TransforM – Transparency for Machinery in Personal Pervasive Smart Spaces

转换项目是SPP-2199项目Perform的后续项目，该项目调查了个性作为机器人智能空间的交互范式。Performance构建了所谓的“房间智能”，即拥有机器人元素和智能厨房电器的智能生活环境，可以作为一个连贯的实体来处理，并对其用户表现出一定的个性。在用户研究中对这一范例的探索过程中，很明显，希望这种环境透明(用户可以理解幕后发生的事情)和希望技术不可见之间存在冲突，特别是在家庭环境中(用户希望事情简单地工作，而不是在舒适的气氛中处理所有细节)。因此，提供适当程度的透明度对接受这样的环境至关重要。基于可解释人工智能和人际沟通中的透明度概念，我们调查了如何在房间智能等个人智能空间中构建和沟通透明度。我们建议设计一个透明构建块的语言和工具箱，我们将在几个用户研究中对其进行评估。其中，这些研究将帮助我们了解用户如何感知透明度构建块，透明度与信任和接受之间的关系，以及是否有可能确定“透明度的最佳点”。此外，我们将探索考虑到个人用户特征和偏好(例如，技术专长、个性差异)的适应性透明度的潜力。基于这些研究，我们将为未来的智能空间提取一般的可扩展交互范例和设计指南，以确保这些范例既不太不透明也不太压倒性，并将通过适当的透明度水平从用户那里获得足够的信任度。

The project TransforM is a follow-up project to the SPP-2199 project PerforM, which investigated personality as an interaction paradigm for robotic smart spaces. PerforM has built a so-called "room intelligence", i.e., smart living environment with robotic elements and intelligent kitchen appliances, which can be addressed as a coherent entity and exhibits a certain personality towards its users. During the exploration of this paradigm in user studies, it became clear that there is a conflict between the wish for transparency of such an environment (the user can understand what is going on behind the scenes) and the wish for invisibility of technology, especially in the home context (the user wants things to simply work, and not deal with all the details, in a cozy atmosphere). Providing the right level of transparency turns therefore out to be crucial for acceptance of such an environment. Building on transparency concepts from Explainable AI and interpersonal communication, we investigate how transparency can be constructed and communicated in personal smart spaces such as the room intelligence and beyond. We propose to design a language and toolbox of transparency building blocks which we will evaluate it in several user studies. Among other, these studies will help us to understand how the transparency building blocks are perceived by the user, how transparency relates to trust and acceptance and whether it is possible to identify a “transparency sweet spot”. Moreover, we will explore the potential of adaptable transparency which considers individual user characteristics and preferences (e.g., technical expertise, personality differences). Based on these studies, we will extract general scalable interaction paradigms and design guidelines for future smart spaces, which ensure that these are neither too intransparent nor too overwhelming, and will – by the right level of transparency – receive an adequate calibrated amount of trust from their users.