Dynamic Personality Adaptation in Large Language Models via State Machines
2026-02-25 • Computation and Language
Computation and LanguageHuman-Computer InteractionMachine Learning
AI summaryⓘ
The authors identify that Large Language Models (LLMs) struggle to change their personality during conversations, which limits their effectiveness in complicated interactions. They created a flexible system using state machines to represent and change personality states based on the ongoing dialogue. Their setup includes a way to score personality traits continuously, which helps guide the model's responses in real-time. Tested in medical training for calming tense situations, their approach adapts to users and even influences user behavior positively. They also show that this method works well even with smaller, simpler classifiers instead of big LLMs.
Large Language Modelsstate machinespersonality simulationInterpersonal Circumplexdynamic personality scoringdialogue systemsmodular pipelinebehavioral alignmentfine-tuned classifiersde-escalation training
Authors
Leon Pielage, Ole Hätscher, Mitja Back, Bernhard Marschall, Benjamin Risse
Abstract
The inability of Large Language Models (LLMs) to modulate their personality expression in response to evolving dialogue dynamics hinders their performance in complex, interactive contexts. We propose a model-agnostic framework for dynamic personality simulation that employs state machines to represent latent personality states, where transition probabilities are dynamically adapted to the conversational context. Part of our architecture is a modular pipeline for continuous personality scoring that evaluates dialogues along latent axes while remaining agnostic to the specific personality models, their dimensions, transition mechanisms, or LLMs used. These scores function as dynamic state variables that systematically reconfigure the system prompt, steering behavioral alignment throughout the interaction.We evaluate this framework by operationalizing the Interpersonal Circumplex (IPC) in a medical education setting. Results demonstrate that the system successfully adapts its personality state to user inputs, but also influences user behavior, thereby facilitating de-escalation training. Notably, the scoring pipeline maintains comparable precision even when utilizing lightweight, fine-tuned classifiers instead of large-scale LLMs. This work demonstrates the feasibility of modular, personality-adaptive architectures for education, customer support, and broader human-computer interaction.