The Pretend-Assessment-Game: Opening a Conversation of the Real Value of Written Student Work in Higher Education in the Time of Large Language Models

There is a silent (maybe not so) but transformative shift underway in the traditional approach to assessing students’ learning outcomes. Students across various disciplines are increasingly turning to large language models (LLMs) to draft, edit, and even entirely compose their written assignments – including coding assisted by intelligent agents, in the case of computer science. Meanwhile, academics, overwhelmed with administrative pressures and high marking loads, are also relying on similar tools to summarise submissions, generate feedback, and assist in the grading process. Despite the widespread use of these practices, higher education continues to behave as an institution that replicates an idea of bureaucratised teaching, idealising itself as a self-centred space of knowledge, and validating students’ learning primarily based on the product, as if the relationship between the production and dissemination of information and knowledge had not radically changed.

The traditional written student work, once viewed as a cornerstone of academic assessment, seems to have lost its reliability as a tool for measuring learning. This trend may be a consequence of decentralisation or questioning of the written language as the sole artifact sufficient to synthesise knowledge derived from reading, interpretation, and argument construction. What was once served as a kind of window into a student’s capacity for critical thought, synthesis, and argumentation now often functions more as a token performance: generated by machines, rubber-stamped by humans, and ultimately filed away with little educational impact. This ongoing charade raises a difficult but essential question: If neither students are genuinely engaging in producing their written work, nor academics are fully engaging in evaluating it, what exactly are we assessing?

The dissonance is striking. In many institutions, a tacit form of complicity exists. Everyone knows that LLMs are being used, but few are willing to admit it. Students are unsure about ethical boundaries, and academics privately express discomfort but lack institutional support on how to respond appropriately to this situation. This confusion deepens further by the absence of reliable tools to identify the use of LLMs. Universities maintain a facade of academic normalcy, adhering to conservative assessment models that no longer accurately reflect the reality of how work is produced or evaluated, particularly in light of the growing use of LLMs associated with premises of productivity and time-saving. In doing so, higher education risks undermining its own social function and credibility.

At the heart of this crisis is a breakdown in the connection between assessment and learning. Written assessments were never just about merely producing text. They were about deeply engaging with ideas, constructing arguments, and refining thought through iteration. If the criticism of the didactic theory in the last decades focused on denouncing an assessment centred on the product (the final assessment instrument – the written text) to the detriment of the learning process with different assessment moments and modes of language expression, today the danger of centring assessment on a written text in the classical model takes on even more concerning proportions. When students outsource cognitive work to an LLM, the benefits of developing autonomous argument construction are lost. They may still earn high marks, but the learning that those marks mean may never have occurred. And when those written pieces are evaluated through a similarly detached process, summarised or scored with minimal human engagement, the opportunity for meaningful feedback evaporates. The entire process becomes a closed loop of automation, devoid of genuine intellectual effort on either side.

This erosion of substance has profound implications. If students graduate with skills in prompt engineering but lack critical thinking, or if they have perfected text generation without mastering argument construction, then the value proposition of higher education becomes emptied of meaning. Academics, policymakers, and students themselves may justifiably begin to question what a degree truly represents.

To move forward, education must first be willing to openly discuss what is happening. We cannot address a problem we refuse to name. Acknowledging the role of LLMs in academic work is not an admission of failure; rather, it is the first step toward developing new models that are more transparent, effective, and future-oriented. This requires rethinking what we assess and how we assess it. Instead of narrowly focusing on the written product, we must consider the learning process: how ideas are formed, challenged, and refined over time. Assessments could involve stages of submission, in-person components, reflective journals, or collaborative projects – formats that emphasise engagement and originality in ways that are difficult to replicate with an LLM.

This shift also demands greater transparency and digital literacy, especially concerning artificial intelligence. Students should be encouraged to disclose how they use LLMs and reflect critically on that use. Rather than prohibiting the use of LLM tools outright, which is becoming an increasingly untenable position, we should teach students to use them responsibly and discerningly, as part of a broader skillset that includes judgment, ethics, and self-awareness.

Academics need support as well. Training and institutional guidance are essential if educators are to use LLM meaningfully without relinquishing their pedagogical role. LLMs can assist with workload management, but they must not replace the human connection that adds value to education. Feedback must still reflect genuine understanding of a student’s strengths, challenges, and potential.

If higher education continues to pretend that traditional teaching and assessment models are still consistent with contemporary educational processes, it risks becoming a hollow and costly credentialing system that offers little value and teaches even less. However, this moment presents an opportunity if we approach it with honesty and creativity. We have the chance to redefine what matters in learning, realign assessment with intellectual development, and reaffirm the value of education in an age where machines can even write, but only humans can truly think.

This essay has been writen in collaboration with:

• by Everton Cavalcante, Assistant Professor in Computer Science at the Federal University of Rio Grande do Norte, Natal, Brazil, https://orcid.org/0000-0002-2475-5075
• by Allyson Carvalho de Araújo, Associate Professor in Physical Education at the Federal University of Rio Grande do Norte, Natal, Brazil, https://orcid.org/0000-0003-0114-8122\

Also published in PT-BR at https://horizontes.sbc.org.br/index.php/2025/08/o-jogo-da-avaliacao-faz-de-conta-iniciando-uma-conversa-sobre-o-valor-real-dos-trabalhos-escritos-de-estudantes-no-ensino-superior-na-era-dos-grandes-modelos-de-linguagem/