In recent months, school service centers have seen the emergence of two new products related to the data management platform Mozaïk-Data of Société GRICS. Based on artificial intelligence, they are intended to facilitate decision-making within school organizations, especially to identify more quickly students at risk of failure or dropping out.
Roberto Bacos Junior, product owner in business intelligence and artificial intelligence of the Mozaïk-Data suite at GRICS, took advantage of the Digital education summit to present the two platforms: Mozaïk-dVision and Mozaïk-dStudio. Initially, these are intended for the management of CSS and educational institutions in the French and English-speaking youth sector.
According to Mr. Bacos, the deployment is underway in all Quebec HSAs and the majority already have access to it.
The development of these new products is possible thanks to the GRICS Data Space, which gathers structured (or unstructured) data from different sources in the school environment. They are essentially tools for the valorization and visualization of these data, in the form of dashboards, which allow the prediction of certain situations.
"School workers get a picture of their clientele more quickly at the beginning of the school year. The information provided does not replace their judgment, but it can guide them and allow them to intervene more quickly in certain cases. It represents an additional resource among those available," explained Mr. Bacos.
dVision allows to see the information processed from the GRICS database and to analyze it in order to guide decisions. dStudio is a self-service workspace in which CSS can manage their data. It is therefore primarily intended for use by CSS IT departments.
First data in grade 4 writing
Initially, the development of the platforms focuses on the predictions of success or failure of students in grade 4 in writing. This is in order to be able to identify students at risk of failure on the mandatory departmental writing test. Similarly, early detection of students at risk of failure in writing proficiency will allow for more appropriate support and reduce delays in their further schooling, Bacos says.
New dashboards can be generated at the beginning of the school year (based on the previous year's results), at the end of the first stage in November and at the end of the second stage in March. Data such as gender, age, mother tongue, number of absences, previous results are used. Each variable can be isolated and the effect is then analyzed.
To date, the accuracy rate is between 87 % and 90 % and has identified "silent failors," students who no one suspected might fail and who did.
Ethics of artificial intelligence
Obviously, the work is not done without a constant concern to limit the possible biases associated with the development of tools based on artificial intelligence. Nesrine Zemirli, a specialist in the field, who is accompanying GRICS in the project, came to testify about the ethical governance framework that is also being built. The objective is not to create discrimination from the interpellation of data.
The 3 principles that guide the work:
- Every individual must have the same chance to be well predicted by the algorithm.
- The algorithm should not underperform for a category.
- A trend detected by the algorithm must not become stigmatizing.
The development of the governance framework is strongly inspired by the Montreal Declaration for Responsible AI.