Large language models are increasingly used to simulate social and political audiences, but it remains unclear whether persona-conditioned generations provide coherent reception patterns or align with real public opinion. We study audience reactions to morally salient political messages, using Trump’s posts as a relevant case in point, through two theory-grounded human value frameworks, Inglehart-Welzel cultural map and Moral Foundations Theory. Across issue-specific message bundles, we analyze stance together with agreement, legitimacy, offensiveness, and endorsement, asking whether synthetic personas produce systematic reception differences, which persona dimensions explain those differences, and whether aggregate synthetic distributions align with matched Pew public-opinion patterns. The paper examines three related dimensions of synthetic-audience evaluation: internal differentiation, explanatory structure, and external plausibility. It contributes a reproducible study design for political audience simulation and an empirical account of where synthetic moral audiences are informative, and where they should not be treated as calibrated population proxies.

Detecting Schwartz values in political texts is hard: cues are often implicit, and neighboring values differ by fine distinctions. Two remedies are widely assumed to help: more surrounding document text, and explicit moral knowledge. Knowledge-based retrieval has improved benchmarks elsewhere, but whether either transfers here is untested, because published systems vary context, knowledge, and model family at once. We separate these factors under matched conditions on the ValuesML/Touché ValueEval format. The input ranges from the target sentence to a local window to the full document. Retrieval is either absent or drawn from a curated moral knowledge base, injected by early, late, or cross-attention fusion. Supervised DeBERTa-v3 encoders are compared against zero-shot LLMs from 12B to 123B. More context is not uniformly better: full-document input improves the encoders by 2.5-3.8 macro-F1 points but does not consistently help the LLMs. Retrieved knowledge helps more reliably, improving every model family and context under early fusion. A control substituting random knowledge-base entries shows the families gain differently: LLMs from relevance, encoders from exposure to the value ontology. Neither larger encoders nor larger LLMs guarantee gains, and early fusion outperforms both trainable variants. Value-sensitive NLP should evaluate context, knowledge, and model family jointly.