Resisting cognitive laziness: A DASHBOARD to make AI use visible and transform critical thinking

Team name: Reboot Team

Use of AI tools : We have used Gemini and ChatGPT to translate and summarise our project into English

External feedback & contributions : L'Atelier Pédagogique d'UPV has helped us to be more pragmatic and realistic in carrying out the project, challenging us with reflective questions.

Initial contribution: Resisting cognitive laziness: a DASHBOARD to make AI use visible and transform critical thinking

Final contribution:

Imagine a very common scene.

Step 1. A student opens ChatGPT to summarise a dense academic text because time is short and the reading load is high.
Step 2. The AI produces something fluent, fast and convincing, but the student is left with a hidden doubt: Did I really understand this, or did I just outsource the effort?
Step 3. The teacher receives the final work and faces the opposite frustration: How can I distinguish genuine intellectual work from passive dependence without turning evaluation into suspicion?

That is exactly why our project matters. Today, AI is already embedded in student life, and many students use it for core academic tasks such as explaining concepts, summarising material and structuring ideas. At the same time, institutions are still struggling to provide clear guidance, and students report anxiety about being falsely accused of misconduct. Our dashboard responds to both frustrations at once: the student’s uncertainty and the teacher’s lack of visibility.

Our contribution is a dashboard as a pedagogical visual interface, almost like a compact infographic, that makes the invisible visible. Instead of only showing the final answer, it shows the process: the quality of the prompt, the degree of transformation between AI output and final text, the level of source verification, the reformulations made by the student, and the points where meaning may have been simplified, corrected or lost. In other words, it creates visibility where there is now opacity, and transparency where there is now only suspicion. This matters because recent evidence shows that generative AI can improve immediate task performance while weakening deep learning when it is used as a shortcut rather than a learning tool.

Technically, the project works as a structured sequence. First, the dashboard captures the interaction and displays a set of clear indicators. Second, it checks whether real understanding is present through short reflective prompts and comprehension checks, not just through surface traces. Third, if weak points appear, it activates a simple improvement protocol: compare sources, rewrite in one’s own words, justify a correction, or ask AI to support thinking without replacing it. This logic is coherent with the PMAISE model, which shows that AI becomes educationally meaningful when it is combined with structured feedback, interactive scaffolding and personalization alignment. It is also consistent with constructive alignment, because the real question is not what the tool produces, but what the student actually does with it and how that activity is aligned with learning goals.

Our proposal also includes a technical and ethical backbone. If students detect and verify hallucinations or factual errors, and if these are validated, the university can aggregate anonymised, discipline-specific contributions and potentially redirect them to research or partner ecosystems. A wallet system can then redistribute value through campus credits for students and recognition for teachers, while preserving privacy, transparency and human control. This is why ethics is not an appendix in our project: it is built into the architecture itself, through agency, intelligibility, non-surveillance and careful governance. The broader policy direction is aligned with current calls to place human judgement, feedback and oversight at the centre of educational AI use.

What finally strengthens our proposal is the richness of its bibliography. We did not build this idea only on intuition. It is grounded in recent higher education evidence on student AI use and assessment anxiety, in OECD work on the difference between performance gains and actual learning, in pedagogical research on engagement and mediation, and in foundational theories of constructive alignment. This combination gives the project both conceptual depth and practical credibility: it is not just a nice interface, but a serious educational response to one of the most urgent tensions in higher education today.

Reflection on the process

During Phase 2, our contribution evolved from a relatively simple dashboard concept into a broader pedagogical, ethical and institutional proposal https://docs.google.com/document/d/1vVdMwtJcyJizCM9h0Ano0S4z2nf9ApL18Lb_V4TWDyw/edit?tab=t.0. Several sources of feedback shaped this evolution. Peer comments helped us clarify how our idea was being perceived from outside, which encouraged us to reinforce the dashboard as a reflective and non-punitive tool. Another comment questioned the realism of our original RLHF, reminding us that major AI providers do not simply accept structured user feedback without an actual partnership or channel. Expert discussions and keynote inputs also pushed us to think beyond a technical tool and to address the wider educational ecosystem: the role of the university, the risks of dependence, the importance of transparency, and the need to protect critical thinking rather than only improve productivity. As a result, we made several major changes. First, we restructured the project around four pillars: social, emotional, material and intellectual. This helped us move from a single-tool idea toward a more coherent model. Second, we clarified that the dashboard is a visual interface, almost like a compact infographic, that makes the student’s intellectual process visible during a summarisation task. Third, we strengthened the intellectual pillar by distinguishing three analytical layers: prompt quality, observable interaction traces, and pedagogical/cognitive interpretation. We also added a clearer comprehension stage and an active improvement protocol, so that weak indicators lead to practical guidance rather than mere diagnosis. Fourth, we introduced a more explicit ethical framework, inspired by trustworthy AI principles, to ensure that the dashboard supports agency, transparency and intelligibility instead of surveillance. Finally, we revised the RLHF contribution pathway: instead of imagining a direct student-to-Big-Tech pipeline, we now describe the university as a trusted intermediary that validates, aggregates and anonymises qualified feedback before it can be shared with research or partner ecosystems.

This process made our final proposal much stronger. It forced us to test not only whether the idea was attractive, but whether it was pedagogically meaningful, ethically defensible and institutionally plausible. The project became more rigorous because we clarified what the dashboard measures, what it cannot infer automatically, how reflection is elicited, and how improvement is supported. It became more realistic because we refined the implementation logic and the value redistribution model. And it became more human because we integrated the emotional dimension of AI use, especially guilt, uncertainty and the need for intellectual sovereignty. In the end, Phase 2 did not just improve our wording: it transformed our proposal into a more robust vision of how higher education could respond to AI without giving up its formative role.

Video pitch

Please provide a link to your 1-minute video pitch presenting your work: https://docs.google.com/videos/d/1QRlj98sGg5iBMN89AWwbpyp4Kb0HNuILE4yz0HsnM5Q/edit?scene=id.p#scene=id.p