AIDE is a project aimed at developing computational thinking through the creation of an interactive machine learning environment. This project will be used to study the development of computational thinking, while using a digital expertise approach that is both creative and critical.
A complementary approach to the links between artificial intelligence and education.
In education, AI is usually applied as a tool, that is often disruptive (for e.g., digital assistant), that uses machine learning algorithms. In this approach the learner interacts with machine learning tools. However, it is possible to consider that the learner continues to carry out disconnected activities, but the environment and traditional tools are enriched in order to capture, analyze, and understand learning processes better, while providing feedback in real time. Thus, machine learning can be seen as a means of enriching these learning devices in which analogous and disconnected activities can take place.
On the other hand, we also highly recommend the need for citizen training in artificial intelligence in order to master, and thus understand, these mechanisms and implement them.
Here we propose a third axis that appears to be little or not developed:
use the formalisms and mechanisms of machine learning as a paradigm in the science of education.
The three operational components of the AIDE project.
The AIDE project has three operational components, complementary to what already exists:
- At the theoretical level, use the machine learning formalisms as a model of the learner, as it can be done in computational neuroscience on another scale.
- At the experimental level, use machine learning tools to obtain more formalized, more reliable, and more automated measures of variables related to a learning situation.
- At the pedagogical level, help the learner to understand in his own process of learning to learn, which is a matter of mechanical processes of what is creative and making these two aspects explicit.
In other words, before using artificial intelligence to, for example, improve human learning, let’s try to use it to better understand how a human learns.
A specific object of study: disconnected activities.
Learning of computational thinking as the foundation of digital education seems to be really effective with activities disconnected for several well-understood reasons but not easy to establish experimentally.
We propose here to build a measuring device that allows a learner (or a small group) to practice a unplugged activity to solve a problem related to computers and educational robotics in connection with the existing work. The use of machine learning tools will make it possible to:
– Make measures based on the gestures or attitudes of the learner in order to automate and ensure reliable collection of data, and to connect them better to cognitive processes, such as the exploitation and exploration processes.
– Confront the cognitive mechanisms involved in human learning.
with underlying models in machine learning (e.g., duality actor / critic in learning situations with reward (reinforcement)).