HarmonicAI: Human-guided collaborative multi-objective design of explainable, fair and privacy-preserving AI for digital health

人工智能（AI）是现代医疗系统数字化转型的最重要支柱之一，它将利用通过可穿戴设备和传感器收集的越来越多的真实世界数据，并考虑疾病与个人/人群之间的多种复杂相互作用。虽然人工智能支持的数字健康服务和产品在数量和种类上都在迅速扩大，但大多数人工智能创新仍然是概念验证的形式。关于AI是否值得信任，一直存在争议。欧盟AI HLEG定义了值得信赖的AI系统应该是合法的，道德的和强大的。为了将其转化为可操作的做法，提供可解释性、公平性和隐私至关重要。在可解释AI、公平AI和隐私保护AI领域已经进行了大量的研究。然而，目前的研究工作，以解决这三个挑战是分散的，并最终在各种解决方案与异构，非互操作性，甚至冲突的能力。HarmonicAI的宏伟愿景是建立一个人机协作的多目标设计框架，以促进数字健康的可解释、公平和隐私保护的AI。HarmonicAI汇集了人工智能，医疗保健，物联网，数据科学，隐私，网络安全，软件工程，HCI和工业设计领域的专家，共同目标是为人工智能从业者制定具体的技术和操作指南，以设计以人为本，特定领域，以需求为导向的值得信赖的人工智能解决方案，加速人工智能驱动的数字医疗服务的可扩展部署，并向公众保证数字医疗中的人工智能正在以合乎道德和值得信赖的方式开发和使用。

Artificial Intelligence (AI) is one of the most significant pillars for the digital transformation of modern healthcare systems which will leverage the growing volume of real-world data collected through wearables and sensors, and consider multitude of complex interactions between diseases and individual/population. While AI-enabled digital health services and products are rapidly expanding in volume and variety, most of the AI innovations remain in the form of proof-of-concept. There is a continuous debate regarding whether AI is worthy of trust. The EU AI HLEG has defined that trustworthy AI systems should be lawful, ethical and robust. To translate it into actionable practices, provision of explainability, fairness and privacy is crucial. A considerable volume of research has been conducted in the areas of explainable AI, fair AI and privacy-preserving AI. However, the current research efforts to tackle the three challenges are fragmented and have culminated in a variety of solutions with heterogeneous, non-interoperable, or even conflicting capabilities. The ambitious vision of HarmonicAI is to build a human-machine collaborative multi-objective design framework to foster coherently explainable, fair and privacy-preserving AI for digital health. HarmonicAI draws together proven experts in AI, health care, IoT, data science, privacy, cyber security, software engineering, HCI and industrial design with an underlying common aim to develop concrete technical and operational guidelines for AI practitioners to design human-centered, domain-specific, requirement-oriented trustworthy AI solutions, accelerating the scalable deployment of AI-powered digital health services and offering assurance to the public that AI in digital health is being developed and used in an ethical and trustworthy manner.