Speaking the news How conversational agents influence trust and issue salience

This dissertation studies the role of conversational agents (CAs), such as Google Assistant, as news recommender systems and their impact on public issue salience and trust in recommender systems. Chapter 1 establishes the research framework, emphasizing the influence of CAs as information gatekeepers and highlighting the research gap in understanding CAs’ role as news recommenders. Chapter 2 examines the societal implications of news recommendations provided by CAs, focusing on how trust in these agents shapes public perceptions of important issues (i.e., issue salience). It compares the effects of CAs with other news sources, such as social media and traditional websites, and evaluates the moderating role of general media trust and channel-specific trust. Chapter 3 explores how the interface—whether a CA or a traditional website—impacts trust in news recommenders, evaluating specifically the influence of CAs’ dialogic interactions. Chapter 4 investigates the effects of global and local explanation types across different modalities (audio and text), considering the regulatory context and its impact on user trust. Chapter 5 integrates findings from the previous chapters, concluding that CAs, due to their uniqueness, can potentially influence issue salience. Additionally, CAs can negatively influence trust in news recommenders through factors such as perceived enjoyment. The dissertation also identifies perceived personalization and privacy concerns as important determinants of trust. However, the modality or the dialogic interaction with CAs does not significantly affect these factors. As CAs evolve, future research should further investigate trust in CAs, mainly focusing on how different dialogic features (e.g., content) can affect user enjoyment.