Machine Teaching: An Explainable Machine Learning Model for Individualized Education

• Eduardo Vargas Ferreira ITA / UFPR https://orcid.org/0000-0002-9241-9778
• Ana Carolina Lorena ITA https://orcid.org/0000-0002-6140-571X

Resumo

Education is the single most important investment that people can make in their futures, and since the Universal Declaration of Human Rights in 1948, the goal of achieving universal education has been on the international agenda. In this regard, there is no doubt that the Web has had a profound impact on making education both universally available and more relevant. Nevertheless, online courses are based on “static” learning material (one-size-fits-all). For that reason, it is not straightforward to assess learning with a great number of learners who differ considerably in their educational background, engagement styles, and cognitive skills. In this work, we aim to address these aforementioned challenges by proposing an explainable Machine Learning algorithm for personalizing web-based education systems. The method has a “deep” architecture mimicking the information representation structure in human brains, and it is continuously adapted based on the signals of the students, understanding their performance through micro-steps and maximizing the learning outcome. While our methodology is general and can be applied in numerous scenarios, we demonstrate its performance by a real case study which comprises a non-mandatory, standardized exam, that evaluates high school students in Brazil.