Andrew McConnell, a course director and instructor in York University’s Faculty of Education, is using AI to preserve Indigenous languages and cultural knowledge.
McConnell is an Anishinaabe of Nipissing First Nation who is on secondment from his role as an Indigenous education coordinator at the York Region District School Board. At York University, he is exploring how to develop a large language model (LLM) – a type of AI trained on vast amounts of text data to understand and generate human-like language – with an information base about Indigenous languages that is produced entirely by members of Indigenous communities.
He wants to ensure the tools can be collaboratively developed and owned by Indigenous community members and respect Indigenous ways of knowledge sharing.
“A major concern in the Indigenous community is language loss – we don’t have many first-language speakers left to teach the next generation,” says McConnell, who is also pursuing a PhD in digital media at the Lassonde School of Engineering. “AI has been designed to be useful for the dominant society and particularly for corporate culture, which tends to be at odds with Indigenous culture. We need a distinctly Indigenous process for the development of AI.”
McConnell is experimenting with this idea using York’s Automated University Response Assistant (YU AURA) research GPT for building custom AI solutions. He uploaded a variety of scholarly materials produced exclusively by Indigenous people. He also programmed the tool to respond in ways that respect Indigenous people and – in keeping with Anishinaabe tradition and to avoid the problem of false AI-generated output – to refrain from responding when it lacks sufficient information.
“As proof of concept, it worked – and there’s where I stopped, because I hadn’t gotten permission from any of the scholars to add their work, a courtesy that is essential in our culture,” McConnell says. “We need an Indigenous council – people in the community who can think through the next steps together.”
To this end, McConnell is participating in Abundant Intelligences, a global Indigenous-led research program focused on designing and developing AI technologies based on Indigenous knowledge systems and practices. Through the program’s Toronto pod, which includes York University and OCAD University, he connects with Indigenous knowledge-holders, cultural practitioners and language keepers to conceptualize and prototype new AI practices and tools in partnership with scientists, artists and engineers.
The next step in developing the tool involved testing the capacity of AI chatbots – such as ChatGPT’s Indigenous Language Supporter, Claude and Microsoft Copilot – to interact with him in Anishinaabemowin, the main language of Anishinaabe people. He found all AI models tested have significant limitations on how accurately they could communicate in the language.
“These LLMs process words through a linguistic model that is Anglo colonial and not properly trained on a good Ojibway dictionary and robust grammar rules,” he says.
As the project continues to unfold, McConnell and his collaborators prioritize the First Nations principles of ownership, control, access and possession to ensure the Indigenous community maintains sovereignty over all development processes, computer hardware and resulting data. He says this approach could lead to the establishment of an independent server farm on an Indigenous reserve, but he is still exploring the idea.
Throughout the process, he says, the team is taking a slow-and-steady approach so it can carefully consider the ethical dimensions of AI, including its intensive use of energy that contributes to carbon emissions.
“We want to approach this in a way that is sustainable and is respectful to our relations with others, which is core to our way of being and seeing the world, and to our mindset of how to be a good human being.”
With files from Sharon Aschaiek