Renee Morrison

Many herald Generative-AI (Gen-AI) as game-changing in education given its capacity for immediate access to knowledge. Others warn of the potential risks of misinformation and of declining data sovereignty -particularly for Indigenous groups -when texts are artificially generated. Machine learning has repeatedly been found to narrowly represent groups already subject to historical discrimination, and this, coupled with limited Indigenous involvement in the development of Gen-AI, not to mention its amplified capacities, raises multiple concerns. This ongoing study aims to begin addressing some of these concerns by learning from First Nations academics as to the potential harms borne of synthetically created Gen-AI texts. No other studies are found which consider these unique perspectives, nor that investigate the texts generated when Gen-AI is specifically prompted (problematically) to take on the persona of a First Nations person. Using a qualitative research design, incorporating both thematic and Indigenous content analysis, our study investigates the representation of First Nations people in Gen-AI texts, and First Nations academics’ perspectives on misappropriating Gen-AI outputs. Data-wise, the study utilises ChatGPT4.0 outputs and data collected during interviews with seven First Nations academics. We consider the opportunities and risks reported by participants regarding Gen-AI texts not produced/ guided by First Nations experts. The portrayal of Indigenous knowledges and narratives has always been important, but the scaled capacities of new Gen-AI technologies and their efficiency in countless arenas encourages homogenization, meaning that the risks of harmful stereotyping and to data sovereignty are heightened. Preliminary findings and implications are discussed in this presentation.