Towards a societal AI alignment benchmark for evaluating human-machine value convergence

Ljubisa Bojic; Dylan Seychell; Milan Cabarkapa · 2026 · Humanities and Social Sciences Communications evidence low priority coded

Main argument

Thesis: proposes a societal AI alignment benchmark measuring human-machine VALUE CONVERGENCE, prototyped by comparing sentiment toward AGI across seven LLMs (scores 3.32-4.12/5; GPT-4 most positive) against three independent human samples, with temporal variation tracked over three consecutive days - finding LLM sentiment diverges from human samples and varies across models and days.

Why it matters here

Empirical LLM-vs-human sentiment comparison toward AGI (7 LLMs vs 3 human samples, Likert, temporal variation over days) as a prototype 'societal alignment benchmark'. Methodologically close to the xphi corpus comparisons; also an instance of exactly the benchmark thinking LaCroix critiques.

Reading notes

Compact treatment. Abstract read.

Bojic, L., Seychell, D., & Cabarkapa, M. (2026). Towards a societal AI alignment benchmark for evaluating human-machine value convergence. Humanities and Social Sciences Communications.

Close reading — 1 coded units

#1 · pp. 1 · evidence

“Seven LLMs, including GPT-4 and Bard, were analyzed and compared against sentiment data from three independent human sample populations. Temporal variations in sentiment were also evaluated over three consecutive days. The results highlighted a diversity in sentiment scores among LLMs, ranging from 3.32 to 4.12 out of 5.”

AG-MORAL-CON ? TU-METHOD ?

Synthesis-matrix row

contradicts T1-ISOUGHT-OPEN

practices distribution-matching as alignment - instance of the fallacy

supports T5-AGENCY-DENIED-EVALUABILITY-KEPT

cross-day sentiment instability (minor empirical)

Memos (1)

comparison · unit #1

Double use: (a) methodological neighbor - comparing LLM output distributions to human population distributions is exactly what the folk corpus enables at scale and with reasons attached (their 3 Likert samples vs the corpus's 366k coded comments); (b) cautionary instance - treating sentiment-distribution match as 'alignment' is the benchmark fallacy LaCroix diagnoses (sociological concordance ≠ ethicality), so cite it as both precedent and foil. The cross-day instability finding adds a minor seventh datum to the LLM-evaluative-instability pile.