Overview
Antonio has been running structured experiments to understand how generative AI is changing economics education. The work moves beyond concerns around ‘cheating’ and misuse and instead rebuilds pedagogy around these tools.
Creating a specialised, course-aligned AI tutor
Antonio developed a specialised, course-aligned AI tutor with Professor Ronny Razin, whose key insight is that the best way to verify understanding is to ask students to explain concepts interactively.
The Socratic chatbot refuses to give final answers, prompting students to take the next step in a mathematical or logical derivation and clarifying notation or concepts when needed. The aim is to provide 24/7 personalised tutoring that levels the playing field.
Student feedback and impact to date
It was piloted in Autumn Term 2025 alongside Razin's first-year Microeconomics teaching (around 850 students), and saw more than 500 students engaging.
Feedback from a targeted game theory exercise was overwhelmingly positive, with students asking for weekly sessions and fewer time limits.
A preliminary analysis shows that students that completed the session with the chatbot on a single exercise obtained 6 extra marks on average in the exam with respect to students that could not use the chatbot because of technical issues.
Reverse Engineering Learning with AI
Antonio also developed a new pedagogical approach called RELAI (Reverse Engineering Learning with AI) – a framework for helping students analyse and extend code with AI as scaffolding rather than solution-provider.
The work reflects a shift towards teaching students to read, organise, and verify code with AI support, as well as introducing students to AI coding agents. Students start from a full Python snippet, use AI to explain what it does and probe the underlying concepts, and then extend or modify the code – debugging with AI when needed.
Comparing AI-generated feedback with human feedback
Antonion also experimented comparing AI-generated feedback with human feedback for highly technical projects within the MSc programme.
Early findings show that, while students found the AI feedback to be more positive and encouraging, it often lacked the depth and precision required for rigorous technical critique.
This outcome implies that a hybrid approach could be advantageous, where teachers collaborate with AI tools to produce a quality-assured set of feedback comments, combining the supportive tone of AI with the technical expertise of human evaluators.
Scalability
The Socratic tutor model is designed to transfer to other courses and institutions; RELAI has applicability beyond economics.