Toward a Theory of Justice for Artificial Intelligence (Daedalus 151(2):218-231)

Iason Gabriel · 2022 · Daedalus 151(2), Spring 2022, AI & Society issue interlocutor medium priority coded

Main argument

Thesis: the Rawlsian basic structure of society 'is best understood as a composite of sociotechnical systems', and AI increasingly shapes elements of the basic structure (criminal justice risk assessment, welfare allocation, economic mobility) - therefore egalitarian norms of distributive justice apply to AI deployed in these contexts, requiring (1) a standard of PUBLIC JUSTIFICATION, (2) support for citizens' RIGHTS, and (3) substantively FAIR OUTCOMES with particular attention to impacts on the WORST-OFF. Key reframing: the moral properties of algorithms are products of the social systems of deployment, not internal to models. Political theory has failed technology twice over - liberal bracketing (technology exogenous to justice) and Marxist determinism (technology explains everything, leaving no moral vantage point); the essay's two-way interaction view (we shape technology; technology shapes us; the choices are morally evaluable) restores AI to justice's jurisdiction.

Why it matters here

The middle term of the Gabriel arc (2020 alignment -> 2022 justice -> 2025 claims): AI enters the Rawlsian BASIC STRUCTURE because the basic structure is a composite of sociotechnical systems - so egalitarian justice norms (public justification, citizens' rights, fair outcomes, priority to the worst-off) apply to deployed AI directly. The political-philosophy bridge for the governance chapter, and the reframing that 'the moral properties of algorithms are not internal to the models... but a product of the social systems within which they are deployed.'

Reading notes

Targeted extraction from the Daedalus volume (pdf pp. 219-233 of 384; essay pp. 218-231). The volume's other relevant essay (Tasioulas) coded separately. Note the essay explicitly diagnoses political theory's neglect of technology (Rawls mentions technology 3 times) and rejects Marxist determinism as the alternative.

Gabriel, I. (2022). Toward a Theory of Justice for Artificial Intelligence. Daedalus, 151(2), 218-231.

Close reading — 3 coded units

#1 · pp. 218 · claim

“[The demand for justice] reframes much of the discussion around 'AI ethics' by drawing attention to the fact that the moral properties of algorithms are not internal to the models themselves but rather a product of the social systems within which they are deployed.”

RL-INST

#2 · pp. 219–220 · claim

“The first [claim] is that the basic structure of society is best understood as a composite of sociotechnical systems: that is, systems that are constituted through the interaction of human and technological elements. [...] The second is that AI increasingly shapes elements of the basic structure in relevant ways, and hence that its design, development, and deployment all potentially interface with principles of justice in this context.”

RL-INST

#3 · pp. 218 · argument

“These norms entail that the relevant AI systems must meet a certain standard of public justification, support citizens' rights, and promote substantively fair outcomes, something that requires particular attention to the impact they have on the worst-off members of society.”

RL-LEGITIMACY

Synthesis-matrix row

complicates T3-PROCEDURALISM-INCOMPLETE

supplies the political-institutional home; worst-off standard vs G&K scoping tension

Memos (1)

comparison · unit #2

The Gabriel arc is now fully coded: 2020 (fair principle-selection for alignment) -> 2022 (AI in the basic structure; justice norms apply to deployment) -> 2024 STELA (empirical elicitation) -> 2025 G&K (claims-based deliberation) -> 2026 Fischli (agentic autonomy). The 2022 essay supplies what the 2020 paper lacked - a POLITICAL-INSTITUTIONAL home for alignment norms - and is the direct ancestor of Ferretti's institutional-change thesis. For the governance chapter: the worst-off priority (unit 3) gives the Immigration chapter a Rawlsian test with teeth - immigration AI patently affects some of the globally worst-off, so on Gabriel's own standard it faces the strictest justification demands, yet (per G&K unit 14) its subjects are excluded from the justificatory community. The tension between Gabriel-2022's standard and G&K-2025's scoping is a publishable observation.