Even this year we had the pleasure of presenting our scientific contribution to the National Conference Didamatica, long a point of reference for discussing emerging topics in the evolution of teaching and the school of the future, regarding Human-Computer Interaction: this year’s edition was dedicated to the theme “Artificial Intelligence for Education” and to the ways in which AI algorithms, systems, and approaches can impact teaching, learning, and education in general.
The conference was organized by AICA – Italian Association for Computer Science and Automation, in collaboration with the Institute of Didactic Technologies of the National Research Council, the Ministry of Education, and under the patronage of the Agency for Digital Italy.
We presented the paper “Enhancing Teachers – AI Collaboration: Human Computer Interaction Techniques for Recommender Systems in Educational Platforms”, which outlines the methodologies and approaches identified to significantly improve the management of the interaction between teachers and the Recommender System of the WhoTeach platform in the personalized creation of courses.
How to Conduct an Evaluation of Human-Computer Interaction of Intelligent Recommender Systems
While the literature is abundant on the technical and implementation aspects of Recommender Systems, it is scant on the aspects related to interaction and users’ decision-making processes. How can the performance of recommendation systems equipped with Artificial Intelligence be evaluated with regard to Human-Computer Interaction?
To answer this question, we conducted a study on WhoTeach, the e-learning platform developed by Social Thingum. WhoTeach includes a Recommender System equipped with Artificial Intelligence that helps experts and teachers quickly and effectively aggregate high-quality content into courses: it allows teachers to filter, select, and suggest educational resources in any format, so that they can be integrated into complete courses according to the needs or requirements of both teachers and students.
We followed 3 guidelines identified among the Guidelines for Human-AI Interaction particularly suitable for evaluating model-based recommendation systems:
- G2: make clear how well the system can do what it is capable of doing: help the user understand how often the AI system might make mistakes;
- G13: learn from the user’s behavior: personalize the user experience by learning from their actions over time;
- G14: update and adapt cautiously: mimic disruptive changes when updating and adapting the AI system’s behavior.
Following these guidelines, evaluation tests of Human-Computer Interaction were carried out, taking into account human decision-making behavior.
The Key Concepts to Improve an Intelligent Recommender System
Through this analysis, we were able to identify some key concepts of Human-Computer Interaction that can improve the performance of Recommender Systems.
- User Control: the user must be in control of the experience. This means establishing an appropriate level of trust regarding how data are managed, while also addressing issues related to privacy.
- Adaptivity: resource recommendations should not be aimed solely at achieving the highest accuracy, but must adapt to the needs and desires of users.
- Emotional Aspects: emotions have a profound influence on decision-making processes. A recommendation system should be able to utilize behavioral data when proposing resources.
- High-Risk Domains: not all fields of knowledge can be treated in the same way: there are areas where the level of risk is higher and decision-making becomes more complex. For example, in the medical-surgical field it is more difficult to place trust in Artificial Intelligence.
Conclusions
In this study, we have outlined a framework useful for the evaluation and improvement of AI-based Recommender Systems. The chosen methodologies demonstrate that combining the data used by recommendation systems with visualization and interaction techniques could enhance both the quality and accuracy of the recommendations, while also increasing the system’s transparency.
In future editions of Didamatica, we will continue our studies on the influence of Artificial Intelligence in the field of Education. For this reason, we invite you to follow us and stay updated on the world of Innovation!