Agentic Literacy Debt: A Structural Problem the AI Literacy Field Has Not Yet Named

Existing artificial-intelligence literacy frameworks are structurally inadequate for a world in which autonomous agents make decisions on behalf of humans, according to a paper submitted on 21 Apr 2026 [1]. The authors argue the deficit is not a temporary gap but a compounding societal debt. The paper, posted to arXiv, introduces the term “agentic literacy debt” to describe the accumulating deficit that grows when agentic AI systems are deployed at scale without corresponding literacy infrastructure [1][2]. Autonomous AI agents now plan, decide, and act for users in healthcare, financial services, and workplace settings, often without step-by-step human approval [1][2]. Current literacy frameworks were built for a world where humans evaluate AI outputs and then choose whether to act; they lack vocabulary for a user who has delegated authority to an agent whose actions may not be observable, reversible, or controllable [1][2]. The debt compounds through three reinforcing channels: normalization of opaque delegation, multi-agent ecosystem complexity, and institutional path dependence [1][2]. It is incurred by the organizations that deploy agents but paid by the users, patients, and citizens on whose behalf the agents act [1][2]. Evidence drawn from healthcare, financial fraud, and global equity contexts suggests the gap is already consequential [1][2]. The authors contend the problem is structural and will not be closed by curriculum reform alone [1][2]. They call for a reframing of AI literacy as a governance capability rather than an evaluative one [1][2]. The paper does not propose a specific regulatory mechanism but frames the debt as a governance challenge that demands institutional responses before agentic systems become further entrenched in critical sectors [1][2]. The argument arrives as autonomous and semi-autonomous systems proliferate across industries. In agriculture, for example, unregulated chemical inputs have produced long-term health consequences that were not priced into initial productivity gains; studies published between 2011 and 2020 attribute 45 different types of cancers afflicting rural farm workers in India to pesticide exposure, with occupational contact identified as a major trigger [3]. The parallel drawn by the paper’s framework is that opaque delegation to agents, like unregulated chemical adoption, externalizes costs onto end-users while benefits accrue to deploying institutions [1][2]. The concept of path dependence invoked in the paper has historical analogues. In Italian economic history, protectionist tariff laws passed between 1877 and 1887 penalized agricultural exports from the south while favoring industrial production in the north, creating a structural gap that persisted into the 21st century [5]. The paper warns that institutional path dependence in agentic AI deployment could similarly lock in literacy deficits that become progressively harder to reverse [1][2].