# Navigating the future of education: how Institución Educativa ALEPH is trying to advance learning with AI

Author: Fernando Nieto Lobato
Original publication: 2024-09-11
Spanish original: https://estrategiabyaleph.substack.com/p/estrategia-50-la-transformacion-de
English URL: https://elcontemplador.github.io/estrategia-english/essays/050/
Status: Published translation

This is a translation of the original Spanish essay published on 11 September 2024. Its claims, examples and forecasts retain that historical context.

English publication: 2026-09-29

Artificial intelligence, or AI, has brought about what Ethan Mollick, a professor at the University of Pennsylvania's Wharton School, calls the “Homework Apocalypse”. The figures are striking: 82% of university students and 72% of primary and secondary school students in the United States already use AI to do their schoolwork. This widespread adoption poses **an unprecedented challenge to the traditional education system**, which Mollick explores in depth [in a recent and very interesting article](https://www.oneusefulthing.org/p/post-apocalyptic-education).

Mollick, whose [excellent latest book was the subject of issue 38 of our newsletter](https://estrategiabyaleph.substack.com/p/estrategia-38-descubra-las-cuatro), warns of two crucial phenomena: **“Illusory Knowledge”**, whereby students overestimate their understanding when using AI, and the **“Detection Illusion”**, whereby educators mistakenly believe they can easily identify AI-generated work. A study in Turkey found that unguided use of ChatGPT for homework led to final exam scores that were 17% lower, despite improving homework grades.

**At Institución Educativa ALEPH, we have decided to confront this challenge.** We can clearly see that rather more students than we would like are using these new tools simply to do their [master's degree](https://institucioneducativaaleph.com/master-en-marketing-comunicacion-y-consultoria-politica/) assignments for them, rather than to learn, just as some students once copied from Wikipedia or various websites without properly citing them or working through and absorbing the material. [Instead of resisting change, we have embraced it by developing AI tutors based on GPT proto-agents for our Master's in Political Marketing, Communication and Consultancy, ready for the coming 2024–25 academic year, which begins in October](https://institucioneducativaaleph.acblnk.com/show/2815751/fSn5Zdgo8vaS9Vi9kUGbjsLWi991m/c8e40f932aaf4c9e2914b810d2cb3eee/). As director of estrategIA and leader of this project, I have been fortunate to work on this pioneering initiative. It is important to stress that it is still in an advanced beta phase, as we are working with entirely new technology. It is also subject to the limitations of AI models and to the decisions large companies, in this case OpenAI, make about their products. Finally, we have yet to begin observing how students actually use it, so that we can improve it step by step.

[Our AI tutors have been designed for each module of the master's programme, with the aim of helping students better understand the course material](https://institucioneducativaaleph.com/tutoresia-aleph/). Developed as proto-agents integrated into ChatGPT, these agents have already undergone extensive testing and many iterations. They are entirely additional to everything the programme offered in previous years: they are new complements rather than replacements, joining the outstanding international teaching staff and the excellent gamified methodology that are hallmarks of education at Institución Educativa ALEPH and its [Master's in Political Marketing, Communication and Consultancy](https://institucioneducativaaleph.com/master-en-marketing-comunicacion-y-consultoria-politica/). These AI tutors are intended to be available at all times to answer questions, offer further explanations and guide students through practice assessments.

Developing these tutors has been a fascinating, challenging and lengthy process. The first prototypes date from November last year, but at the time free users could not use GPT proto-agents. That was an insurmountable limitation if we wanted to integrate them directly into ChatGPT to make them easy to use: it is still the only AI technology many people know and use. One difficulty has been adapting each module's educational content into a format the AI could retrieve effectively. Above all, we have faced the challenge of getting the agents to stick strictly to the syllabus, without straying into irrelevant matters or answering unrelated questions. This is probably the most significant difficulty these systems face with current AI models when political leaders want to make them publicly available, although the next generation will probably resolve much of it. With a little knowledge and malicious intent, it is relatively easy to get these models to “talk” about irrelevant matters. With more sophisticated *prompt* injection, it is even possible to hack the system into making outrageous statements that could destroy the credibility of a politician who wanted to offer such a tool to people interested in their policies, even where considerable effort has gone into protecting its instructions.

[![Spanish-language interface of ALEPH's tutor for topic 12, Social media communication in an election campaign, created by Fernando Nieto. Its two suggested actions are to ask questions about the topic and take a knowledge test.](https://elcontemplador.github.io/estrategia-english/assets/images/3ee00db9-414e-4f23-b054-dde6c3074d8d_869x1063.png)](https://substackcdn.com/image/fetch/$s_!ffpi!,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2F3ee00db9-414e-4f23-b054-dde6c3074d8d_869x1063.png)

*Screenshot in Spanish: the tutor supports topic 12 of ALEPH's Master's in Political Marketing, Communication and Consultancy, “Social media communication in an election campaign”. The suggested actions are “I would like to ask questions about topic 12” and “I would like to take a knowledge test”. The interface also warns that ChatGPT can make mistakes and that important information should be checked.*

Another challenge, which required several iterations before it worked reasonably well, has been creating assessment tests that are genuinely useful to students and that reproduce the format of real assessments by default. Students can, however, ask the proto-agent for other formats to assess their own understanding and command of the topics.

Integrating AI into education raises fascinating questions about the future of learning, the role of educators and the very nature of knowledge in the digital age. As we move forward, we must ensure that technology extends, rather than simply replaces, the fundamental human capacities for critical thinking, creativity and empathy.

**We are only scratching the surface of what is possible. The journey towards education enhanced by AI will be complex, but also incredibly exciting.** At Institución Educativa ALEPH, we are proud to be at the forefront of this transformation, working to create an educational future that is more accessible, personalised and effective than ever.

As a **“gift”** to you, dear readers, we are sharing an early version of one of the projects we hope to pursue in the near future. It is still at a very early beta stage, although it is already fully usable. We are confident that more powerful new models, and [especially the integration of voice and video](https://x.com/kimmonismus/status/1831604030994575613), will enable us to take it further: creating GPT proto-agents of significant figures, in our case from political thought, with whom we can converse and who can explain their ideas and knowledge.

Here is one of our first creations: a Machiavelli proto-agent. It has been given [a public-domain version of *The Prince* as its main source for conversation](https://es.wikisource.org/wiki/El_pr%C3%ADncipe_%28S%C3%A1nchez_Rojas_tr.%29), together with instructions, still being developed, to enable it to discuss its thinking in the first person.

**[Machiavelli ChatGPT 0.1 beta](https://chatgpt.com/g/g-YzsnrjsUw-maquiavelo-bot-0-1-beta)**

[![Spanish-language interface of Machiavelli bot 0.1 beta, by Fernando Nieto, inviting readers to discuss Machiavelli's thought and work.](https://elcontemplador.github.io/estrategia-english/assets/images/1450fd02-ad9d-4804-9797-02b8e4cc8bdb_851x988.png)](https://chatgpt.com/g/g-YzsnrjsUw-maquiavelo-bot-0-1-beta)

*Screenshot in Spanish: “Talk to Machiavelli about his thought and work.” Suggested questions ask about the most important aspects of his thought, the qualities of a great leader, how to avoid flatterers and how a leader can acquire a good reputation. The interface warns that ChatGPT can make mistakes and that important information should be checked.*

Fernando Nieto Lobato

*Director of Digital Innovation, Institución Educativa ALEPH*