Speech in the Panaudikon: On the Relationship Between Speech Processing Research and Surveillance

Abstract

The potential of using speech technologies for surveillance purposes are tacitly acknowledged within the field of speech processing, however there has been little research on how theories from surveillance studies interact with speech processing research. In this paper, we analyse 20% of papers published in 2025 from Interspeech, one of the main venues for speech research, through Steve Mann’s compass of veillances. We find that engagement with considerations of surveillance is almost entirely absent in the literature, despite risks being readily apparent in a number of research tasks, with some tasks being more prone to risks of surveillance. Encouragingly, we find tasks and findings that provide avenues for counter-surveillance research, yet these introduce other risks. We argue that identifying mechanisms for addressing the risk of surveillance requires discussion, and therefore provide a framework for researchers to identify how their work engages with surveillance, and how they can limit the risks of it. To this end, we present a list of questions for researchers to engage in how their work might contribute to surveillance.