Driving Generative Agents With Their Personality
This work addresses the challenge of enhancing realism in video game NPCs for game developers and players, but it is incremental as it applies existing LLM capabilities to a new domain-specific context.
The research tackled the problem of making video game characters more human-like by using Large Language Models (LLMs) to incorporate personality information from psychometric values, resulting in the LLM consistently representing given personality profiles and accurately generating content based on personality, as shown through repurposing the IPIP questionnaire.
This research explores the potential of Large Language Models (LLMs) to utilize psychometric values, specifically personality information, within the context of video game character development. Affective Computing (AC) systems quantify a Non-Player character's (NPC) psyche, and an LLM can take advantage of the system's information by using the values for prompt generation. The research shows an LLM can consistently represent a given personality profile, thereby enhancing the human-like characteristics of game characters. Repurposing a human examination, the International Personality Item Pool (IPIP) questionnaire, to evaluate an LLM shows that the model can accurately generate content concerning the personality provided. Results show that the improvement of LLM, such as the latest GPT-4 model, can consistently utilize and interpret a personality to represent behavior.