•  
  •  
 
Indiana Law Journal

Document Type

Article

Publication Date

6-2025

Publication Citation

100 Indiana Law Journal 1673

Abstract

The increasing use of AI rather than human surveillance puts pressure on two long-used cultural and (sometimes) legal distinctions: as between human and machine observers and as between content and metadata. Machines do more and more watching through advancing technology, rendering AI a plausible replacement for humans in surveillance tasks. Further, machines can commit to surveil only certain forms of information in a way that humans cannot, rendering the distinction between content and metadata increasingly relevant too for crafting privacy law and policy. Yet despite the increasing importance of these distinctions, their legal importance remains in four key domains of privacy law: Fourth Amendment law, wiretap law, consumer privacy law, and the privacy torts. Given the failure of privacy law to settle conclusively the import of the human/AI and content/metadata distinctions, this Article proposes looking to empirical measures of the judgments of ordinary people to better understand whether and how such distinctions should be made if law is to be responsive to reasonable expectations of privacy.

There is incomplete empirical evidence as to whether the AI/human surveillance and content/metadata distinctions hold weight for ordinary people, and if so, how. To address this empirical gap, this Article presents the results of a vignette study carried out on a large (N = 1000), demographically representative sample of Americans to elicit their judgments of a state surveillance program that collected either content or metadata and in which potential surveillants could be either human or AI. Unsurprisingly, AI surveillance was judged to be more privacy preserving than human surveillance, empirically buttressing the importance of a human/AI distinction. However, the perceived privacy advantage for an AI surveillant was not a dispositive factor in stated preferences regarding technology use. Accuracy—a factor rarely discussed in defenses of state surveillance —was more influential than privacy in determining participants’ preferences for a human or AI surveillant. Further, the scope of information surveilled (content or metadata) strongly influenced accuracy judgments in comparing human and AI systems and shifted surveillance policy preferences as between human and AI surveillants. The empirical data therefore show that the distinction between content and metadata is important to ordinary people, and that this distinction can lead to unexpected outcomes, such as a preference for human rather than AI surveillance when contents of communications are collected.

Share

COinS