Existing approaches to infer user traits and generate responses consistent with a persona rely on static prompting. However, these methods have drawbacks such as lack of calibrated uncertainty, ignoring sequential evidence, and drifting during long interactions. To address this, we present AI YOU, a framework that continually updates a personality profile with 22 dimensions from conversation and embodies it in a personal digital twin.
The system combines prompting, Bayesian updating, and conformal prediction for persona inference. A periodically refreshed memory anchor and cognitive memory with three layers preserve persona consistency over long interactions.
The main results show that AI YOU achieves:
- Conformal coverage ranging from 0.921 to 0.976;
- Improved uncertainty calibration and reasoning grounded in memory;
- Enhanced persona fidelity over static prompting in role playing over 100 turns while reducing trait drift, for most evaluated backbones under adversarial settings with multiple agents.
The prototype AI YOU Town initializes an imaginative twin world for future interaction. The online demo is available at AI YOU Town.
Blogger's Review: The introduction of the AI YOU framework opens new possibilities for personalized interactions, especially in maintaining personality consistency over long conversations, which is crucial for enhancing user experience. The dynamic updating mechanism combined with Bayesian methods undoubtedly lays a solid foundation for future applications of digital twins.