Explainable Artificial Intelligence (XAI) aims to bridge the understanding between decisions made by an AI interface and the user interacting with the AI. When the goal of the AI is to teach the user how to solve a problem, user-friendly explanations of the AI’s decisions must be given to the user so they can learn how to replicate the process for themselves. This paper describes the process of defining explanations in the context of a collaborative AI platform, ALLURE, which teaches the user how to solve a Rubik’s Cube. A macro-action in our collaborative AI algorithm refers to a set of moves that takes the cube from initial state to goal state - a process that was not transparent nor accessible when we revealed back-end logic to the front-end for user engagement. By providing macro-action explanations to the user in a chatbot as well as a visual representation of the moves being performed on a virtual Rubik’s Cube, we created an XAI interface to engage and guide the user through a subset of the solutions that can later be applied to the remaining solutions of the AI. After initial usability testing, our study provides some useful and practical XAI user interface design implications.
CITATION STYLE
Bradley, C., Wu, D., Tang, H., Singh, I., Wydant, K., Capps, B., … Srivastava, B. (2022). Explainable Artificial Intelligence (XAI) User Interface Design for Solving a Rubik’s Cube. In Communications in Computer and Information Science (Vol. 1655 CCIS, pp. 605–612). Springer Science and Business Media Deutschland GmbH. https://doi.org/10.1007/978-3-031-19682-9_76
Mendeley helps you to discover research relevant for your work.