The Unknowability of Autonomous Tools and the Liminal Experience of Their Use

Abstract

Recently companies are increasingly adopting intelligent technologies, such as autonomous tools and artificial intelligence, to assist complex knowledge works that are traditionally carried out by human experts. These tools can independently learn and execute novel actions. The input–output relationships of these tools, however, are unknowable to human experts. This calls for analysis of how humans may work differently while interacting with such tools. To this end, we conduct a comparative case study at one of the world’s largest semiconductor manufacturers. We investigate how chip designers interact with two families of design technologies: one following a traditional designer-centric approach in which the designer knows what outputs particular inputs to the tools will generate, and another relying on autonomous tools that continually surprise the user. Our inquiry reveals that, when using autonomous tools, designers can hardly understand the design generated by the tools and become more like laboratory experimentalists in that their primary job is experimenting with different inputs and assessing the quality of the finished design. Their interactions with the tools are marked by ambiguity, and the design is moved forward along multiple design trajectories in accordance with a multifarious temporality.