Designing User interactions around deep learning for applications that define everyday habils (eg diet, lifestyle, shopping)

为了使人工智能变得有用、可接受和有吸引力，必须了解用户需求和机器学习（ML）模型与HCI研究的原因。这种理解将导致高效、有效和透明地使用数据，通过旨在从用户交互中学习的ML模型来满足用户需求。我们的研究旨在使最终用户能够对他们日常使用的人工智能系统进行交互式控制（例如用于精神/身体健康，工作/休闲/社会支持的应用程序），并通过自适应可解释界面传达系统如何工作和学习。研究Q1 -将人工智能作为更好地支持人们日常生活的工具的理想场所是什么？心理/身体健康，工作/休闲/社会支持）？研究Q2 -在识别需求时，我们如何构建反映用户交互的深层模型？研究Q3 -用户与智能系统交互的好处是什么？这是否会带来积极的体验和可解释的AI？研究Q4：我们能否围绕深度学习构建一个用户交互框架，以适应不同的应用程序？我们计划与最终用户和HCI/AI领域专家进行用户访谈和观察，以深入了解如何构建反映用户交互的ML模型。我们还将使用定量调查和AMT研究来进一步了解和建模。通过确定将人工智能设计为改善人们日常生活的交互工具的含义，并使用研究原型对其进行测试，我们的目标是评估研究原型并构建人类算法交互的指导方针。

For AI to be useful, acceptable, and attractive, it is essential to understand user needs and reason for Machine Learning (ML) models with HCI research. This understanding would lead to using data efficiently, effectively, and transparently to answer user needs with ML models that are designed to learn from user interactions. Our research aims to empower the end users with interactive control over the AI systems which they use everyday (such as applications for mental/physical wellbeing, work/leisure/social support), and to communicate how the system works and learns with adaptive explainable interfaces. Research Q1 - What are the ideal places to place AI as a tool to better support people's everyday life (eg. mental/physical wellbeing, work/leisure/social support)?Research Q2 - On identifying the needs, how might we build deep models mirroring user interactions?Research Q3 - What are the benefits of user interactions with intelligent systems? Does this result in a positive experience and explainable AI?Research Q4 - Can we build a framework for user interactions around deep learning that could be adapted to different applications?We plan to conduct user interviews and observations with end users and HCI/AI domain experts to find insights into building ML models that mirror user interactions. We will also use quantitative surveys and AMT studies for further understanding and modelling. By identifying the implications to design AI as an interactive tool to better people's everyday life and testing them with research prototypes, we aim to evaluate the research prototypes and construct guidelines for human-algorithm interactions.