Habeas Rogare: Reversing the Perspective in the Harm Principle and Cognitive Sovereignty in the Age of AI
Year: 2026
Type: preprint
Abstract: ABSTRACT In practice, the Harm Principle (Mill, 1859) is applied from the perspective of the one who acts: the actor is considered legitimate until harm is proven, and the burden of proof falls on the affected party. For harms that are slow, cumulative, or invisible to the affected party, proof becomes practically impossible, and the harm remains outside the very principle meant to prevent it. The paper proposes reversing the perspective: "The other's freedom ends where my freedom begins." The reversal consists of two operations: shifting the presumption onto the emitter and specifying the protected object, namely the preconditions of autonomy (bodily integrity, cognitive capacities, access to information about risk). It applies between any parties (individuals, groups, companies) and is limited by four trigger criteria, in response to Mill's own objection to the theory of "social rights." Leaded gasoline, secondhand smoke, the REACH regulation, and the precautionary principle are earlier partial applications. Applied to the capacity to question, the reversal grounds Habeas Rogare: the right not to have one's own capacity for reasoning eroded by a third party's design, without one's knowledge or without a real choice. Part II tests the criteria against generative AI assistants delivered with zero friction as the default mode. Five studies with different designs point in the same direction: the loss occurs when AI eliminates cognitive participation, not from the use of AI as such, while pedagogically configured tutors produce gains. The mechanisms are described through FCPT and CCPT. The proposed remedy is an alternative obligation: the provider either demonstrates that the default mode does not reduce users' capacity to solve tasks without assistance, or offers a visible, graduated control of cognitive friction (the Mechanism of Cognitive Stimulation, Cognitive Slider). The request addressed to the industry thus becomes an obligation, without prohibiting the frictionless mode for informed adults. The record contains the English version and the original Romanian version of the paper.
Author: Stan, Adrian
Language: English
Subfield: Artificial Intelligence
Field: Computer Science
Domain: Physical Sciences
Open Access status: green
OpenAlex ID: W7225731111
Cites:
Cited by:
Related to: