Using Source Tracking AI to Analyze News Coverage about First Nations, Indigenous and Métis Communities

  “This paper explores the interdisciplinary, creative development of an  artificial intelligence (AI) tool designed to analyze sourcing practices in  journalism, with a focus on news coverage of Indigenous, First Nations, and  Métis communities in Canada. Rooted in theories of journalistic routines,  framing, and media representation, the tool categorizes sources into seven  key types: political, authority, expert, organization, unaffiliated, media, and  celebrity. Analysis of a corpus of articles of interest to Indigenous  communities reveals statistically significant imbalances in sourcing practices.  Political and institutional sources were overrepresented, while unaffiliated  sources, representing grassroots or lived experiences, were  underrepresented. These findings reflect persistent biases in Canadian  media’s portrayal of Indigenous communities, reinforcing institutional  narratives over diverse perspectives. While the AI tool offers a systematic  method to identify and quantify such patterns, limitations in its current  iteration temper its broader applicability. Despite these limitations, the tool  demonstrates potential for promoting accountability in journalism by  enabling newsrooms to critically assess and refine their sourcing practices.  Future iterations should address these shortcomings by incorporating more  inclusive training data, refining category definitions, and improving accuracy  for underrepresented and misclassified groups. This work underscores the  need for ethical and methodological rigour in developing AI tools to address  systemic inequities in media coverage