We examined training for a job interview through a role-play simulation using a ChatGPT-based chatbot. We questioned whether a humanized large language model
(LLM)-based chatbot avatar that mimics human-to-human interaction conventions through its appearance – i.e., a formal-looking chatbot who plays the role of the interviewer – will improve the effectiveness of the
training.
Three avatars, each visually different, were
used: formal, sporty, and casual avatar. The 85 trainees were randomly assigned to one of the three groups, and each received job interview training by the assigned avatar as
the interviewer. Trainees received an evaluation of their performance during the interview and then repeated the job interview simulation again.