Everywhere, university humanities departments are diving into generative AI to instrumentalise the very critical thinking that warns against such a move. Australia leads the way!
The rapid spread of the Internet and the consumer culture of generative AI—the most well-known being ChatGPT—is increasingly mirrored in our institutions, particularly universities, which do not want to be left behind by any significant change. AI and even machine learning have been around for decades, but it is the large language model (LLM) that has driven much of the current interest, thus prompting universities to act. The business model is like Uber’s, insofar as OpenAI entered a market, created significant hype, and then waited for everyone to respond. The interest from the public, including students, has been substantial, with many fascinated by the quality of the outputs, but also by claims about the future benefits. All the other major tech companies have followed, flooding the market with new applications and devices. Australian universities are now scrambling to incorporate AI into their curricula and research, nudged by tech companies seeking new markets and the Federal government, which wants Australia to be a leader in AI use. The latter is enforced through such mechanisms as TEQSA (Tertiary Education Quality and Standards Agency) audits. The problem with such auditing systems is that they do not address fundamental questions about the university’s role and its disciplinary traditions. Instead, the TEQSA discourse reiterates hackneyed lines about responsible use, privacy and integrity, which are then incorporated into university guidelines and planning documents with very little variation. Such discourse assumes that generative artificial intelligence (Gen AI) is just another information tool that can be managed through a set of external ethical principles and procedures. Despite this rhetoric, the reason why universities and the Federal government are looking for ways to incorporate Gen AI is that they accept the external market-driven logic that AI will fundamentally change everything we do, and by accepting the industry’s logic, they effectively make it happen. It has never been about an internal disciplinary-specific demand to use the new tools. AI has been used for many years in many disciplines, from music composition to neural networks in cognitive science, without much of a response. It is only since the rise of the large publicly available commercial systems, mostly from Silicon Valley, and the snowballing promotion that there has been a push to extensively change how universities operate. Now, there is a desperate search for applications that are suddenly deemed necessary to teaching and research, irrespective of the discipline.
Universities are adapting to pressures of an information marketplace rather than drawing upon their own disciplinary traditions to produce new knowledge—a fundamental feature of research—or even starting with the most basic questions about what we want our students to know. Knowing how to use a commercial product, something that will rapidly change, differs significantly from knowing the rudiments and principles of a discipline. For example, the deployment of AI in IT should differ significantly from its use in music, as each discipline has a different object of analysis and a different history. Few would argue that AI should replace human performance in music, even though the systems can certainly read scores and produce musical outputs. Yet, in many other disciplines that are equally human-centric, from philosophy through to sociology, with methods established over hundreds and even thousands of years, the increasingly popular institutional line is that both students and staff should be developing strategies to use Gen AI. Now, I am not arguing that they shouldn’t, but such use should not comprise a knee-jerk response to a techno-capitalist imperative. Any questions about use should be formulated through the disciplinary tradition. Indeed, Gen AI’s success has largely come about through the erasure of epistemological differences in favour of a universal sign system managed by constantly evolving algorithms. Unlike academic research, which has progressed through laying bare the very processes of reasoning and its historical conditions, Gen AI disguises its own reasoning and asks us to respond to its infinitely producible outputs.
In the universities, we are constantly looking for ways to adopt Gen AI to demonstrate a commitment to the technologically-imagined future. Since 2022‒23, many academics have started incorporating AI into their teaching to prove their worthiness in helping universities satisfy the government audits. The most common exercise in the social sciences and humanities during this period involved students appraising the outputs of LLMs against commonly accepted knowledge in a discipline. A typical exercise would assert the value of human knowledge while still accepting the imperative to use the technology, for example, in questions like the following: does the LLM’s definition of dialectical materialism match what is said in Marx’s text? Depending on the year level, such exercises can be pointless. Will a student in their first year have the capacity to make such a judgment about the quality of a definition when they have not yet sufficiently studied Marx? Would it be more useful to wait until a student has studied the original texts over a number of years before they start judging a technology that is designed to mimic human thought? Even if the students are capable of making informed judgements simply because they are human, we still have to ask what they have learnt by appraising the technology. The most likely result is that they have learnt something about the affordances of Gen AI. Great! But is the subject or course designed to teach AI’s affordances or social theory? Such an approach says more about a desperation to find some use for the technology than about furthering the understanding of the discipline. In any case, it supports the business model of the tech giants, for, regardless of the degree of scepticism, Gen AI becomes central to university activity.
Naturally, there are the early adopters and boosters, who, in any institution, desperately seek uses for an emerging technology so they can place themselves at the vanguard of the paradigmatic shift. Careers will be built on the promotion of new technology, irrespective of whether it aligns with existing university principles. The most common arguments in favour of the adoption are: students are already using technology, so we just have to teach them responsible use, or that we must prepare students for an AI-driven workplace. These arguments are not without merit, but unfortunately, they have not formed the basis of clear planning. The question is not really about use or non-use, rather the focus should be on when and how the technology is incorporated into a curriculum. Depending on the discipline, we could easily run dedicated units at the end or beginning of a degree to address the ethics and pragmatics of AI use. This seems more sensible than asking every unit coordinator to show how they will be using it in their course. Demonstrating that Gen AI is supplementary to the development of human thought accords priority to the discipline over workplace or commercial considerations. We are employed to train students in a discipline and to help them develop a sufficient level of expertise in whatever they have chosen to study. The first question should always be: will the adoption of AI increase a student’s level of expertise? An ancillary question in the humanities would be: Will the use of Gen AI further the maintenance and progression of human culture? Unfortunately, these questions rarely arise because the instrumental reasoning of governments, tech companies and university administrations holds sway. Of course, many academics have expressed concern, especially with regard to academic integrity and the future of the workforce, but also with regard to the environmental and social impact of Gen AI. The problem, however, is that universities have already accepted an instrumentalist rationale before even considering the particular way in which Gen AI will be used. Even with the clearly provable claims that it uses a lot more energy and water than existing information technologies, I have yet to hear anyone make the ethical claim that we should restrict the use of Gen AI to absolutely necessary tasks—tasks that definitely further research or learning. We can all have small bins to apparently cut down on office waste, but there are no similar restrictions on the widespread adoption of Gen AI.
Despite the apparent novelty of AI, it is just another example of the exteriorisation and instrumentalisation of knowledge, a central theme in the contemporary university. Regardless of the effects of Gen AI, this instrumentalisation already shapes how the contemporary university operates. Research is increasingly characterised by metrics ranging from citations to journal rankings. Even collaboration—what should be a peculiarly human inclination—is now encouraged because it contributes to university rankings and the quest for funds. Notably, most of the research factors are derived from education technology companies, publishing conglomerates and tech giants such as Google. As with the adoption of Gen AI, we are working from external categories and measures rather than from within the internal logic of a discipline. This is demonstrated in one of the most typical responses to an academic’s statement that they have a book coming out: who is the publisher? This question usually precedes any questions about the topic because the name of the publisher can also contribute to the university ranking systems. The aim of the university is to perform well within an audit culture, and this often diverges from disciplinary aims and its own internal process of critique. Academics will definitely still produce good work, but the technological tail is increasingly wagging the dog.
In a similar manner, teaching has been reduced to rubrics, learning outcomes, student evaluation, etc. The typical assessment task in a university involves a set of instructions and a rubric that states how the task will be graded—how the markers will distinguish between higher and lower grades. This mechanism ostensibly arose from a need to ensure consistency across the marking cohorts; an external framework that serves a legislative function with regard to the administration of grades. The expansion of the external architecture of assessment must be contrasted with the desire to know, underpinning human learning and culture. Like academics who are beholden to ranking systems, students are increasingly writing to meet the telos of the rubric. The question for the student becomes how to produce something that satisfies the external criteria of assessment by reproducing and mimicking existing models of success. However, the role of assessment should be to provide a forum for coming to terms with the problems and procedures of a discipline. For example, if the rubric states that the assignment must have ten references, the references will be inserted irrespective of whether they properly contribute to the argument. For many years, I have seen students just use references as a form of garnish to increase the ostensible value of their texts. They are objects to be cited, not read. This focus on the process of adjudication and the external form of academic writing shifts all attention to the output—the finished piece of work. From an efficiency perspective, the faster one arrives at the output, the better. Why waste time struggling with the specific problems and the history of a discipline when all of this will be reduced to a single mark? This issue is also central to why students use Gen AI to produce work. It produces documents that resemble typical university assignments, from essays to lab reports or even mathematical proofs. The user simply has to write a prompt, and the particular platform produces something that resembles an academic text, and you can even specify things like the number of references, the length of paragraphs, the use of headings and so on. Some of the material may be fallacious and some accurate, but what is most important is that it resembles the required output for assessment. Thinking here is outsourced to a technology, but does this really matter in a system in which all value is outsourced and externalised?
The promotion of Gen AI is typical of this process, for administrators want to claim that graduates are AI-ready and can ethically use the systems without considering how technology itself shapes thought. They assume a productivity model in which doing more at a faster rate is the ultimate aim, something that does not easily align with the acquisition of human knowledge. Students themselves are caught up in a similar model, for the idea is to produce assignments that meet external guidelines as quickly as possible. All that matters in both cases is that the external measures of value are satisfied from a TEQSA audit, through citation counts, to a student’s grade. Gen AI is just another symptom of a culture increasingly concerned with the simulation of knowledge and the simulation of value. It allows the technology to produce information with increasingly less human intervention. Why let the slowness of human thinking and learning impede a techno-instrumentalist fantasy of automated knowledge production?