NormGenesis: Multicultural Dialogue Generation via Exemplar-Guided Social Norm Modeling and Violation Recovery

Abstract

Social norms govern culturally appropriate behavior in communication, enabling dialogue systems to produce responses that are not only coherent but also socially acceptable. We present NormGenesis, a multicultural framework for generating and annotating socially grounded dialogues across English, Chinese, and Korean. To model the dynamics of social interaction beyond static norm classification, we propose a novel dialogue type, Violation-to-Resolution (V2R), which models the progression of conversations following norm violations through recognition and socially appropriate repair. To improve pragmatic consistency in underrepresented languages, we implement an exemplar-based iterative refinement early in the dialogue synthesis process. This design introduces alignment with linguistic, emotional, and sociocultural expectations before full dialogue generation begins. Using this framework, we construct a dataset of 10,800 multi-turn dialogues annotated at the turn level for norm adherence, speaker intent, and emotional response. Human and LLM-based evaluations demonstrate that NormGenesis significantly outperforms existing datasets in refinement quality, dialogue naturalness, and generalization performance. We show that models trained on our V2R-augmented data exhibit improved pragmatic competence in ethically sensitive contexts. Our work establishes a new benchmark for culturally adaptive dialogue modeling and provides a scalable methodology for norm-aware generation across linguistically and culturally diverse languages.

Figures (17)

• Figure 1: Comparison of generation outputs in Korean. Prior methods li2023normdial produce pragmatically inconsistent responses, including honorific misuse and unnatural tone (highlighted in red). In contrast, our framework yields culturally and pragmatically coherent outputs (highlighted in blue).
• Figure 2: NormGenesis Overview. Our framework consists of four stages: (1) culturally grounded norm and style design, (2) scenario–situation construction across norm adherence, violation, and resolution types, (3) exemplar-based iterative refinement using semantically aligned exemplars, and (4) multi-turn dialogue generation with turn-level annotations. Each stage is evaluated and refined iteratively to ensure pragmatic consistency and cultural alignment, as described in Section \ref{['Sec:3']}. The RQ evaluation protocol is described in Section \ref{['Sec:4.2']}.
• Figure 3: Subnorm examples
• Figure 4: Adherence norm example (EN)
• Figure 5: Adherence norm example (KR)