High school teacher introduces AI project for students

Sylvain Tissier, an instructor at a high school in Savigny, has introduced a project where first-year students train an artificial intelligence to detect infections in blood samples. The project aims to help students apply their immunology knowledge and understand how AI systems can be biased.

Students train AI to distinguish infected blood samples

Students are tasked with creating an AI that can differentiate between infected and healthy blood samples using images of blood smears taken under a microscope. They use the Vittascience platform, which allows AI training for image recognition without requiring programming skills.

Activity builds on immunology lessons and includes a debate

The project follows lessons on inflammatory reactions and adaptive immunity, allowing students to apply their knowledge, such as the increase in white blood cells during an infection. It is followed by a debate on the role of AI in medicine, focusing on how AI learns and the critical perspective on its use.

Students explore AI bias and create image databases

During the session, students first watch an interactive video from Inserm on AI in medicine and answer a questionnaire on their opinions. They then develop a strategy in groups, considering how to train an AI to recognize infected blood. They use a biased example, such as an AI trained to distinguish between men and women without considering men with long hair or women with short hair.

Hands-on training and evaluation of AI performance

After a group discussion, students create two image databases: infected and healthy blood samples, photographing slides under a microscope with a camera. They then train their model on Vittascience. In the evaluation phase, the AI is tested against a human on an unknown slide, comparing speed and reliability. Results are shared in a collaborative spreadsheet, followed by a debate game and a second questionnaire.

Project emphasizes critical thinking and ethical questions

This approach differs from traditional activities by requiring students to explicitly define what characterizes infected blood to teach a machine, thus applying their immunology knowledge concretely. They also discover that the quality of AI depends on the data provided. The comparison between AI and humans raises ethical questions, such as whether some errors in medical diagnosis are acceptable.

Implementation tips and minimal costs

Teachers considering this activity should ensure it is placed at the right time in the curriculum, after immunology has been covered. They should test the equipment beforehand, as the image capture via the browser with the camera may be problematic on some devices. The debate is crucial for providing a critical perspective on the activity, and the cost is minimal, using standard laboratory equipment.

Student engagement and impact on perceptions of AI

Students have shown significant engagement in the activity, which addresses the issue of a shortage of doctors. They have developed their own strategies, used microscopes, and tested an AI on a real diagnostic problem, making the activity very concrete. A questionnaire before and after the activity showed that students' confidence in AI for analyzing medical results increased, as did their belief that AI is a good thing for the future.

Students value doctors but see AI as a diagnostic aid

However, students remain strongly attached to the role of doctors, with the idea that medical decisions should be made only by doctors being the most shared. They see AI as an aid in diagnosis rather than a substitute for healthcare professionals. While no students have explicitly stated they want to become bioinformaticians, many were unaware of the profession before the activity. The debate on the shortage of doctors has been particularly impactful, giving the discussion a concrete relevance. Students were especially sensitive to issues related to mental health and data privacy when discussing the detection of depression.