Evolving Agents: Interactive Simulation of Dynamic and Diverse Human Personalities

TL;DR

This work tackles the challenge of modeling human personality as a dynamic, evolving construct within embodied agents to serve as versatile design probes. It introduces Evolving Agents, a two-system architecture where Behavior (Plan and Action) and Personality (Emotion, Cognition, Character Growth) interact in a feedback loop to produce believable, diverse behaviors over time, validated through objective metrics (Big Five, Behavioral Similarity) and human evaluations. The authors implement a Unity-based Sandbox with a three-level environment description, memory-driven cognition, and planning guided by GPT-4-turbo, and they demonstrate personality evolution over days with ablation studies confirming the importance of each module. They also show designers can gain inspiration from the agents via workshops, highlighting the platform’s potential as a proactive tool in user-centered design and socio-psychological research, while acknowledging limitations of large models and ethical considerations for human-AI interactions. The work provides a concrete, interpretable pipeline for simulating dynamic personalities and offers practical metrics and workshops to quantify and leverage agent believability and evolution in early-stage design and research contexts. The core contribution is a coherent, testable architecture that integrates emotion, cognition, and growth with planning and social interaction, enabling continuous personality evolution driven by daily experiences, with measurable effects on behavior and design utility.

Abstract

Human-like Agents with diverse and dynamic personalities could serve as an essential design probe in the process of user-centered design, thereby enabling designers to enhance the user experience of interactive applications. In this article, we introduce Evolving Agents, a novel agent architecture that consists of two systems: Personality and Behavior. The Personality system includes Cognition, Emotion, and Character Growth modules. The Behavior system comprises two modules: Planning and Action. We also build a simulation platform that enables agents to interact with the environment and other agents. Evolving Agents can simulate the human personality evolution process. Compared to its initial state, agents' personality and behavior patterns undergo believable development after several days of simulation. Agents reflect on their behavior to reason and develop new personality traits. These traits, in turn, generate new behavior patterns, forming a feedback loop-like personality evolution. Our experiment utilized a simulation platform with ten agents for evaluation. During the assessment, these agents experienced believable and inspirational personality evolution. Through ablation and control experiments, we demonstrated the effectiveness of agent personality evolution, and all of our agent architecture modules contribute to creating believable human-like agents with diverse and dynamic personalities. We also demonstrated through workshops how Evolving Agents could inspire designers.

Figures (19)

• Figure 1: Evolving Agents architecture is inspired by psychology stimulus-response theory, encompasses two subsystems: Behavior and Personality
• Figure 2: Each Agent can create a schedule that aligns with its character. For example, the enthusiastic and open Isabella prefers to leave her dorm and make social plans compared to the shy Benjamin
• Figure 3: Post process procedure of appointment plan
• Figure 4: Agent chooses their communication partner and defines the topic to talk about based on the dialogue memory
• Figure 5: The Agent discusses different topics with various partners, and the depth of each topic continuously increases