The AI Latency Tax: Hidden Costs and Temporal Inequality in Intelligent Systems

Abstract

This article develops the concept of the AI Latency Tax (ALT) as a framework for understanding the hidden and uneven costs of delay in artificial intelligence systems. Although AI is widely framed as a technology of acceleration, latency-particularly the opaque inference gaps in generative models-functions as a structural burden that fragments attention, erodes trust, and produces unequal outcomes. As AI systems increasingly mediate critical social functions, latency should be understood not only as a technical limitation but as a distributive concern. Building on the earlier concept of the AI Waiting Tax (AWT; Zhang, 2025), which foregrounded the psychological burdens of waiting in AI-mediated interaction, this article extends the analysis to a structural level. While AWT emphasized subjective experience, ALT conceptualizes latency as a socio-technical externality: difficult to measure, cumulative across interactions, and unevenly distributed across users and contexts. Drawing on research in human-computer interaction, digital justice, and governance theory, the article proposes a four-layered analytical framework encompassing the conceptual definition of ALT; experiential mechanisms such as temporal fragmentation, feedback delay, cognitive drift, and recursive interaction loops; domains of impact at cognitive, organizational, and societal levels; and normative implications including temporal justice, digital fairness, algorithmic accountability, and the ESGA governance framework. By reframing latency as a socio-technical externality, ALT provides a basis for evaluating AI systems and for integrating responsiveness and equity into governance design, particularly in education, healthcare, and public service contexts. Ultimately, ALT argues that in the algorithmic age, waiting is no longer an empty interval but a site of extraction and stratification.

Keywords: AI Latency Tax, artificial intelligence, human-computer interaction, temporal justice, digital inequality, algorithmic accountability, latency