Temporal Quotas, Interrupted Task Resumption and Sleep Displacement in Generative AI Use

Generative artificial intelligence (GenAI) systems now serve as open-ended collaborators in coding, writing, study and planning. A conversational model can always produce another revision, so a working session has no natural end point. Emerging studies associate compulsive or problematic GenAI use with anxiety, burnout, cognitive overreliance and sleep disturbance, but the field remains conceptually fragmented and gives no grounds for treating "AI addiction" as a clinical diagnosis. Most of this work asks how much people use AI. This paper asks when they stop. It develops the Temporal-Quota Reinforcement Hypothesis (TQRH): short, predictable cycles of access replenishment, especially when the reset time is shown to the user, can act as temporal anchors that shape stopping decisions, anticipatory checking, return-to-use and bedtime in a susceptible subset of users. The hypothesis draws on research on problematic AI use, the I-PACE model, bedtime procrastination, the bidirectional relationship between technology and sleep, chronotype and the resumption of interrupted tasks. The paper makes five contributions: (i) operational definitions of quota events, reset-cue precision and the habitual sleep window; (ii) nine falsifiable hypotheses, with predictions that separate TQRH from five competing explanations, including the possibility that quota events merely mark heavy use; (iii) a two-phase protocol for Delhi-NCR that combines a four-week observational study (privacy-preserving metadata, ecological momentary assessment, sleep diaries and optional wearables) with a randomised interface experiment manipulating reset-time visibility and replenishment timing; (iv) an analysis plan with a smallest effect size of interest and equivalence-based falsification criteria; and (v) a design simulation showing that the number of quota events near bedtime, more than the number of participants, determines statistical power, so recruiting users who reach limits regularly is the most efficient design choice. The paper also separates consumer usage quotas from API rate limiting. TQRH does not claim that quotas cause insomnia. It proposes that the way an access architecture communicates future availability may change when people disengage, and it states the evidence that would show it does not.