Augmenting Dialog with Think-Aloud Utterances for Modeling Individual Personality Traits by LLM
This addresses the problem of better mimicking human personality in AI chatbots for applications like personalized interactions, though it is incremental as it builds on existing personality modeling with LLMs.
The study tackled modeling individual personality traits in text chat by augmenting dialog data with think-aloud utterances (TAUs) for LLM training, resulting in persona LLMs that more closely aligned with speakers' Agreeableness and Neuroticism traits compared to using original dialog data.
This study proposes augmenting dialog data with think-aloud utterances (TAUs) for modeling individual personalities in text chat by LLM. TAU is a verbalization of a speaker's thought before articulating the utterance. We expect "persona LLMs" trained with TAU-augmented data can mimic the speaker's personality trait better. We tested whether the trained persona LLMs obtain the human personality with respect to Big Five, a framework characterizing human personality traits from five aspects. The results showed that LLMs trained with TAU-augmented data more closely align to the speakers' Agreeableness and Neuroticism of Big Five than those trained with original dialog data. We also found that the quality of TAU-augmentation impacts persona LLM's performance.