AI literacy and Perceptions of Agency and Gatekeeping in AI News Chatbots : A focus group discussion on Hej Aftonbladet

Sammanfattning

This thesis explores how users perceive Artificial Intelligence (AI) news chatbots and how their level of AI literacy shapes these perceptions, using the Swedish chatbot Hej Aftonbladet as a case study. As AI in journalism develops from a backend newsroom tool to an audience-facing interface that communicates directly with readers, the relationship between news outlets and their audiences is changing rapidly. The study is grounded in Human-Machine Communication (HMC), which conceptualizes AI as an active communicative participant, and in AI literacy as a multidimensional skillset. It uses a qualitative case study approach, based on two focus group discussions with eight university students in Uppsala, divided by applied AI literacy. Participants interacted with the chatbot over one to two weeks before the focus group interviews were conducted. The gathered material was examined through abductive thematic analysis. Findings show that while users perceive the agency of the chatbot as machine-like, they attribute moments of higher agency to it when it appears to understand the user or the conversation. Themes such as gatekeeping became prevalent through the constraints on the chatbot such as suggested prompts, answer formats and topic refusals observed by the users. Some users perceive gatekeeping as deliberate editorial choices, while others attributed restrictions to a more diffuse and expressed more skepticism towards the chatbot’s integrity. AI literacy does not increase or decrease these perceptions linearly but changes where users notice signs of gatekeeping and AI agency. The thesis concludes with a discussion on the implications for news organizations, in terms of how the same design choice can perceived as both a responsible editorial decision and a source of suspicion supporting the literacy paradox in the setting with a discussion on what these findings mean for news organizations.