This post is part of the ongoing series of reading reflections on HCI papers and articles. You can find the other posts ‘here.’
Guidelines for Human-AI Interaction (CHI 2019)
Authors: Amershi, S., Weld, D., Vorvoreanu, M., Fourney, A., Nushi, B., Collisson, P., Suh, J., Iqbal, S. T., Bennett, P., Inkpen, K., Teevan, J., Kikin-Gil, R., and Horvitz, E.
Available at Microsoft Research.
This article provides 18 design guidelines for human-AI interaction, which are distilled from over 150+ recommendations from various sources. The paper also describes developing and validating the guidelines through interviews with multiple stakeholders and expert evaluations. The guidelines are intended to be applicable across a wide range of AI-infused systems.
Prompts
Why is this article important?
This article explains how the authors arrived at the 18 guidelines and details the 4 phases they took to develop the guidelines and the methods they used to validate them. I believe this article best explains how different guidelines from various domains can be generalized to benefit the broader AI community. The article also provides a good overview of the current state of the art in AI design guidelines.
The 18 guidelines, as validated by the authors, are the most applicable guidelines for most AI systems. The authors have gone to great lengths to ensure they are simple, straightforward, and easy to evaluate. These guidelines apply to designing, validating, and evaluating AI systems. The guidelines apply to various AI systems, from voice assistants to autonomous vehicles, making it a valuable resource.
What do the authors seem to be assuming about the future of A.I. and Human Interaction?
The authors envision a future where all AI systems are designed with the guidelines in the article, which they believe will make AI systems more predictable and intuitive for users to interact with. They also wish AI systems are more transparent and understandable to users, which will help users trust the systems more. The authors also hope that the guidelines will allow designers and developers of AI systems to evaluate and improve their AI systems’ design to be more human-centric.
How might you integrate the reading into your academic/professional work?
I intend to use the guidelines in my work to evaluate and improve the design of my AI systems. I also plan on using the guidelines to assess and improve the design of other AI systems I encounter. I believe the guidelines will be helpful in my work because they apply to a wide range of AI systems and are easy to evaluate.