Abstract
As interactive systems shift toward probabilistic, generative models, algorithmic bias evolves into implicit hallucination and epistemic erasure, creating ecosystemic ripple effects. This paper proposes Restorative Critical Play, a theoretical design framework in practice that looks beyond interaction to interrogate and repair the epistemic violence within Generative AI. Drawing on Critical Design and Black Game Studies, the work argues that designers must expose the entangled realities of generative outputs rather than merely subverting mechanics.
We introduce three design patterns, Algorithmic Inoculation, Counter-Narrative Jamming, and Data Visceralization, to transform the AI black box from an oracle of truth into a site of critical inquiry. These work to inspire detection, motivation and correction. By reconfiguring users from passive consumers into active auditors of systemic bias, these patterns make visible the cultural forces obscured by frictionless interfaces. Ultimately, this work demonstrates how play can function as community repair, equipping users to navigate and reshape the complex sociotechnical impacts of emerging