Abstract Artificial intelligence (AI) systems for decision-support are being introduced across multiple domains. One precondition for successful implementation is that those affected by the decisions trust the AI system. In the literature, many drivers of trust in AI have been proposed. In this study, we investigate the relative importance of six potential drivers of trust across three different decision-making domains, i.e., finance, legal, and healthcare in the Danish population. Based on an initial literature review and focus groups we designed a discrete choice study using conjoint analysis. Respondents were presented with a decision involving an AI system and asked to choose which of three possible AI systems they would trust the most, each system being characterized by a combination of levels of six attributes. All participants were presented with six choice situations. The decisions presented in the three domains were chosen to have approximately equal importance based on the focus group discussions. The three decisions were decision-making about a loan for a house purchase, decision-making about guilt in a case of assault, and diagnosis of skin cancer in primary care. Three samples of approximately 300 respondents each, representative of the Danish adult population were recruited through the panel provider Wilke. The main findings are that the most important drivers of trust differ between domains, with professional responsibility for the final decision being most important in finance and legal contexts, and precision being most important in the healthcare domain. Across all three domains, the absence of a particular trust-making feature has a strong negative effect, stronger than the presence of the same feature.

Abstract AI is developing rapidly, as are concerns about the environmental impact of its training and deployment. Studies about the environmental sustainability of AI have begun to emerge in the past five years, stressing the need for critical reflection on the discursive underbelly of this emerging scholarship. For example, how do authors frame the problem of the environmental impact of AI? Are there any ethical reflections accompanying their reporting, and if so, which ethical theories and principles guide normative considerations about the environmental impact of AI? In this study, we conduct a scoping literature review on (1) how authors refer to and frame the problem of the environmental impact of AI systems, (2) who is ascribed responsibility for mitigating said impact, (3) what mitigative measures are proposed, and (4) what normative commitments justify such prescriptive normative statements. Our findings indicate that most literature on the topic is concentrated in computer science, engineering, and natural sciences, and the humanities are mostly absent. This results in a dominance of technofix attitudes towards the problem, and a narrow and limited engagement with ethical principles and theories. As such, we argue for more interdisciplinary work on the environmental sustainability of AI, leading to more diverse solutions and more explicit and pluralistic ethical starting points, grounded, for example, in relational and more-than-human ethics. The findings of this review highlight gaps in the literature and opportunities for developers, social scientists, and AI ethicists for more effective and diverse responses to AI’s environmental impact.

Abstract This article offers a critical review of how mainstream cinema constructs futures of AI and normalises modes of governance for those futures. Drawing on scholarship on AI narratives, socio-technical imaginaries, and film analysis, and using an illustrative and replicable corpus of nine feature films (selected from IMDb with at least 200,000 votes), the article identifies three cross-cutting patterns. First, the dystopian/cyberpunk canon in live action is linked to technopessimistic social imaginaries and has received sustained attention. Second, hybrid futures shift the grammar of threat towards coexistence, especially when AI is integrated into care/management plots and diegetic violence is contained. Third, the rebellious role of AI is associated with representations of AI as independent artificial general intelligence (AGI) and high thresholds of violence, reinforcing socio-technical imaginaries characterised by fear of machine rebellion. Overall, the article provides a sociological reading that links reception, futures, and roles, with implications for AI governance and its social implementation.

This study evaluates conversational large language models (LLMs) as pedagogical brainstorming tools for historical source code analysis through structured prompt-based approaches adapted from Critical Code Studies (CCS). The research tests whether conversational interfaces like ChatGPT-4o can support initial exploration of complex historical codebases by adapting CCS perspectives into conversational prompt formats. The dual-prompt evaluation separates technical parsing from interpretive reasoning, assessing how effectively conversational interfaces extract structural information while generating preliminary interpretive hypotheses. Using the Apple Lisa source code as a case study, this analysis documents both pedagogical utility and systematic limitations. The findings demonstrate that while conversational LLMs can preserve developer annotations, parse visual artifacts such as ASCII diagrams, and generate educationally valuable insights for approaching unfamiliar programming languages, the conversational user interface significantly constrains systematic analytical capabilities. Through three case studies examining ASCII typography, architectural diagrams, and interface implementations, the analysis illustrates how conversational interfaces function as structured brainstorming partners while necessitating rigorous validation against primary sources. This evaluation contributes to understanding AI’s pedagogical role in digital humanities by positioning conversational LLMs as question-framing tools rather than interpretive authorities, while identifying systematic implementation requirements necessary for rigorous computational approaches to historical software analysis.

Abstract This article examines generative AI models as sociocultural actors, focusing on how they reproduce and constrain notions of authorship and identity within the contemporary U.S. literary field. Through a simulation of 101 “AI authors” modeled on real-world counterparts, we analyze how language models articulate literary distinction across race, gender, and publishing contexts. Our findings show that generative outputs are dominated by a reductive black–white binary that both flattens intra-group diversity and marginalizes alternative axes of distinction. Comparative analysis with human-authored works highlights the homogenizing tendencies of AI systems, suggesting that the model operationalizes identity as a limited categorical variable rather than as a dynamic cultural practice. Prompt-based interventions reduce racial fixation but often efface race altogether, raising critical questions about the limits of technical fixes for bias. We argue that these dynamics illuminate how AI systems participate in cultural production by encoding and reshaping social categories of difference. This article contributes to debates on AI agency, algorithmic bias, and cultural equity, and points toward the need for design practices attentive to identity, representation, and creative diversity.

Healthcare systems often delegate surgical consent-seeking to members of the treating team other than the surgeon (e.g., junior doctors in the UK and Australia). Yet, little is known about public attitudes toward this practice compared to emerging AI-supported options. This first large-scale empirical study examines how laypeople evaluate the validity and liability risks of using an AI-supported surgical consent system (Consent-GPT). We randomly assigned 376 UK participants (demographically representative for age, ethnicity, and gender) to evaluate identical transcripts of surgical consent interviews framed as being conducted by either Consent-GPT, a junior doctor, or the treating surgeon. Participants broadly agreed that AI-supported consent was valid (87.6% agreement), but rated it significantly lower than consent sought solely by human clinicians (treating surgeon: 97.6% agreement; junior doctor: 96.2%). Participants expressed substantially lower satisfaction with AI-supported consent compared to human-only processes (Consent-GPT: 59.5% satisfied; treating surgeon 96.8%; junior doctor: 93.1%), despite identical consent interactions (i.e., the same informational content and display format). Regarding justification to sue the hospital following a complication, participants were slightly more inclined to support legal action in response to AI-supported consent than human-only consent. However, the strongest predictor was proper risk disclosure, not the consent-seeking agent. As AI integration in healthcare accelerates, these results highlight critical considerations for implementation strategies, suggesting that a hybrid approach to consent delegation that leverages AI’s information sharing capabilities while preserving meaningful human engagement may be more acceptable to patients than an otherwise identical process with relatively less human-to-human interaction.