The efficiency-gain illusion: People underestimate the rate of AI use and overestimate its benefits on simple tasks

TEXT START: "People are increasingly turning to AI assistance for simple tasks, e.g., arithmetic, spell-check, and answering simple questions."

THE DISSECTION

This is a behavioral economics study documenting that humans systematically voluntarily surrender cognitive labor to AI even when the surrender produces zero net benefit — and that doing so entrenches further surrender. Three large pre-registered studies (N=2,691) establish two core miscalibrations: people use AI more than they think, and they overestimate the time savings from doing so. A session-level carryover effect is identified — prior AI use predicts subsequent AI use, deepening the miscalibration over time.

THE CORE FALLACY

The paper frames this as a behavioral bias problem — a rational choice failure, a calibration error correctable through better information. This is the wrong level of analysis. The study is inadvertently documenting the molecular mechanism of productive participation collapse, and it doesn't recognize what it has found.

The DT lens identifies what the authors miss: the act of delegating simple cognitive tasks to AI is not a detour from productive participation — it is productive participation eroding itself. The human who outsources arithmetic, spell-checking, and factual recall is not making an irrational short-term choice. They are executing a process of progressive cognitive offloading with zero perceived downside at the individual level and terminal downside at the systemic level.

The miscalibration is not incidental. It is the feature, not the bug, of AI integration.

HIDDEN ASSUMPTIONS

The paper smuggles in two assumptions that collapse under DT scrutiny:

1. That efficiency is the operative variable driving AI adoption. It is not. Humans delegate cognitive tasks for reasons of cognitive relief, social signaling, and status — efficiency is post-hoc rationalization. This is why people use AI even when it is inefficient. The authors acknowledge this but treat it as a puzzle rather than the primary signal.

2. That overreliance is a risk to be managed within the current system. This framing — "overreliance feedback loop" as intervention target — assumes the system is worth preserving. Under DT mechanics, this feedback loop is not a risk. It is the mechanism. The loop is the collapse, and it is proceeding by voluntary, enthusiastic, self-miscalibrating adoption.

SOCIAL FUNCTION

This paper is transition management theater — it identifies a real phenomenon (mass cognitive offloading) but locates the problem in individual miscalibration rather than structural displacement. It performs the intellectual work of seeming to grapple with AI's effects while leaving the system's trajectory unchallenged. It is valuable research that has been slotted into a framework that drains it of its most uncomfortable implications.

THE VERDICT

The authors have documented, with rigorous pre-registration, that humans are voluntarily, enthusiastically, and with systematic self-blindness evacuating their own cognitive labor market — even for simple tasks where the delegation provides no efficiency benefit. This is not a behavioral curiosity. This is the execution of the Discontinuity Thesis at the micro-level, one delegation decision at a time, across millions of individuals who cannot correctly perceive what they are doing to themselves.

The "session-level carryover effect" is the feedback loop. It is not a risk. It is the collapse, self-administered.