The article proposes an interpretive rule, a compositional-effects test identifying the decision unit under Article 22 GDPR together with the allocation of the burden of establishing it, and documentation duties calibrated to inference chains.
Abstract
Two instruments of EU digital law place inference at their centre and mean different things by it. Article 3(1) of the AI Act uses the capability to infer constitutively: it is the central feature separating the regulated category from conventional software. The GDPR never defines inference, yet governs it protectively: the consequences follow from the processing of personal data and from what the inference says about, or does to, a person, whether or not the technology that produced it qualifies as an AI system. The two perimeters are not concentric. Their non-coincidence remained invisible in single-shot systems; agentic architectures make it operationally acute. The thesis: inferential capability does not determine legal scope, and its absence does not create immunity. The framework is two-level. Inference performs two legal functions, constitutive and protective; the protective function operates through three pathways - identificatory, attributive and decisional. Composition is not a fourth pathway but a cross-cutting architectural dimension which, with reach, persistence and reviewability, is what agentic architectures modify. Three concepts support it: the inferential threshold, the inferential reach and the inferential chain, mapped onto the chain of imputation. Regulation (EU) 2026/1744 left the constitutive criterion untouched and inserted a provision contemplating outputs that influence the inputs of future operations, without supplying any rule of aggregation. The article proposes an interpretive rule, a compositional-effects test identifying the decision unit under Article 22 GDPR together with the allocation of the burden of establishing it, and documentation duties calibrated to inference chains.
The paper formulates the Control Responsibility Principle (CRP), which shifts AI ethics from a focus on what systems know to a focus on how their behavior is governed, who holds authority over that governance, and how such authority is to be justified, distributed, and contested.
Generative AI encodes the majority's way of knowing as the default infrastructure of knowledge itself as the default infrastructure of knowledge itself, and law must learn to govern at that level of model training.
Legal theory has strong accounts of why defective legal acts fail. It has a weaker account of why legal systems respond to that failure in categorically different ways. Speech act theory addresses the first question, and legal theorists have drawn on it because it converges with Hart’s insight that nullity is not a sanction. But these frameworks share a structural limitation: binary outcomes. Civil law distinguishes categories of invalidity—nonexistence, absolute nullity, relative nullity—each with distinct procedural consequences. This article argues that speech act theory cannot capture these categories. First, it shows both classical and cooperative accounts produce Boolean outcomes. Second, it demonstrates that civil law assigns defects to discrete categories. Third, it establishes that this assignment tracks normative judgments about protected values, not structural features. Fourth, it responds to objections, including graded speech acts. Even non-binary frameworks cannot supply the evaluative judgments the law’s taxonomy of invalidity requires.
Two systems can supply the same person-linked attribution to the same institutional decision maker yet fall into different legal categories: one uses neural signals, the other text or behaviour. A source-bound rule therefore permits circumvention, while an all-purpose category of"mental data"risks treating fallible outputs as facts about the mind. This article reads the 2025 UNESCO Recommendation on the Ethics of Neurotechnology as non-binding guidance and develops a source-neutral trigger for technologically mediated, person-linked mental-state attribution. Through selective critical synthesis, conceptual engineering, functional legal comparison, and matched counterfactual cases, it separates elicitation, attribution, and use as cumulative objects of regulation. The analysis also distinguishes two harm pathways from two independently assessed duty series. Seven ordered questions and two escalation predicates assign permitted practices to three duty tiers. Presumptive prohibition is reserved for materially autonomy-affecting non-consensual closed-loop intervention and for specified combinations of covert or coercive inference, core attributes, consequential decisions, and manipulation. Common entry does not erase source-related aggravators: invasiveness, embodiment, and closed-loop capacity can add duties or set a higher minimum, while consequential non-neural inference can reach the same tier. The resulting scheme is a classification device, not an empirically validated regulatory outcome.
Gailmard (2026) and Dowding and Miller (2026) draw attention to a methodological gap that political science has yet to adequately address: causal identification, on its own, does not constitute causal explanation. Both contributions help clarify the gap. We extend their analyses in three ways. First, we identify three places where Gailmard’s framework relies on conceptual commitments that operate implicitly within his formal apparatus. The coherence equivalence used in his proofs is stronger than the three coherence properties he states; the bundle reading of theoretical models gives Proposition 5 its content but imposes a tolerance condition that the framework does not address; and Proposition 6 on generalization supplies a necessary-and-sufficient condition for generalization without specifying when that condition obtains. Second, we extend Dowding and Miller’s explanatory pluralism by locating constitutive explanation and explanation by constraint within a structured inferential framework, and we argue that equilibrium explanation comes in distinct varieties — equifinality and comparative statics — that play different roles in political science formal theory. Third, we introduce a nested modeling framework that separates three levels — data, conceptual model, and theory — clarifying how identification, non-causal explanation, and theoretical inference jointly support causal knowledge.
Dwayne Woods· Chinese Political Science Re...· 0 citations
We use cookies to run the site and, with your consent, for analytics and to show ads.
See our Cookie Policy.
