The administration’s aggressive lobbying for ‘more AI everywhere’ could lead to a significant deterioration of working conditions for graduate student workers and faculty, with ripple effects for students’ learning conditions.
We’ve been living with the administration’s fabricated austerity regime now for close to three years. Many students have felt the impact of larger class sizes and significantly less face time with instructors, or in a shift from labs or essays to less labour-intensive multiple-choice exams, simply because instructors don’t get enough TA hours to mentor and grade students. And even though the province recently announced a significant funding increase for Ontario universities, on the backs of low-income students, the provost clarified that he’s intent on continuing to starve the university of the resources needed to improve our teaching and learning environment.
What does austerity have to do with AI? Well, the administration may do what some schools have already done, which is to try to cut labour costs by automating aspects of teaching. We say “try” because this calculation rarely works out: automation usually just siphons money into the tech industry instead of cutting costs. This dynamic follows a playbook which well predates the emergence of generative AI: under the guise of technological inevitability, employers lay off employees, rehire them at lower wages, intensifiy workloads, and normalize precarity. All of which would pose significant strains on the quality of education at Queens, because taking away resources from educators always means taking away resources from students.
If you think that this is a faraway scenario, think again: previous rounds of bargaining suggest that the administration has been quietly trying to lay the groundwork for automating aspects of teaching. For example, in their negotiations with PSAC 901 last year, the administration tried to remove existing protections so that they could replace TAs with AI. In QUFA’s last round of bargaining, the administration tried to gain control over our intellectual property, which would’ve allowed them to use our materials as training data for courses designed and taught by generative AI software.
Through drastic cuts in TA hours, the administration also has already informally created a situation where many professors are forced to rely on AI because they wouldn’t otherwise be able to shoulder their workload. Indeed, the provost’s ‘AI advisor’—another expensive position on Queen’s already bloated higher admin salary bill—has actively encouraged QUFA members to use AI to deal with TA cuts or increases in administrative workload. The push for automation in teaching is likely going to be intensified in light of the administration’s plan to expand remote micro-credential offerings, first revealed in The Journal’s recent reporting.
What would increasing automation of teaching mean for students concretely? Because generative AI models don’t have a humanlike understanding of the words that they use, and often “hallucinate,” it would mean lower-quality teaching materials, lower-quality feedback, inaccuracies in grading, and additional student fees. At the same time, it would allow educational-tech corporations to access a massive trove of student data to monetize or use to train AI models.
Chemistry students have already experienced these effects when Stemble, an AI grading system, was adopted in first year Chemistry courses. As a PSAC 901 member reported at a recent QUFA event on workplace surveillance, the grades students received from Stemble were often wrong and contradicted what the instructor had said in class. The system also didn’t save time for TAs, who still had to comb through its output to check it for correctness. And every student in the course had to pay a mandatory fee of $90 to submit their assignments through this system—a concrete example of how, instead of saving money overall, high-tech systems tend to funnel more money to tech companies.
At the same time, it’s well documented that generative AI reproduces and amplifies social biases from its training data—including biases against female, racialized, queer, nonbinary, trans, disabled, immigrant, or first-generation students. This could result in all types of marginalized students being given lower grades across the board. Even if the information about someone’s race or gender isn’t given to the AI, it can guess these traits based on subtle patterns like word choices used in an assignment. Corporations often don’t take accountability for harmful inaccuracies in their products’ output, which means that the labour to correct them falls on students, TAs, and instructors.
This means that students would pay three times for ed tech: first, with additional fees; second, with their data; and third, with their time spent on correcting inaccuracies.
A recent survey conducted by Queen’s suggests that the administration’s push for more AI isn’t well aligned with students’ perspectives and needs. For example, students are quite concerned about how the use of generative AI impacts independent and critical thinking. With good reason: as educators, we know that learning can’t be automated (memorizing isn’t the same as understanding), and there’s emerging scientific evidence to support these concerns. The administration’s current path also doesn’t seem to align with current labour market trends.
Plans to drastically slash FAS, gutting arts, humanities, and social sciences in favour of STEM and AI sit uneasily with indications for a rising labour market demand in the humanities and social sciences, and an entry-level hiring crisis in software jobs, where technology corporations first experimented with the above-mentioned Gen AI-austerity playbook on their own employees. These misalignments are, of course, not unique to Queen’s, and in response, students at other universities are beginning to put pressure on their administrations to reconsider their investment in the AI bubble.
For the higher education sector as a whole, the result appears to be increasing stratification: as some commentators have pointed out, the introduction of AI in higher education is heading in a direction in which “small numbers of elite students will have access to a more traditional, largely tech-free liberal arts education, while everyone else has a ‘degraded, soulless form of vocational training administered by AI instructors.’” One wonders which of these two paths the current administration is pursuing for Canada’s oldest university.
QUFA has been working on these issues through the creation of a task force, co-chaired by the authors of this opinion piece, which prepared suggestions for how to address problems related to generative AI in our upcoming bargaining this spring.
To protect our University from the automation of labour and related consequences for equity, academic freedom, and privacy, we’ve proposed to strengthen our collective agreement with respect to technological change and data rights. As we move into bargaining, we hope to have more conversations with the Queen’s community—students, graduate student workers, and staff—about how AI impacts their education and work, so we can strategize together about how to keep Queen’s a thriving learning environment.
For the last couple of years, the discussion on generative AI in higher education has been largely centred on academic integrity—wrongly so, in our view, because as we’ve argued here, the issues at stake for universities are much more expansive. Often, critical views such as ours are dismissed as ‘doomerism’ or ‘luddism’. But this isn’t just an ill-informed historical understanding of what the luddism was as a labour movement and tactic for labour struggle, but completely misses the points we’ve laid out here. It’s not about being against technology. It’s about the struggle for good working conditions—and our working conditions are your learning conditions.
Norma Möllers is an Associate Professor of Sociology. Carolyn Lamb is an Assistant Professor at the School of Computing.
Tags
artificial intelligence, Opinions, QUFA
All final editorial decisions are made by the Editor(s) in Chief and/or the Managing Editor. Authors should not be contacted, targeted, or harassed under any circumstances. If you have any grievances with this article, please direct your comments to journal_editors@ams.queensu.ca.