Skip to main content

Cookie settings

We use cookies to ensure the basic functionalities of the website and to enhance your online experience. You can configure and accept the use of the cookies, and modify your consent options, at any time.

Essential

Preferences

Analytics and statistics

Marketing

Changes at "Resisting cognitive laziness: a DASHBOARD to make AI use visible and transform critical thinking"

Avatar: Nicol D. Nicol D. 12/04/2026 12:45

Title (Français)

  • +Resisting cognitive laziness: a DASHBOARD to make AI use visible and transform critical thinking

Body (Français)

  • +
  • +
    Team name
  • +
    Reboot Mental
  • +
  • +
    Team members (First name, LAST NAME, University)
  • +
    ARPASANU-SUSOI Nicoleta-Roxana ; nicoleta-roxana.arpasanu-susoi@etu.univ-montp3.fr
  • +ARAMBURU Josela ; josela-ines.aramburu@etu.univ-montp3.fr
  • +MIRANDA Anthony ; anthony.miranda@etu.umpv.fr
  • +BOUQUET Maxime  ; maxime.bouquet@univ-montp3.fr
  • +
  • +All students are from Univesité Paul Valéry
  • +
    What area does your use case primarily fall under?
  • +
    Training / education / pedagogy
  • +
    The AI use case you are working on
  • +
    We focus on a very specific and realistic use case in higher education: students using generative AI to summarise theoretical course material when they feel cognitively overloaded. Our target learner is not a disengaged student looking for an easy shortcut, but a capable and motivated student who faces dense lectures, heavy reading loads, fragmented attention, and frequent deadlines. In this context, AI becomes attractive because it is fast, fluent, and immediately available when the student feels saturated.
  • +Our project starts from the profile of a student like Emma (PERSONA), a demanding master’s student who uses AI to summarise lessons, explain difficult concepts, and structure revision notes, but who lacks a reliable way to judge whether the AI output is accurate, nuanced, and pedagogically useful. The problem is therefore not AI use itself, but the opacity of AI-assisted summarisation: students may receive a fluent summary without knowing what has been preserved, distorted, oversimplified, or invented. The dashboard we propose is designed to support this precise moment of use. It is not meant to replace studying, but to make the quality and limits of AI-assisted summaries more visible.
  • +
  • +
    Why this use case matters
  • +
    This use case matters because AI-assisted summarisation sits at the intersection of two very real pressures in higher education: academic overload and the need to maintain genuine understanding. Students increasingly use AI to explain concepts, summarise material, and support assessed work, often because university learning conditions compress too much information into too little time. In that sense, AI is often a coping tool before it is a cheating tool. At the same time, this convenience creates a pedagogical risk: students may perform academic tasks more efficiently without clearly seeing whether they are still engaging in the intellectual work that learning requires. Recent evidence shows that overreliance on AI can reduce active engagement, weaken deeper learning, and create false mastery, especially when fluent outputs are accepted without verification.
  • +Our persona illustrates this tension clearly. On the one hand, the student hopes AI will save time, reduce overload, and produce more accessible summaries. On the other hand, what she actually needs is help checking whether summaries preserve nuance, whether claims are grounded in the original material, and whether assisted fluency is being confused with real understanding. The invisible risk is that better task completion can hide weaker engagement with the intellectual work the task was meant to develop. This is why summarisation is such a strong entry point for the project: it is common, concrete, and pedagogically central.
  • +
  • +
    Your team's motivation and learning objectives
  • +
    Our motivation is to respond to a widespread but insufficiently addressed reality in higher education: many students use AI because it helps them survive academic pressure, yet they often lack tools to evaluate whether this help is still compatible with real learning. We do not want to moralise AI use or portray students as lazy. Instead, we want to understand how AI-mediated study practices can remain intellectually honest, reflective, and educationally meaningful. Our project is therefore motivated by a double concern: helping students cope with overload while protecting the cognitive processes that university learning is supposed to develop.
  • +Our learning objective as a team is to move from an abstract concern about “cognitive laziness” or “dependence” toward a more rigorous and usable pedagogical model. In particular, we want to learn how to translate broad concepts such as source rigour, human contribution, and metacognition into indicators that are not speculative or impressionistic. Rather than claiming to measure inner cognition directly, we aim to work with observable traces of use, structured comparison with source material, and reflective prompts that make students’ study practices more visible and discussable. In that sense, the project is both pedagogical and methodological: it is about designing a useful tool, but also about learning how far such a tool can make valid claims about AI-supported learning without exaggerating what it can know.
  • +
  • +
    Your initial contribution
  • +
    Our initial contribution is the development of a pedagogical dashboard focused on a specific and highly realistic use of generative AI in higher education: the summarisation of academic content. The real idea of the project is not to build another AI assistant, but to create a device that makes visible the cognitive process that takes place when a student uses generative AI to summarise a text.
  • +
  • +1. What is the REAL idea of the project?
  • + A device that makes visible the cognitive process when a student uses generative AI.
  • +
  • +The dashboard is activated when a student generates a summary using AI. From that moment, it brings into relation three elements: the original text, the AI output, and the student’s intervention. This comparison makes it possible to reveal, in a concrete and interpretable way, what happened during the transformation process: what was preserved, what was omitted, what was simplified, what appeared without grounding, and what was or was not reviewed by the student.
  • +
  • +2. What does the dashboard concretely do? Three things : A, B and C.
  • +A. Compares, brings into relation:
  • + the original text, the AI output, and the student’s intervention
  • + to reveal what happened in the transformation
  • +B. Signals, but not in a moral way, it shows:
  • + · what was lost
  • + · what was simplified
  • + · what appears without a source
  • + · what was not reviewed
  • +C. Shows the student their interaction and their own usage through patterns such as:
  • + · automatic use vs active use
  • + · editing vs copying
  • + · verification vs acceptance
  • +Concretely, the dashboard does three things. First, it compares and relates the source text, the AI-generated summary, and the student’s modifications in order to make the transformation process visible. Second, it signals key issues without moralising them: for example, what was lost, what was oversimplified, what appears without a source, and what was accepted without revision. Third, it shows the student their own pattern of interaction with AI, making visible tendencies such as automatic use versus active use, editing versus copying, and verification versus immediate acceptance.
  • +
  • +3. What are the dashboard indicators? fron I to IV.
  • +I. Content fidelity — To what extent does the summary respect the original?
  • + Indicators:
  • + key concepts present / absent
  • + distortions or simplifications
  • + inventions or hallucinations
  • +II. Level of transformation — What did the student do with the output?
  • + Indicators:
  • + direct copy vs reformulation
  • + text editing
  • + integration of other sources
  • +III. Verification rigor — Did the student check anything?
  • + Indicators:
  • + source search
  • + verification questions asked to the AI
  • + comparison with original material
  • +IV. Reflective activation — Was there explicit thinking?
  • + Indicators (non-automatic):
  • + responses to prompts such as:
  • + what did the AI omit
  • + what did you correct
  • + what do you not understand
  • +The dashboard is structured around four main dimensions. The first is content fidelity, which asks to what extent the summary respects the key concepts, distinctions, and nuances of the original material. Its indicators include the presence or absence of key concepts, distortions or simplifications, and possible inventions or hallucinations. The second is the level of transformation, which examines what the student actually did with the AI output: direct copy versus reformulation, degree of text editing, and integration of additional sources or ideas. The third is verification rigor, which indicates whether the student critically checked the information rather than accepting it directly. Its indicators include source searches, verification questions asked to the AI, and explicit comparison with the original material. The fourth is reflective activation, which concerns whether there was explicit thinking during the process. This dimension should not be inferred automatically, but supported through short prompts such as: What did the AI omit? What did you correct? What remains unclear?
  • +
  • +A key strength of this contribution is that it is realistically implementable precisely because it does not attempt to address cognition in the abstract, but focuses on one clearly delimited academic task: summarisation. Behind the dashboard, the indicators would not be arbitrary. They would be grounded in psychometric measures and in a strong semiotic analysis of the summarisation task, so that the dashboard can be interpreted through a combination of psychological, cognitive, and semiotic indices. Psychometric scales can be used to measure the presence and intensity of typical cognitive and affective biases that may shape AI-assisted summarisation, such as confirmation bias, anchoring, and forms of impostor syndrome, while also providing a structured way to assess metacognition, particularly the student’s awareness, monitoring, and regulation of their own use of AI throughout the summarisation process. In this way, when the dashboard is analysed, it would not merely display surface metrics, but reflect deeper dimensions of the task: how meaning is reduced, reorganised, altered, verified, or appropriated by the student.
  • +
  • +The dashboard therefore does not aim to evaluate or punish students. Its purpose is to provide a structured representation of their own use of AI. By making visible the relationships between source, output, and revision, it transforms an opaque practice into an interpretable and discussable pedagogical experience, both for the student and for the teacher.
  • +
  • +

Confirm

Please log in

The password is too short.

Share