Cognitive Debt: The Cumulative Cognitive Cost of AI-Augmented Knowledge Work
Citation. Ruttenberg, D. (2026). Cognitive debt: The cumulative cognitive cost of AI-augmented knowledge work. PsyArXiv. https://doi.org/10.31234/osf.io/yhvec_v1
Summary. Ruttenberg proposes cognitive debt as the cumulative, delayed cost to attention, learning, and mental health when chronic AI reliance reduces active cognitive engagement in knowledge work. The conceptual preprint distinguishes this longitudinal liability from momentary cognitive load, adaptive offloading, and decision-level automation bias. It specifies attentional erosion, effort displacement, and affective depletion as interacting mechanisms, and proposes threshold-sensitive, longitudinal tests including a neurodivergent early-warning hypothesis.
Key claims
- Cognitive debt is a temporal and potentially hard-to-reverse liability: short-run AI-assisted output can remain intact while repeated reliance weakens sustained attention, effortful learning, uncertainty tolerance, and later unassisted transfer.
- The proposed model integrates attentional erosion, effort displacement, and affective depletion over time; it predicts non-linear acceleration after an individual capacity threshold is crossed rather than a smooth correspondence between AI use and immediate performance loss.
- The preprint distinguishes passive offloading from reflective scaffolding: the relevant design question is whether AI assistance displaces conflict resolution and problem structuring or preserves productive engagement while reducing extraneous burden.
- Neurodivergent knowledge workers are proposed as an early-warning population because narrower sensory and cognitive tolerance windows may reveal unsustainable AI-mediated work patterns sooner; the paper frames this as design-relevant expertise rather than deficit.
- The paper proposes falsifiable longitudinal tests using transfer performance, sustained-attention measures, burnout indicators, and AI-use patterns, including an operational threshold defined as concurrent SART d-prime decline and emotional-exhaustion elevation without a corresponding decline in AI-assisted output quality.
Evidence & limitations
- This is a single-author conceptual PsyArXiv preprint, not a direct empirical test of cognitive debt, its mechanisms, the proposed thresholds, or neurotype differences in workplace knowledge work.
- Its numerical threshold definitions and time-window predictions are proposed operationalizations, not fitted estimates or validated diagnostic criteria; the cited causal learning evidence is largely from educational settings.
- The author declares a patent relevant to ethical wearable sensing devices for sensory accommodation and reports no external funding; any future physiological monitoring application requires participatory, privacy-preserving, non-punitive governance.
- The Drive PDF SHA-256 exactly matches the public OSF download; Crossref and the PsyArXiv API independently identify the same title, author, DOI, publication date, and public preprint record.
- This is related to, but not a duplicate of, 2026-ruttenberg-cognitive-debt-ai-research and 2026-ruttenberg-neurodivergent-expertise-ai-work: it foregrounds a distinct cumulative model with specified mechanisms, operational threshold, and falsifiable predictions.