The Inadequacy of Laboratory Studies and the Usefulness of Interviews in Artificial Intelligence and Robotics

• In Event: Lab Studies Reloaded? Machine Learning, Ethnography, and Critical STS

Abstract

Artificial intelligence (AI) and robotics laboratories appear very compelling as ethnographic sites, probably because of the strong fictional and journalistic narratives behind the scientific routine; places where robots and artificial minds are created must be interesting observation settings. Following my journey into methodology throughout the early stages of my doctoral research, this presentation compares strengths and weaknesses of laboratory studies and interviewing through anecdotes and a summary of my preliminary findings. During piloting, and while I was considering a mixture of ethnography and interviewing as my main approach to learn about expectations and actual capabilities of AI/robotics, I have conducted a two-months ethnographic micro-study at a prominent UK machine learning robotics laboratory. During that time, I learned a number of interesting lessons about conducting everyday research: academic/technological competition, circuits of financial interest, the very slow technical progress, and the ways all these factors impact heavily the very existence of robots. However, after conducting two pilot interviews, these lessons became again immediately clear in a much shorter time, with deeper insights and the time spent in the laboratory appeared nearly redundant, sometimes even leading to awkward misunderstandings. During in-depth interviews, AI/robotics specialists have the opportunity to express themselves in an unrushed, straightforward manner, reflect upon the social dimensions of their field, provide with definitions, flag out research obstacles, and more. Hence, my PhD’s methodology was eventually entirely interview-based. This panel will be a great opportunity to compare my sceptical view towards AI/robotics-related laboratory studies with other researchers’ (hopefully) more positive experiences.

Author

• Vassilis Galanos, University of Edinburgh