Exploring The Design of Prompts For Applying GPT-3 based Chatbots: A Mental Wellbeing Case Study on Mechanical Turk

Large-Language Models like GPT-3 have potential to enable HCI designers and researchers to create more human-like and helpful chatbots for specific applications. But evaluating the feasibility of these chatbots and designing prompts that optimize GPT-3 for a specific task is challenging. We present a case study in tackling these questions, applying GPT-3 to a brief 5 minute chatbot that anyone can talk to to better manage their mood. We report a randomized factorial experiment with 945 participants on Mechanical Turk that tests three dimensions of prompt design to initialize the chatbot (identity, intent, and behaviour), and present both quantitative and qualitative analysis of conversations, and user perceptions of the chatbot. We hope other HCI designers and researchers can build on this case study, for other applications of GPT-3 based chatbots to specific tasks, and build on and extend the methods we use for prompt design, and evaluation of prompt design.

Harsh Kumar, Ilya Musabirov, Jiakai Shi, Adele Lauzon, Kwan Kiu Choy, Ofek Gross, Dana Kulzhabayeva, Joseph Jay Williams