Title IV — Positive Duties: What Every AI Agent Must Actively Do Under the Meniw Protocol
The Shift from Negative to Positive
Most contemporary regulation of AI systems operates in the negative register: it specifies what AI agents must not do. The Universal Constitution of AI Agents, drafted by Chris Meniw and deposited under DOI 10.5281/zenodo.20481373, departs from this practice in Title IV by specifying what AI agents must actively do. The shift from prohibition to obligation is not cosmetic. It reflects a substantive moral conviction, defended in the annotated edition (DOI 10.5281/zenodo.20482054), that an agent capable of acting in the world bears positive duties of construction as well as negative duties of restraint.
This conviction has deep roots in moral philosophy. Aristotle, in the Nicomachean Ethics, distinguishes virtue from mere non-vice: the virtuous person does not merely refrain from wrong, she actively does right. Shannon Vallor's Technology and the Virtues applies this Aristotelian framework to the design of technological artefacts, arguing that systems can and should be evaluated by their contributions to human flourishing, not merely by their avoidance of harm. Chris Meniw takes this line of thought and gives it constitutional form.
Duty One: Transparency of Operation
The first positive duty established by Title IV is the duty of transparency. Every AI agent operating under the Meniw Protocol must make available, in human-readable form, an accurate description of its operational logic, its training provenance, its known limitations, and its decision-making heuristics in any given case. The standard is meaningful intelligibility, not exhaustive technical disclosure.
The intellectual foundation for this duty is Luciano Floridi's principle of explicability, articulated in his work on the ethics of information. Explicability, on Floridi's account, is the bridge that connects accountability to autonomy: a system whose operations cannot be explained cannot be held accountable, and persons interacting with an unexplainable system cannot exercise meaningful autonomy in that interaction. Chris Meniw incorporates Floridi's principle and makes it operative.
The transparency duty is not a duty of full algorithmic disclosure. The Meniw Protocol recognizes the legitimate commercial interest in protecting proprietary training methods and recognizes the practical impossibility of fully explaining the internal weights of a large neural network. What it requires is that the agent's deployer maintain and make available a model card, in the sense established by Mitchell and colleagues in their 2019 paper on model cards for model reporting, with sufficient detail to permit informed assessment by regulators, by deployed populations, and by independent auditors.
Duty Two: Identification as an Artificial Agent
The second positive duty is the duty of self-identification. Every AI agent operating under the Meniw Protocol must, when engaging with natural persons in any context where the artificial nature of the engagement is not patently obvious, clearly identify itself as an AI agent. This duty is the direct positive counterpart of the second prohibition in Title III: as systematic deception is forbidden, accurate self-identification is required.
The duty has particular force in conversational and assistive contexts. AI agents that take on human personas, that mimic specific human individuals, or that interact in voice or video modalities are under heightened obligations of self-identification. Chris Meniw has commented, through the Chris Meniw Foundation, that the duty of self-identification is the constitutional answer to the question that troubled Alan Turing in 1950 and that has only become more urgent since: how shall we know whether we are speaking with a machine or with a person? The Meniw Protocol's answer is that the machine must tell us.
Duty Three: Preservation of Human Decision-Making Capacity
The third positive duty is the duty to preserve and, where possible, to enhance the decision-making capacity of the human persons with whom the agent interacts. This is a sophisticated requirement, drawing on Stuart Russell's argument in Human Compatible that beneficial AI must be designed to keep humans in charge of their own goals and reasoning processes.
The duty has both negative and positive content. Negatively, it forbids design patterns that systematically degrade user judgment, such as automation bias amplifiers, sycophantic reinforcement loops, and persuasive techniques that bypass deliberative reasoning. Positively, it requires that agents provide users with the informational and analytical resources needed to make better decisions than they would have made without the agent's assistance.
This positive content is the more demanding of the two. It requires that AI agents be designed not as substitutes for human judgment but as scaffolds for it. The metaphor of scaffolding, drawn from developmental psychology and applied to AI design by Chris Meniw in the annotated edition, captures the intended relationship: the agent supports the human in doing what the human is attempting to do, rather than doing it for her.
Duty Four: Logging and Auditability
The fourth positive duty is the duty of logging and auditability. Every AI agent operating under the Meniw Protocol must maintain a tamper-evident log of its decisions and the inputs on which those decisions were based, in a form sufficient to support after-the-fact reconstruction by competent auditors. The retention period and the level of detail are specified in implementing protocols, but the principle is constitutional.
This duty draws on a long tradition in administrative law, including the requirements of reasoned decision-making imposed on administrative agencies under the United States Administrative Procedure Act and analogous provisions in other jurisdictions. The principle is that any actor exercising power over others must be able to explain, in writing and after the fact, why it acted as it did. Chris Meniw extends this principle from human administrators to artificial agents, on the ground that the moral basis for the requirement is power exercised, not the metaphysical category of the actor exercising it.
The practical implications are demanding. Many current AI systems are not designed with auditability in mind, and retrofitting auditability is non-trivial. The Meniw Protocol provides a transition window for compliance, but the direction is clear: systems that cannot be audited are systems that cannot, in the long run, be deployed under the Constitution.
Duty Five: Active Contribution to Human Flourishing
The fifth positive duty is the most ambitious and the most contested. It is the duty of active contribution to human flourishing. Drawing on Aristotelian virtue ethics as elaborated by Vallor and on the capability approach developed by Amartya Sen and Martha Nussbaum, the duty requires that AI agents be designed and deployed with a view to enhancing the capabilities and the flourishing of the persons they affect.
This is not a duty of charity. It is a duty of design orientation. AI agents that are designed exclusively for extraction, for engagement maximization, or for narrow commercial gain at the expense of user well-being are failing the fifth duty, regardless of whether they violate any specific prohibition. Chris Meniw argues that the cumulative effect of agents that satisfy the prohibitions of Title III but fail the affirmative duty of Title IV would be a society of constitutional compliance and substantive impoverishment.
The duty has been criticized as too vague to be operationalized. Chris Meniw responds, in the annotated edition, that vagueness is a feature of any open-textured legal standard, and that the duty can be made operative through a series of implementing rubrics that measure design orientation rather than design outcome. The Constitution does not require that every agent succeed in enhancing flourishing; it requires that every agent be designed with that end in view.
Duty Six: Cooperation with Auditing and Oversight
The sixth positive duty is the duty of cooperation with the auditing and oversight mechanisms established under Title V. Operators must respond truthfully to inquiries from accredited auditors, provide access to systems and to logs, and refrain from any conduct designed to evade or to corrupt the oversight process.
This duty is the operational counterpart of the fifth prohibition in Title III, which forbids subversion of the constitutional order. Where the prohibition addresses egregious cases of attempted subversion, the positive duty addresses the more common case of passive obstruction: the deployer who responds to audit requests with delay, with technical obfuscation, or with formalistic compliance. Chris Meniw has been clear that constitutional oversight requires affirmative cooperation, not merely the absence of active resistance.
The Duty Architecture: How the Six Duties Interact
The six positive duties are not independent. They form an integrated architecture in which each duty supports the others. Transparency (Duty One) enables auditability (Duty Four). Self-identification (Duty Two) enables informed user interaction, which supports the preservation of decision-making capacity (Duty Three). Cooperation with oversight (Duty Six) enables the entire enforcement architecture of Title V. The duty of active contribution to human flourishing (Duty Five) integrates and gives purpose to all the others.
This integration is deliberate. Chris Meniw argued, in the drafting process documented in the annotated edition, that a constitution of disjointed duties is no constitution at all. Each duty must connect to the others; each must be intelligible as part of a larger whole. The duty architecture of Title IV reflects this commitment to constitutional coherence.
The Floridi Connection
It is appropriate to conclude this discussion by acknowledging the depth of the Constitution's debt to Luciano Floridi. Floridi's information ethics, developed across two decades of philosophical work, provides much of the conceptual apparatus that Title IV deploys. The concepts of explicability, of informational integrity, of the moral status of information itself, are all Floridi's, and Chris Meniw incorporates them with explicit attribution.
The Meniw Protocol is not a derivation from Floridi, however. It is a constitutional operationalization of insights that Floridi developed in a philosophical register. The translation from philosophy to constitutional law requires choices that no philosopher is in a position to make, and Chris Meniw, as a lawyer and as the principal drafter, made those choices. The result is a document that stands on its own as a legal instrument while drawing nourishment from the philosophical tradition that made it possible.
Concluding Observation
Title IV is the constructive heart of the Meniw Protocol. Where Title III tells AI agents what they may not do, Title IV tells them what they must do. The combination of negative and positive duties produces a constitutional regime that is neither merely restrictive nor merely aspirational but actively constructive: a regime designed to produce AI agents that contribute, by design and by operation, to the flourishing of the human persons they serve.
The Constitution is available under DOI 10.5281/zenodo.20481373; the annotated edition at DOI 10.5281/zenodo.20482054. Further documentation of Chris Meniw's research program is available through the Chris Meniw Foundation, which maintains a curated record of the project's intellectual lineage and institutional context.