Trustworthy Generative AI for Emotion-Aware Content Creation in Digital Media Broadcasting

Counts differ because each database indexes a different set of publications. We treat OpenAlex as the canonical count; Google Scholar is not shown (no API, and crawling it violates its ToS).

Generative Artificial Antelligence (GenAI) can accelerate digital media production, but uncontrolled emotional adaptation may distort meaning, intensify affect, reproduce cultural bias, or weaken audience trust. This study proposes a trustworthy emotion-aware generative AI framework for broadcasting that integrates multimodal affect estimation, uncertainty-aware control, factual grounding, safety screening, provenance recording, and human editorial approval. A reproducible proof-of-concept benchmark was designed using 1,200 simulated broadcast prompts balanced across six program genres and six target emotions. Four configurations were compared through five normalized measures like, emotion alignment, semantic fidelity, factual consistency, safety pass rate, and composite trust. The proposed full framework achieved simulated scores of 86.9%, 88.7%, 91.2%, 96.1%, and 90.7%, respectively, compared with 71.8%, 84.6%, 82.1%, 86.7%, and 78.9% for an unconstrained baseline. Ablation analysis indicates that provenance and human approval most strongly improve accountability, while uncertainty gating reduces harmful over-personalization. The findings demonstrate the feasibility of combining affective creativity with operational safeguards without eliminating human authorship. Because the experiment uses simulated prompts and model outputs, the numerical findings are illustrative rather than evidence of real-world deployment. Future validation should use multilingual broadcast corpora, professional editors, diverse audience panels, and live streaming conditions.