional trajectories. We identified three recurrent trajectories: congruent uptake (integration), pragmatic filtering (instrumental use) and defensive conservation (dismissal). These trajectories did not align with discipline or gender but reflected human-AI comparisons regulating receptivity. Drawing on Maturana’s structural determinism, we reframe variability not as design failure but as a lawful outcome of history-shaped organization, shifting design goals from persuasion towards perturbation calibration and interpretive autonomy.