AI Hype, Imaginaries, and Industry Discourse
From Narratives to Infrastructures: Three Cuts at the Politics of AI Discourse
The papers gathered here share a common starting point — that talk about generative AI is never just talk — but they push that insight in different directions. Wang2025-zy treats AI discourse as a cartography of competing visions; Stanusch2026-ec treats it as strategic work performed by industry actors during moments of controversy; and Dodds2026-df treats it as a structuring infrastructure that conditions journalism itself. Read together, they trace an arc from mapping imaginaries, to unpacking how they are deployed, to theorising hype as a material force with redistributive consequences.
Mapping the Imaginary Field
Wang2025-zy sets a baseline by operationalising Cave and Dihal’s framework of AI hopes and fears across UK, US, Chinese, and Indian newspapers. The contribution is comparative and quantitative: utopian/dystopian tropes recur transnationally, but their salience varies systematically with national political-economic context. This grounds the more interpretive work of the other two papers — confirming that imaginaries are patterned, nationally inflected, and discernible at scale. Yet the framework remains largely descriptive: it shows what visions circulate without strongly theorising who benefits from their circulation or how they perform work in the world.
Imaginaries as Strategic Action
Stanusch2026-ec takes up precisely that question, shifting from press coverage to industry self-representation during the November 2023 Sam Altman controversy. Here imaginaries are not just floating narratives but active devices of issuefication — instruments that determine which concerns become salient and which are rendered invisible. The paper’s identification of Longtermism, Regulatory Ambivalence, and Techno-Hagiography on LinkedIn, paired with the strategies of premediation (displacing critique into speculative futures) and preclusion (absorbing critique by self-nominating as problem-solver), gives analytical teeth to the dystopian/utopian binary mapped by Wang2025-zy. Crucially, both papers converge on a striking absence: present-day harms — environmental costs, labour, racialised bias — are systematically marginalised in dominant discourse, surfacing only in counter-imaginaries on X/Twitter.
Hype as Infrastructure
Dodds2026-df pushes the theoretical stakes furthest by reframing AI hype not as a discursive bubble but as a systemic infrastructure that mobilises resources, redistributes legitimacy, and configures futures before they materialise. Where Wang2025-zy documents imaginaries and Stanusch2026-ec shows industry deploying them, Dodds2026-df asks what hype does once embedded in professional fields — particularly journalism, where reporters occupy a contradictory dual role as hype-makers and hype-watchers. Sourcing patterns favouring tech executives, vague union agreements, and the relabelling of “automation” projects as “AI” projects all illustrate hype’s redistributive effects. The piece’s call for “breakdown and repair” responds directly to the foreclosure dynamic that Stanusch2026-ec names as preclusion: making hype’s mechanisms visible is the precondition for imagining alternative futures.
Convergences and Open Questions
Three threads tie the cluster together. First, all three papers privilege the concept of sociotechnical imaginaries (via Jasanoff and Kim) but extend it in distinct ways — comparative-quantitative, controversy-based, and infrastructural. Second, each identifies the same conspicuous absences: labour, ecology, and historically situated critique drop out of dominant discourse across newspapers (Wang2025-zy), industry platforms (Stanusch2026-ec), and journalistic sourcing (Dodds2026-df). Third, all three resist treating discourse as epiphenomenal — narratives allocate resources, legitimacy, and political possibility.
The open question across the cluster concerns agency and intervention. Wang2025-zy largely brackets it; Stanusch2026-ec locates partial resistance in platform counter-publics; Dodds2026-df articulates the most programmatic response in its breakdown-and-repair agenda. A productive next move would be to ask how the cross-national variation documented by Wang2025-zy interacts with the industry strategies catalogued by Stanusch2026-ec — whether premediation and preclusion travel differently across the geopolitical framings that Dodds2026-df identifies as central to hype’s operation.