Researchers are increasingly using artificial intelligence to construct measures of social, organizational, and occupational characteristics that are missing from conventional surveys. We introduce the AICOME (AI Contextual Measurement) framework to evaluate whether AI‑derived respondent‑level measures can recover individual and group effects in contextual models. The key idea is to obtain an AI measure at the respondent level, then compute its aggregate within the respondent's occupational group and the individual deviation, enabling estimation of both between‑group and within‑group associations rather than treating the AI measure merely as a response predictor.
We validate the framework with the 2022 China Family Panel Studies (CFPS), using occupations as the grouping structure and several job‑related survey variables as benchmarks. For four dimensions—computer use, foreign‑language use, weekly working hours, and management responsibilities—we compare survey measures with AI‑derived measures across four validation types: response‑level, model‑level, contextual, and boundary‑condition. The results show that when rich respondent and job characteristic information is available, AI contextual measurement can recover a substantial portion of the contextual‑model information embedded in the survey variables. Weekly working hours provides the strongest case, with AI measures reproducing the large negative between‑occupation and within‑occupation associations with job satisfaction observed in CFPS.
The framework also identifies clear boundary conditions: performance deteriorates when information is limited to occupation and basic demographics, and recovery weakens further when several related concepts are simultaneously unobserved. Overall, AICOME is most valuable for recovering a limited set of theoretically important constructs from existing rich datasets.
Review