A didactic management team carries out monitoring and tutoring activities of the different phases of the learning path in order to reach the fixed index of effectiveness.
In particular, tutors are requested to conduct a continuous analysis and an optimal interpretation of learning process data, followed by the definition of corrective strategies.
The quantitative data analysis of learning paths involving a wide number of learners moreover requires the functional and well-timed selection of information, as well as a flexible key to understand the dynamics existing between the connected variables.
In such situations it may be useful to refer to expert systems based on mathematical models, similar to the experimental software tool developed by our Research and Development team: the Guardian Tutor 1.0.
The Guardian Tutor 1.0 is an Artificial Intelligence application based on an Artificial Neural Network (ANN) aimed at managing the non-effectiveness Risk in those situations characterized by a high level of complexity and containing a considerable amounts of data.
This expert system is able to:
• identifying risk variables affecting the learning process;
• providing the process tutor with aggregate data related both to the whole group of users and to a specific cluster of learners;
• allowing the process tutor to screen a real time monitoring report;
• facilitating the decisional process related to the tutoring actions toward learners.

The architecture of the model and the phases of design and realization are illustrated in the paper:
1. specification of the steps of the process (flow-chart);
2. analysis and normalization of collected data;
3. design and logical development of the network;
4. training;
5. testing and implementation.

The Guardian Tutor 1.0 represents an experimental and innovative tool, at disposal of the process tutor, that extends the possibility of analyzing the non-effectiveness Risk and allows the planning of in itinere interventions in order to mitigate this Risk.
This way, the tutoring activities could be standardized and, as a consequence, they could be considered as a sort of “therapy” responding to the symptoms identified by the Artificial Neural Network related to a specific cluster of learners.