Pre-university students’ perception on algorithmic biases of artificial intelligence

Abstract

This research focuses on pre-university students’ perceptions of algorithmic biases in artificial intelligence. Six types of biases (generational, gender, functional diversity, ethnicity, geographical origin and economic reasons) are examined on the basis of four variables (age, sex, educational level and academic year) of young people. A quantitative method is employed using a questionnaire. ANOVA, T-test and Kruskall-Wallis test are used. The results show statistically significant differences in the variables analysed and, in general terms, young people have a medium-high perception of possible biases. The highest number of differences between groups was found in the level of education (secondary education/baccalaureate/vocational training). The least differences were found in age (less than 12 years/12–14 years/15–17 years/18–21 years) and sex of the participants (male/female). Students in vocational training have a higher perception of bias and those in baccalaureate have the lowest means. In this case, significant differences were found. The results also show significant differences in biases produced by functional diversity, geographical origin and economic reasons. In relation to age, significant differences were found in two groups of students. According to sex, males have a higher perception of gender and ethnicity biases. These results have consequences for educational practice, as they highlight the aspects that should be addressed in the training of young people in artificial intelligence. It also has implications for research as it opens up new questions to be analysed.