Learning the Value Systems of Societies from Preferences

Andrés Holgado-Sánchez; Holger Billhardt; Sascha Ossowski; Sara Degli-Esposti · 2025 · ECAI 2025; arXiv:2507.20728 evidence low priority coded

Main argument

Thesis: since 'it is more adequate to conceive the value system of a society as a set of value systems of different groups, rather than as the simple aggregation of individual value systems', the value-learning problem should be reformalized as learning (a) socially SHARED value groundings (computational representations of what each value means) plus (b) a DIVERSE SET of group value systems (weightings/orderings), via heuristic deep clustering over qualitative value-based preferences from sampled agents. Argument type: formalization + method + empirical evaluation (real travel-decision data). Key framing concessions in the intro: values vary across time and cultures; preferences may be incomplete due to incommensurable values and context-specificity; value-aware AI must reason explicitly about consequences relative to specific human values and adapt to different stakeholders' value systems; manual specification is misspecification-prone, motivating learning from behaviour.

Why it matters here

The technical mirror of the dissertation's core empirical premise: a society's value system is a SET of group value systems, not an aggregate of individuals - and it can be learned from preference data via clustering. Their deep-clustering pipeline on travel decisions is what the xphi corpus does for AI ethics at scale; useful as the ML-venue citation that value-system PLURALITY is now an engineering requirement, not just a philosophical claim.

Reading notes

Read abstract + intro (11pp ECAI paper; technical middle sections skimmed - the formal machinery is beyond the dissertation's needs). Madrid (URJC/CSIC). The intro's framing sentence on incommensurable values + local coherence + normative reasoning cites the same conceptual stack as the philosophical literature.

Holgado-Sánchez, A., Billhardt, H., Ossowski, S., & Degli-Esposti, S. (2025). Learning the Value Systems of Societies from Preferences. ECAI 2025. arXiv:2507.20728

Close reading — 3 coded units

#1 · pp. 1 · claim

“social science and humanities literature suggest that it is more adequate to conceive the value system of a society as a set of value systems of different groups, rather than as the simple aggregation of individual value systems. Accordingly, here we formalize the problem of learning the value systems of societies [...] The method learns socially shared value groundings and a set of diverse value systems representing a given society.”

TU-METHOD VC-INTRA-VALUE

#2 · pp. 1 · argument

“Defining human values and value-based preferences (or value systems) is a challenging task because values vary across time and cultures. In addition, at the time of acting, human preferences may be incomplete due to incommensurable values and context-specificity.”

VC-INTRA-VALUE

#3 · pp. 1 · argument

“truly value-aligned AI systems must be able to explicitly reason about the consequences of their behaviour [...] based on specific human values, allowing their adaptation to the value systems of different stakeholders. [...] As manual design is prone to misspecification, value learning suggests to induce them automatically from demonstrations of value-aligned behaviour.”

AG-UNDERSTANDING

Synthesis-matrix row

complicates T2-PREFERENTISM-BROKEN

learns FROM preferences but into group-plural value systems

Memos (1)

thesis-link · unit #1

Unit 1 is the ML venue's version of the dissertation's stakeholder premise: group-structured value plurality is now a FORMAL modelling requirement at ECAI, not just a humanities claim. Two uses: (a) cite alongside STELA and Lloyd's latent classes as the third methodological cousin of the xphi corpus design (deep clustering : their travel data :: stakeholder/category stratification : folk corpus) - and note the corpus's advantage: their data is revealed travel preferences (thin), the corpus is reason-annotated normative discourse (thick); (b) their 'socially shared value groundings' vs 'diverse value systems' distinction (shared MEANING of values, diverse WEIGHTINGS) maps precisely onto the folk corpus finding-structure - e.g. shared invocation of 'accountability' with divergent weightings across stakeholder groups - and onto the overlapping-consensus structure (shared groundings) the convergentist argument needs. A useful formal vocabulary to borrow.