RHealthTwin: Towards Responsible and Multimodal Digital Twins for Personalized Well-being (2506.08486v1)

Abstract: The rise of LLMs has created new possibilities for digital twins in healthcare. However, the deployment of such systems in consumer health contexts raises significant concerns related to hallucination, bias, lack of transparency, and ethical misuse. In response to recommendations from health authorities such as the World Health Organization (WHO), we propose Responsible Health Twin (RHealthTwin), a principled framework for building and governing AI-powered digital twins for well-being assistance. RHealthTwin processes multimodal inputs that guide a health-focused LLM to produce safe, relevant, and explainable responses. At the core of RHealthTwin is the Responsible Prompt Engine (RPE), which addresses the limitations of traditional LLM configuration. Conventionally, users input unstructured prompt and the system instruction to configure the LLM, which increases the risk of hallucination. In contrast, RPE extracts predefined slots dynamically to structure both inputs. This guides the LLM to generate responses that are context aware, personalized, fair, reliable, and explainable for well-being assistance. The framework further adapts over time through a feedback loop that updates the prompt structure based on user satisfaction. We evaluate RHealthTwin across four consumer health domains including mental support, symptom triage, nutrition planning, and activity coaching. RPE achieves state-of-the-art results with BLEU = 0.41, ROUGE-L = 0.63, and BERTScore = 0.89 on benchmark datasets. Also, we achieve over 90% in ethical compliance and instruction-following metrics using LLM-as-judge evaluation, outperforming baseline strategies. We envision RHealthTwin as a forward-looking foundation for responsible LLM-based applications in health and well-being.

• The paper presents RHealthTwin, a framework that mitigates LLM hallucinations and ethical risks in digital twins via a structured, multimodal approach.
• It employs a Responsible Prompt Engine to convert unstructured inputs into controlled, context-aware responses for personalized healthcare.
• Evaluations across diverse datasets show state-of-the-art performance with BLEU 0.41, ROUGE-L 0.63, and BERTScore 0.89, ensuring high ethical compliance.

RHealth Twin: Towards Responsible and Multimodal Digital Twins for Personalized Well-being

The paper "RHealth Twin: Towards Responsible and Multimodal Digital Twins for Personalized Well-being," authored by Rahatara Ferdousi and M. Anwar Hossain, presents a robust, ethical framework for the development and application of AI-powered digital twins in healthcare. The research targets complex issues inherent in deploying LLMs within consumer health contexts, addressing hallucinations, biases, lack of transparency, and ethical misuse, all through the implementation of the RHealthTwin framework.

Overview

Digital twins in healthcare provide virtual replicas of patients, offering personalized assistance across various health domains. This paper proposes the Responsible Health Twin (RHealthTwin), which centralizes the concept of multimodal digital twins supported by a Responsible Prompt Engine (RPE). RPE is designed to mitigate the risk of hallucinations typically brought on by unstructured user prompts in traditional LLM configurations. By dynamically organizing these inputs into controlled and structured slots, RHealthTwin produces contextually aware, fair, reliable, and explainable outputs beneficial for mental health support, symptom triage, nutrition planning, and activity coaching.

Key Components

RHealthTwin integrates several modules:

1. Context-Aware Task Personalization Module: It tailors responses based on individual health data, aligning with personal goals.
2. Adaptive System Behavior Management Module: Facilitates dynamic model roles and tone configuration, enhancing accountability.
3. Filter Constraints Module: Applies ethical safeguards to prevent bias, ensuring fairness.
4. Justification and Grounding Module: Utilizes evidence-backed outputs to provide transparency and rationale.

Additionally, a feedback loop refines the system’s responses based on user satisfaction, allowing RHealthTwin to evolve as a personalized health companion.

Evaluation and Results

RHealthTwin was evaluated across four datasets covering distinct health domains, including mental and physical well-being. Utilizing metrics such as BLEU, ROUGE-L, and BERTScore, RPE achieved state-of-the-art results, with scores of BLEU = 0.41, ROUGE-L = 0.63, and BERTScore = 0.89. Additionally, the framework consistently surpassed instruction-tuned baselines in ethical compliance and instruction-following, attaining scores over 90%. Specific results from datasets, such as MentalChat16K and MTS-Dialog, confirmed RHealthTwin's ability to generate structured and ethically aligned responses.

Implications

The implications of RHealthTwin are substantial both theoretically and practically, influencing future developments in AI for healthcare. The framework not only fosters consumer trust through responsible AI practices but also serves as a groundwork model for embedding ethical AI systems into the healthcare domain. The structured prompt generation and ethical constraints exemplified here can particularly impact roles where AI might augment human decision-making without replacing human judgment.

Future Directions

While RHealthTwin advances the digital twin concept and explores new horizons in healthcare AI, challenges remain, particularly concerning the integration of stakeholder input, adaptability to diverse cultural contexts, and scalability. As the paper delineates, collaboration with healthcare professionals and ethicists is crucial to refine the system further and ensure its alignment with established medical standards. Further research is warranted to explore deeper integration methods and evaluate the framework's application across varied linguistic settings.

Conclusion

This paper encapsulates a significant framework that contributes to the personalized healthcare field through AI-driven digital twins, marking progress towards responsible, ethical, and reliable applications. The RHealthTwin exemplar, illuminated by rigorous evaluation and alignment with WHO guidelines, lays the groundwork for future innovations and responsible applications across healthcare support systems.