Suspicious Indigenous language books spark concerns about AI authors.
APTN News: Thomas Bear claims to be a distinguished Plains Cree language expert from Canada and the author of several books.
His bio celebrates his work in the community, sharing his expertise through workshops and seminars.
But there’s suspicion that Bear isn’t who he claims to be–or if he even exists.
A series of Indigenous language books on Amazon with near-identical covers are being questioned on social media. The books, ranging from Plains Cree to Ojibwe to Blackfoot, retail between $45 to 65 CAD.
Outside the Amazon marketplace, the self-professed language experts lack a traceable digital footprint. The International Standard Book Numbers (ISBNs) are not formally registered.
After purchasing an Ojibwe verbs book, Facebook user Laurie McLeod-Shabogesic claimed it was “riddled with inaccurate translations”.
Some angry reviewers and commenters allege the authors used generative artificial intelligence (AI).
Reading between the lines
To put Bear’s book, The Most Frequently Used Plains Cree Adjectives to the test, APTN News reached out to Kevin Lewis, a Plains Cree language instructor from the Ministikwan Lake Cree Nation in Treaty 6.
Within seconds of skimming the sample text, Lewis sensed something was off. Most words were either incorrectly translated or completely made up.
‘nikan’, which Lewis says is the Plains Cree word for “in the front” was translated as “all” in Bear’s book. Other ‘words’, such as ‘âkayâs and ‘kostaw’ were unrecognizable to him.
A few other things stood out to Lewis. Unlike in Bear’s text, he said there are no capital letters in the standardized Cree that instructors have developed over decades. Despite the book’s offerings, Lewis said adjectives don’t fully exist in the polysynthetic language.
Instead, Cree uses what Lewis calls “word sentences”.
“It’s a verb-based language,” he said. “It’ll be like ‘I in the past moved by running I did’ and then that would be one sentence. And then in English, with the English mind, you would look at it and say, woah, that’s just one word.”
Lewis had never heard of a Plains Cree expert named Thomas Bear. He added that it’s rare to come across an unfamiliar name in the tight-knit circle of Plains Cree instructors.
“Usually, you kind of know who’s who as far as, you know who’s publishing, who’s working and who’s researching, just because we’ve been getting together for a very long time.”
APTN wasn’t able to contact the author Thomas Bear unless we purchased the book. Amazon wouldn’t confirm who was selling it.
The rise of AI-generated content
Professor David Gerhard heads the University of Manitoba’s Department of Computer Science.
Over the past few years, he said AI-generated books have become more commonplace on websites, like Amazon, which allow self-publishing.
“Generative AI has the flavour of real language, and so, depending on the way you prompt it, it’s possible to make it so that it sounds realistic,” Gerhard said during an interview with APTN.
While the text can look legitimate, it’s often unreliable.
“These generative tools, especially when they are working in domains where they don’t have a lot of information, can sometimes just make up an answer, rather than say they don’t know,” he said.
In early October, Gerhard came across a reportedly AI-generated Manitoba mushroom foraging book that left out potentially lethal details.
“It was just wrong,” he said. “It had some examples of ‘here’s how to identify this mushroom’ that were incorrect and dangerous because some mushrooms if you’re not careful with them, can really hurt you.”
APTN plugged the introduction of Bear’s Plains Cree adjectives book into three AI detectors.
While the tools are not guaranteed to be accurate, all three suggested the excerpt was 100 per cent AI-generated.
Gerhard notes a few other warning signs, including little to no reviews, fake author profiles and a non-existent ISBN.
“You can do a search for these authors and find that they’re not real people. But the big tell is that the ISBN, the international book number, is just made up,” he said.
At the time of writing, Bear’s Plains Cree adjectives books had three reviews, all accusing the publication of being a scam.
“Cree really doesn’t have adjectives. Is this book an AI production – along with its two companion volumes? I hope innocent learners don’t waste money on this!” said one review.
The ISBN listed garnered no results from a search tool. APTN could not find any digital record of a Plains Cree language instructor named Thomas Bear.
A call for more Indigenous language programming
Whether or not the language books are AI-generated, there’s a human side to the story; someone, somewhere, uploaded the books to the marketplace.
However, Gerhard said sellers face few–if any–consequences for uploading inaccurate or AI-generated resources.
“I don’t think there’s any way that we can police this. I don’t there are any consequences we can bring on the people doing this,” Gerhard said. “Instead, what we can do is educate the public and let them know that these things are out there and to be careful of the sources of the information that you get.”
Still, Lewis worries the dubious language books could mistakenly fall into the hands of well-intentioned beginners.
“There’s a lot of people that really want to learn their languages, the Indigenous languages, and then it’s just a shame that somebody sort of capitalized it in that way,” he said.
On the other hand, he believes their existence underscores the need for more Indigenous language programs.
“We need them all over the place because there are so many people that are living in the cities, in the towns, not only in the First Nations or the settlements, the Métis communities. We’re everywhere now,” he said. “If we can sort of get online and start teaching these languages, then I think that’s probably the best way to do it.”
Tim Gillman, an Amazon spokesperson issued the following statement in an email to APTN:
“Amazon is committed to providing the best possible shopping, reading, and publishing experience for authors and customers,” he said. “We have content guidelines governing which books can be listed for sale, whether AI-generated or not. We invest significant time and resources to ensure our guidelines are followed, and when we identify titles that don’t comply with those guidelines, we take appropriate action such as removing non-compliant books and suspending publisher accounts to prevent repeated abuse.”
Amazon’s Kindle Direct Publishing content guidelines, which govern books published on Amazon, require that authors disclose AI-generated, but not AI-assisted, content.
The guidelines define AI-generated content as “ text, images, or translations created by an AI-based tool”, and AI-assisted content as content created by the author but edited, refined, error-checked or otherwise improved with an AI-based tool.
AI-assisted content also includes using an AI-based tool to brainstorm and generate ideas to inspire an author’s original content.
After APTN contacted Amazon, The Most Frequently Used Plains Cree Adjectives was removed from the marketplace, along with several other titles.
However, a handful of similar Indigenous language books remain for sale.
Shortly after the Ojibwe language series under the author name Noah Whitecloud generated outrage on social media, it was listed as “currently unavailable” on Amazon.
The Ojibwe series has since resurfaced on the website under new listings.