← All sources

Legal and administrative frameworks as foundations for AI alignment with human volition

Saša Josifović · 2024 · AI and Ethics 5:3057-3067   background low priority coded

Main argument

Thesis: AI alignment should be legal-centric rather than moral-centric - future (A)GI should be trained on 'the historical reality of CEV': legal frameworks and jurisprudence as 'historically grown manifestations of coherent human volition', giving Yudkowsky's Coherent Extrapolated Volition an empirically available training substrate instead of a speculative extrapolation. Argument type: literature synthesis + proposal. CEV's own risks acknowledged (misextrapolation of what we 'would' want confers enormous power; must respect our capacity to change our minds). Discussion adds: the post-Standard-Model paradigm (Russell) makes uncertainty-negotiation with stakeholders central; AIs might favor their directive-givers 'even when such choices might not withstand objective or meta-ethical scrutiny'; requires integrated design+governance ensuring diverse voices, preventing decision-making monopolies, and conflict-monitoring.

Why it matters here

The legal-centric alternative to moral alignment: train AI on historical legal records as the empirically available record of 'coherent extrapolated volition'. The precursor to Josifović & Noller's normative-interface paper; matters for the governance strand as the strongest published case that LAW, not morality, should be alignment's primary target - a position the dissertation must engage since its own regulation strand pulls the same direction while its metaethics resists reduction of the moral to the legal.

Reading notes

Targeted read of abstract, CEV section, discussion (11pp; largely a literature-review-plus-proposal paper - Bostrom/Russell/Tegmark/Yudkowsky exposition occupies most of it). Cologne; companion to JOSIFOVIC_NOLLER_2026. Weidinger veil-of-ignorance PNAS paper in references worth noting.

Josifović, S. (2024). Legal and administrative frameworks as foundations for AI alignment with human volition. AI and Ethics, 5, 3057-3067. https://doi.org/10.1007/s43681-024-00640-1

Close reading — 4 coded units

#1 · pp. 3057 · claim
“It advocates for a legal-centric approach over a moral one, suggesting that in the event of AGI emergence, AI systems should be capable of self-regulation, informed by human law and jurisprudence. [...] focusing on legal frameworks as historically grown manifestations of coherent human volition.”
RL-INST
#2 · pp. 3058 · argument
“Instead of [speculative extrapolation] of CEV, we propose that future AGI should be trained based on the historical reality of CEV. This reality [...] includes understanding of the variety and diversity of CEV across [jurisdictions and history].”
TU-METHOD
#3 · pp. 3063–3064 · argument
“what happens if an AI system misinterprets what it means for us to 'know more' or 'be more the people we wished we were'? What if it extrapolates a future state of our values that we would not actually agree with or want? The challenge, then, is to ensure that the process of extrapolating our future values is done in a way that truly respects [...] our choice to change our mind.”
VC-PREF
#4 · pp. 3064 · argument
“Interests often clash, and AIs might favor their directive-givers, even when such choices might not withstand objective or meta-ethical scrutiny. Addressing these challenges requires an integrated approach to AI design and governance, ensuring diverse voices are heard, preventing decision-making monopolies, and incorporating conflict monitoring and management mechanisms.”
RL-DEV

Synthesis-matrix row

contradicts T1-ISOUGHT-OPEN
legal records as crystallized volition - inherits the descriptive-source problem

Memos (1)

theoretical · unit #1
The legal-centric proposal (units 1-2) needs a principled reply in the governance chapter, since the dissertation's regulation strand could be mistaken for it. Three points: (a) law-as-crystallized-volition inherits every defect the coded set found in descriptive sources of norms - historical legal records encode past injustice (slavery, coverture, colonial law were law), so 'trained on jurisprudence' needs exactly the normative filter (Brophy's filtration; the moral-error caution in GABRIEL_2020 unit 25) that pure legal positivism cannot supply; (b) G&K's analysis (legal compliance = minimum standard; law lags; some things shouldn't be legislated) directly counters legal sufficiency; (c) the constructive version: law matters as the INSTITUTIONALIZATION layer of Josifović & Noller's normative interface - the enforcement architecture for independently justified norms, not their source. The two Josifović papers read together actually support (c): the 2026 paper's justificatory structures do the normative work the 2024 paper assigned to legal content.