Comp-HuSim: Persistent Digital Personality Simulation Platform
Our paper Comp-HuSim: Persistent Digital Personality Simulation Platform was published at UMAP (ACM Conference on User Modeling, Adaptation and Personalization) 2024.
Comp-HuSim is a multi-agent persona simulation platform powered by generative AI models (GPT-3.5 turbo, GPT-4 turbo, and Mixtral 8x7B). Each Comp-HuSim generative agent is created with random variables such as age, gender identity, economic status, Big-5 personality characteristics, geographic location, and more. Agents are imbued with backstory memories, textured histories, diverse hobbies and interests, and a series of distinct attributes.
In early experiments, the agents engage in conversation with one another, reflect on their past experiences, and play games together in surprisingly sophisticated ways. Complex forms of human-like behaviors become visible in the Comp-HuSim project demonstration, with potential applications in virtual assistants, conversational interactions, and personalized human-robot interaction.
Co-authors: Chengyu Fan, Zaynab Tariq, Nafis Saadiq Bhuiyan, Michael G. Yankoski