This is another post not directly about the ERC grant writing process, but rather a consequence of me being on the receiving end of job applications (and peer review requests). I think it’s important to warn people against trusting AI too much when applying for jobs and grants - so here it is.
The AI talk is everywhere. Some people are excited about new opportunities and possibly saving time on routine tasks, others are terrified - not only about being replaced by AI, but by the prospect that humanity might lose the ability to think for itself. The latter is the refrain in educational settings. I work at a university, and sometimes it feels like all we talk about in terms of pedagogy is about AI.
I don’t mean to get into the debates on whether AI is good or bad, promising or scary. I would like to focus on a set of very practical outcomes that already change how we perceive each other in academia, how it affects the hiring and research funding processes, and how the peer review system will have to change.
I would like to communicate the idea that what looks like a shortcut often does you a disservice on the academic job market and has implications for trust and for how we in academia evaluate competence and effort.
Two things in particular made me think about writing a post about it. First, the hiring process for my ERC Starting grant and the applications that I received. Second, this post:
which, among other things, discusses the implications for the new generation of academics to grow and develop necessary professional skills.
The Erosion of Academic Trust
Based on my hiring process and the discussions surrounding academic writing which requires peer review (journal articles and grant proposals), the observation that struck me is: we are in a crisis of trust in academia, and it will challenge the review system on more than one level.
First, reviewers in journals and research funding systems experience frustration when they realize that they just agreed to read something suspiciously reminding them of the flowery and superficial AI product. We can argue all we want about how using AI in a smart way can improve our productivity, but this won’t change the fact that reviewers are now exposed to a lot of text they don’t trust (but don’t have tools to support their suspicions). More articles and more grant applications are submitted across the board and need to be sent out for review, and accessibility of GenAI may be one of the culprits in this increase (it lowers the cost of producing text and makes this process faster and more accessible to many). It is not the reviewer’s job to prove that something is AI-generated (but whose job is it?), but reading and earnestly commenting on work that you strongly suspect is not entirely human-written is not a task most of us joined academia for. Whatever we think about the unfairness of the peer review system, at its core it’s supposed to be a dialogue a meaningful engagement between the writer(s) and the reviewers. Now there’s a veil of GenAI and GenAI-suspicion.
Second, as someone at the other end of the hiring process, I noticed that I, too, experienced this mistrust and frustration, when I began to realize that all these polished motivation letters I’ve been reading are structurally similar for a reason. As I wrote earlier, quite a few people seemed to take a very logical step in the process of applying for multiple positions to secure some work in academia, and they chose to cut corners with the help of AI, feeding it with their CV and the vacancy text, asking to write a motivation letter for them. My eye soon became attuned to recurring expressions and patterns, and made me mistrust the authors of those letters quite a bit. I understand why people would do it, since I am a relatively recent participant of the cutthroat job market myself, but even my empathy did not make me take these cookie-cutter letters seriously.
The Issue of Professionalization
Learning and critical thinking are difficult. They are supposed to be difficult, as otherwise you don’t really learn things. It also means that, while you as a learner struggle, everyone else has to be patient and wait for you to figure things out (or drop out). But, at the same time, it is so normal for people to save effort and cut corners.
I noticed it also as a teacher in a class, where we encourage students to use AI in classroom. Working on a specific assignment, they are first encouraged to brainstorm without AI, but work with available resources - course readings and more traditional online search if they feel they need more information on the challenge. Then they are encouraged to engage GenAI and see how their non-AI results compare, and fill the gaps in their initial thinking, should there be any. But some students tend to take shortcuts at the first step: do something very basic, and move on to AI for more in-depth analysis almost immediately. Above all, GenAI is a temptation to have it the easy way.
However, this ease comes at the cost of not really getting it, not really mastering the material, the method, the approach. Without proper personalization and critical engagement, AI text, even if technically correct, looks like those green-screen Hollywood movies: beautiful, but fake; and there’s a lot of viewer frustration with the overuse of green screen. In our case, it’s the reviewers who are frustrated.
It is also “easier” for the supervisors and project leaders to delegate tasks to AI instead of hiring a living human being as a trainee or let a student figure out how to perform certain research tasks. If employers take that AI shortcut, an important learning avenue for new generations of academics disappears.
Not an Equalizer, but a Path to More Inequality
GenAI looks like it may have the potential to become an equalizer: non-native writers in English can produce prose closer to publication standards, students from less privileged backgrounds can gain access to individual AI-tutors, and precarious workers can save time on producing application documents faster and without professional coaching help or extensive informal connections.
In reality, I think, the effect will be the same as usual: it will produce new forms of inequality. Those better-off will have the luxury to go analog, will write in notebooks with fountain pens, and create small study groups or employ private tutors. Those who participate in mass factory-like education will lose everything mass education has achieved so far for them: classroom tools and assessment forms that support learners across levels and backgrounds. Class attendance is on the decline, and writing, the large-classroom-friendly form of engagement and assessment, experiences a crisis of trust. Those who already possess strong critical thinking skills and have access to tutors or other cultural and economic resources will benefit the most.
To conclude, don’t get me wrong: I think GenAI can be very useful for academic work. It is the lack of critical engagement with what you produce when collaborating with GenAI that puts you at risk of producing something that looks like a green-screen film.
Comments
Nothing yet. Say the first thing.
Sign in to join the conversation.