Surprising encounters: how moments of surprise illuminate practitioners' beliefs about AI

Abstract

The paper develops the concept of surprise as an analytical lens through which to understand practitioners' beliefs concerning narrow artificial intelligence (AI) adoption in different contexts. The paper draws inductively on empirical data collected across pharmaceutical, higher education and arts sectors, to illuminate how moments of surprise can provide a window to people's beliefs and feelings about the current AI moment. Across the three cases, we conducted a mixture of interviews, focus groups and observations with 65 UK-based practitioners. The analytical approach included thematic analysis and close critical reading. We examine three categories of surprise that shed light on beliefs: “big surprises” (e.g. encountered in relation to AlphaFold, DALL.E), “everyday surprises” (e.g. unexpected smaller scale successes or disappointments) and instances of “lack of surprise” (e.g. regarding AI outputs and political economy). Together these findings reveal a complex interplay of hope, disappointment and, ultimately, power dynamics around AI development and implementation, which brings widespread AI hype further into question. While much prior research focuses on people's expectations and imaginaries of AI, we argue that moments of surprise, i.e. when something happens that contradicts expectations, can illuminate underlying beliefs and how they are challenged in contexts of AI integration.