Synthetic or Authentic? Building Mental Patient Simulators from Longitudinal Evidence
arXiv:2603.22704v1 Announce Type: new Abstract: Patient simulation is essential for developing and evaluating mental health dialogue systems. As most existing approaches rely on snapshot-style prompts with limited profile information, homogeneous behaviors and incoherent disease progression in multi-turn interactions have become key chellenges. In this work, we propose DEPROFILE, a data-grounded patient simulation framework that constructs unified, multi-source patient profiles by integrating demographic attributes, standardized clinical symptoms, counseling dialogues, and longitudinal life-event histories from real-world data. We further introduce a Chain-of-Change agent to transform noisy longitudinal records into structured, temporally grounded memory representations for simulation. Experiments across multiple large language model (LLM) backbones show that with more comprehensive profile constructed by DEPROFILE, the dialogue realism, behavioral diversity, and event richness have c
arXiv:2603.22704v1 Announce Type: new Abstract: Patient simulation is essential for developing and evaluating mental health dialogue systems. As most existing approaches rely on snapshot-style prompts with limited profile information, homogeneous behaviors and incoherent disease progression in multi-turn interactions have become key chellenges. In this work, we propose DEPROFILE, a data-grounded patient simulation framework that constructs unified, multi-source patient profiles by integrating demographic attributes, standardized clinical symptoms, counseling dialogues, and longitudinal life-event histories from real-world data. We further introduce a Chain-of-Change agent to transform noisy longitudinal records into structured, temporally grounded memory representations for simulation. Experiments across multiple large language model (LLM) backbones show that with more comprehensive profile constructed by DEPROFILE, the dialogue realism, behavioral diversity, and event richness have consistently improved and exceed state-of-the-art baselines, highlighting the importance of grounding patient simulation in verifiable longitudinal evidence.
Executive Summary
The article introduces DEPROFILE, a novel framework that enhances mental patient simulation by leveraging longitudinal evidence to construct unified, multi-source profiles. Traditional simulation methods, which rely on snapshot prompts with limited data, often result in homogeneous behaviors and incoherent disease progression. DEPROFILE addresses these challenges by integrating demographic attributes, clinical symptoms, counseling dialogues, and longitudinal life-event histories. Additionally, a Chain-of-Change agent transforms longitudinal records into structured, temporally grounded memory representations. Experimental results across multiple LLMs demonstrate that DEPROFILE improves dialogue realism, behavioral diversity, and event richness, outperforming state-of-the-art baselines. This work underscores the critical value of grounding simulation in verifiable longitudinal data for more authentic mental health interactions.
Key Points
- ▸ Integration of longitudinal evidence into patient simulation
- ▸ Use of DEPROFILE to unify multi-source patient data
- ▸ Chain-of-Change agent improves temporal grounding of longitudinal records
Merits
Comprehensive Data Integration
DEPROFILE's ability to consolidate demographic, clinical, counseling, and longitudinal data into a unified profile represents a significant advancement in simulation fidelity.
Enhanced Realism and Diversity
Experimental validation shows measurable improvements in dialogue realism, behavioral diversity, and event richness, validating the efficacy of the proposed framework.
Demerits
Complexity and Data Dependency
The framework’s reliance on longitudinal data may pose challenges in scalability, data availability, and integration for real-world clinical applications without adequate infrastructure.
Expert Commentary
This paper makes a compelling case for the necessity of longitudinal grounding in mental patient simulation. The DEPROFILE framework is a sophisticated response to a persistent problem in the field—namely, the lack of coherence and realism in multi-turn interactions due to snapshot-based approaches. The Chain-of-Change agent is particularly noteworthy for its ability to operationalize longitudinal data into structured memory representations, a critical step for temporal consistency. While the results are impressive, the authors should acknowledge the potential limitations of data heterogeneity across sources and the need for robust data governance frameworks to mitigate ethical concerns. Moreover, future work should explore the scalability of DEPROFILE across diverse clinical populations and its applicability beyond dialogue systems into broader mental health intervention platforms. Overall, this is a seminal contribution to the intersection of AI and mental health simulation.
Recommendations
- ✓ 1. Encourage open-source availability of DEPROFILE to foster reproducibility and widespread adoption.
- ✓ 2. Initiate interdisciplinary discussions among ethicists, clinicians, and AI researchers to address data privacy and consent issues inherent in longitudinal data usage.
Sources
Original: arXiv - cs.CL
