Abstract Conversational agents designed for emotionally supportive interactions face challenges in balancing affective responsiveness, computational efficiency, and safety in communication. Prior approaches frequently depend on large-scale models, handcrafted affective objectives, or reinforcement learning from human feedback, which can limit scalability and interpretability. This work presents a lightweight, domain-adapted dialogue generation system based on the T5-small architecture, fine-tuned on MentalChat16K, a curated corpus of real and synthetic emotional-support conversations. The proposed model operates without reinforcement learning or emotion-specific training objectives, yet demonstrates encouraging alignment with affective cues and fluent response generation within the evaluated dataset. Empirical evaluation shows improvements over zero-shot and fine-tuned GPT-2 baselines, achieving BLEU (32.14), ROUGE-L (44.72), and BERTScore-F1 (85.11). Expert human assessments indicated high ratings in coherence, emotional appropriateness, and contextual relevance, with substantial inter-rater agreement. Qualitative error analysis indicated generally conservative and context-aware responses within the evaluated sample. During manual review of this sample, no factual hallucinations, medical overreach, or overtly unsafe responses were observed; however, systematic safety benchmarking was beyond the scope of the present study. This study provides initial evidence that compact transformer-based models, when adapted to domain-specific corpora and evaluated under controlled conditions, can support efficient and affectively appropriate dialogue generation in emotionally supportive non-clinical settings, while requiring further safety validation before broader real-world deployment.
Saleela et al. (2026) studied this question.