Edmonton-based Indigenous AI project selected to participate in MIT program

Shani Gwin is the CEO and founder of an Edmonton based AI company - wâsikan kisewâtisiwin. Photo supplied - Jeremy Appel, Local Journalism Initiative Reporter

An Edmonton-based Indigenous-led startup that is developing artificial intelligence tools to assist in identifying anti-Indigenous bias has been selected to participate in a Massachusetts Institute of Technology support program.

wâsikan kisewâtisiwin, which means kind electricity in Cree, is working with Amii, the Alberta Machine Intelligence Institute, to develop two AI tools — the first monitors anti-Indigenous bias and hate speech on social media, and the other is a plug-in that helps correct bias and racism against Indigenous Peoples in writing.

wâsikan kisewâtisiwin was one of thirty successful applicants to the MIT Solve initiative out of more than 2,200 international applicants.

CEO and founder Shani Gwin made the successful pitch to an audience of more than 200 leading philanthropists and investors in New York City in September.

Gwin told Alberta Native News that she was inspired to create these tools by the amount of anti-Indigenous hatred that goes unchecked on social media and the conviction that Indigenous Peoples shouldn’t be burdened with constantly correcting falsehoods.

In addition to her work in AI, Gwin runs a social media communications firm, pipikwan pêhtâkwan, which acquainted her with the “emotional labour” it takes to “educate all the non-Indigenous people about what is problematic.”

While AI has the potential to make it easier to correct anti-Indigenous biases without retraumatizing Indigenous people each time, Gwin said it’s important for there to be AI that is “informed and powered by Indigenous people.”

“AI is being built, like every other system that’s been built to date on these lands, by the white, able bodied, cis male and I just thought, ‘Wow, Indigenous people have to get involved, because this is something that they’re saying is going to touch every aspect of our lives, and if we aren’t involved, it’s really going to harm us,’” she explained.

At this point, wâsikan kisewâtisiwin exists in prototype form as a large language model. The ultimate goal is to have it work as an app like Grammarly, which users can simply plug into their web browser.

“We have a little baby and we’re teaching the baby, so we have to give it instructions. We’re giving it a set of values, and then we’re teaching it about Indigenous people,” said Gwin. “It’s actually about just sharing the history and the context and the bias that people don’t actually even see in their writing.”

A common example is how non-Indigenous writers sometimes use possessives to refer to Indigenous Peoples, which is rooted in a history of Indigenous people “being told that we are wards of the state,” she noted.

Rather than have the tool just correct errors, Gwin plans to have it ask writers questions. If, for example, somebody quotes a teaching from an Elder, it could ask whether the writer has inquired about whether that nation has any protocols for relaying Elders’ teachings.

“It will prompt the person to do their own work too. It’s not going to do everything for them, so there’s lots we still have to do,” she said.

As part of its participation in the MIT Solve program, Gwin’s firm has received a $10,000 grant from MIT Solve and has been invited to participate in a nine-month support program.

“Now we work with them and are part of their network. We get introduced to other people in this field, and we just keep working on developing the tool,” she said.

“There’s other grants and other opportunities to potentially win more money or continue growing and scaling the company.”

The next step is “to just start providing an immense amount of information into the database,” Gwin added.

“What we’re hoping to do in the next few months is to go into the community, start engaging folks, seeing if we can get people to get more involved with what we’re doing,” she said.

Gwin noted that there’s been some skepticism in her community for using AI from a data security perspective, which she was able to alleviate by clearly outlining what the tool will and won’t do.

She emphasized that the tool will be used strictly to correct and question anti-Indigenous biases, not to transmit specific cultural knowledge that has been passed down orally for generations.

“We haven’t always been able to trust how our data is going to be used, and I think that’s why with this tool, I really want to be careful. I don’t want it to have cultural knowledge,” she said.

Gwin said she understands the dangers of AI from an environmental perspective.

“It’s a really resource-heavy tool. It uses a lot of water and energy, and so I think how people are using it right now is pretty irresponsible. We have to think about what this tool is actually going to be used for, and how can we use it ethically and responsibly,” she said.

“But also everyone is using it right now, and I’m worried about Indigenous people just not getting involved and not helping inform it, because we could be in a situation where we’re interacting with something in our day to day life that doesn’t know who we are or know anything about us.”