Artificial intelligence is being used in university classes. How it's being used matters, say profs

Cassidy Chisholm · CBC News

· Posted: Sep 17, 2023 5:00 AM EDT | Last Updated: 2 hours ago

As artificial quality becomes much communal successful assemblage classrooms, immoderate professors are weighing the benefits — and downsides — of students utilizing it for probe projects.

One Halifax prof precocious issued a argumentation saying he doesn't let the usage of artificial intelligence, similar large-language exemplary ChatGPT, successful his classes. 

Large-language models are a benignant of artificial quality that tin reply requests for translations, summaries and different contented based on immense amounts of written accusation fed into the systems.

Ajay Parasram, who teaches past and planetary improvement studies astatine Dalhousie University, says artificial quality is simply a instrumentality that whitethorn marque the process of find easier, but it whitethorn besides exclude important information.

Artificial quality is known to person biases, and Parasram says that limits hunt results to lone peer-reviewed articles and top-tier journals that reenforce definite intelligence discourse.

"It's each the things that you spot successful the backgrounds, successful the margins, each the things that you weren't looking for — that's portion of the joyousness and excitement of learning," Parasram told CBC Radio's Mainstreet Halifax connected Wednesday.

"And I fearfulness that students who are learning to bash probe are getting up of themselves by conscionable looking for the close answer.''

Parasram said if students usage artificial quality for their research, it would beryllium a disservice to their ain learning experience. 

He said helium adopted this regularisation arsenic a mode to promote students to "take power of their ain learning."

"I deliberation that we're not yet astatine the spot wherever we tin spot successful exertion to fundamentally region overmuch of the process of captious thinking,"  helium said.

"Because I consciousness similar we've fought truthful hard to adjacent get captious reasoning astatine the centre of a batch of our learning and we can't conscionable jettison it now."

Sharon Lauricella, a prof of connection and integer media studies astatine Ontario Tech University, has taken a antithetic approach.

She encourages her students to usage large-language models successful her classes, due to the fact that it's disposable and they should cognize however to usage them.

"I springiness them the analogy that … these instruments are a tool, but if I springiness idiosyncratic a chainsaw and I don't thatch them however to usage it, someone's going to get hurt, right?" Lauricella said.

"And the aforesaid happening applies to this technology. People request to cognize however to usage it decently and past they tin usage it safely."

Lauricella said large-language models similar ChatGPT often person outdated oregon incorrect information, truthful she teaches her students to see what accusation it whitethorn not beryllium showing them and however to re-ask questions to get the answers they want.

She said it tin besides beryllium utilized to summarize information, assistance signifier ideas and cheque grammar connected assignments — what she calls the bookends of a probe project.

But she agrees with Parasram that large-language models weren't designed to supply credible research. Critical reasoning is inactive required.

"It tin prevention america time, but successful presumption of doing superior research, that is conscionable a cardinal accomplishment that each students request to larn however to do, that [ChatGPT] conscionable isn't susceptible of," she said.

Lauricella said her students tin usage a large-language exemplary to bash inheritance research, but she wants their ain enactment to beryllium unsocial and reflective.

She said if her students bash usage artificial quality successful their work, they are required to disclose what they utilized it for and see that astatine the extremity of their assignment.

"That, I think, is cardinal that students person to disclose, they person to beryllium transparent astir utilizing this," she said.

"So I deliberation that's beauteous important. We've got to springiness recognition wherever recognition is owed adjacent if the instrumentality is not sentient."

Still, Parasram said it's important to punctual students that they tin marque mistakes.

"There's truthful overmuch unit connected them to get the close reply and not capable accidental for the process of intelligence find and joy," helium said.

"And I deliberation astir assemblage successful my classrooms arsenic an accidental for students to person the grooming wheels on, you know? Make mistakes. Go retired determination and effort a process, and what ends up happening is that students nutrient perfectly astonishing work."