AI writing tools won’t fix academia’s language bias problem

Last week I received a reviewer’s comment on my (rejected) paper. “SStay away from trendy progressive terms like ‘live experience’, ‘interrogation’, etc,” it recommends, adding: “Just talk like a human.”

This comment comes from a reviewer at a leading US psychology journal, where editors are expected to protect the author from such nasty remarks. I don’t mind admitting that I was offended by the last comment, even if most of the editors and reviewers are English speakers who are less sensitive to language-related issues. language.

With so many hurtful comments, it’s no surprise that online courses that teach beginning researchers how to creatively use AI for academic writing are now oversubscribed.

The promise of these language improvement tools is appealing to non-native scholars for two reasons: first, because they are cheaper than language testing services. At Wiley, for example, a standard English editing service for an 8,000-word article costs around $1,000 (£790). By comparison, a full-access subscription to AI-based learning tools costs $25 a month.

Second, AI bots can be easily trained to write in the style of target journal readers, even in the researcher’s own writing style.

That gives me, as a non-native researcher, new possibilities: Now, I don’t just have the option to use the language software program (or annoy my friends). native English speakers) but can also train AI bots to correct their grammar errors. What’s not to like?

If you think innovative AI will democratize access to science, you’re wrong. In fact, these approaches perpetuate the problem of linguistic discrimination in academia by placing the burden on individual researchers.

Yes, the basic version of ChatGPT is free, but to get higher quality output you need to subscribe. Bots speed up writing time but take time to edit text and check for inaccuracies. Thus, non-native researchers still have to invest more time and more money in correcting their papers than their native counterparts.

A recent publication in Plos One, which went viral on academic Twitter, addressed this issue. The study, which surveyed 908 environmental science researchers from different countries, found that academics using English as a second language must spend more time and cognitive effort to complete the tasks. scientific activities. About a third of early-career scientists turned down the opportunity to attend the conference because of language barriers. The additional administrative load and mental burden are greatest for young scientists and lead to discrimination in some contexts.

One way to solve the problem is to provide more resources to non-native speakers. After transferring to a university in Norway, I discovered that Norwegian universities provide a budget specifically for language testing services to support their researchers. Even doctoral students have their own budget allocated for English testing. Language testing is a task assigned to language experts, causing researchers to focus only on what they write, not how they write it.

Norwegian researchers can publish their papers using language services, but this does not necessarily make them better writers in the long run. The important thinking process when writing an article should always be left to the person. In her upcoming book Who wrote this?, Naomi Baron emphasizes that the value of AI is in enhancing writing, not automating it. When people are allowed to write in their native language, they come up with distinctive stylistic expressions, metaphors, and unique features that often enrich writing styles.

When rich countries pay for language services to their researchers, they risk increasing the problem of language discrimination. Globally, it puts other non-Indigenous researchers at the back of the scientific queue. To level the playing field, we need to change the context. As all social justice researchers know, language discrimination is not about changing individual skills or practices; it requires dismantling the systemic structures that maintain inequality.

My suggestion is that we all agree to use AI to tackle language discrimination. What if all higher education staff were allowed to write in their preferred language and dedicated bots could translate it for readers? What if the millions of dollars spent in open access publishing were invested by academic journals in creating AI bots that aid non-native scholars?

Imagine all of the writing and reading of automatically translated academic articles from any language, formatted in the style of the target journal. There may be specialized bots for authors, reviewers, and editors.

With more and more non-natives using AI bots for their scientific writing, we cannot allow further linguistic discrimination in academia. Crude remarks from my reviewer that I “do not sound like a human” may eventually prove useful, but only if they help academia consider unacceptable hierarchies the academic publishing industry currently relies on.

Natalia Kucirkova is a professor of childhood and development at Stavanger University in Norway, and a professor of reading and child development at the Open University.