Point 3(c) of Annex III classifies as high risk AI systems intended to be used to assess the appropriate level of education that an individual will receive or will be able to accessin the context of or within educational and vocational training institutions at all levels . Such systems may serve two main purposes. The first is determining the level of education an individual will receive in terms of progression within an ongoing learning path, such as advancing to the next grade or moving to a higher-level course. The second is assessing the level of education an individual will be able to access by evaluating their readiness and qualifications before they enter or begin a programme. The latter assessment often happens through placement tests or skill assessments used to determine the appropriate starting level or eligibility for specific educational or vocational pathways.
AI systems intended to be used for the first purpose may use as a basis the student ’s current level of knowledge and skills in a particular subject area, such as mathematics or languages, and in this way determine the most suitable level of course, programme, or training module for them to follow next, thus supporting a personalised learning pathway. Such systems may also take into account any special educational needs or talents that a student may have, such as dyslexia or a prodigious ability in music, an d provide personalised recommendations for their education. For example, an AI system may recommend a student with a special talent in mathematics to take advanced courses in calculus or number theory, while recommending a student with dyslexia to receive additional support in reading and writing.
AI systems intended to be used for the second purpose may be used to evaluate an individual’s skills, knowledge, and abilities to identify the most suitable educational or vocationaltraining level for that person. By analysing the learner’s current competencies-and, where relevant, their potential for progression-the system can recommend the appropriate placement within the education pathway and determine whether the student is ready to advance to the next stage of study.
The ultimate decision-making responsibility rests with teachers and educational and vocational training institutions ’ responsible staff who may accept the AI system ’s recommendations or override them based on their professional judgment and knowledge of each student’s needs and abilities. However, if not properly designed and deployed, such systems could misclassify students and negatively impact their learning opportunities and long-term outcomes , which is why they should be classified as high-risk.
a) Practical examples of AI systems falling within the high-risk use case of point 3(c)
- AI-powered adaptive placement tool s intended to be used by an education al or vocational training institution to determine the optimal course level for incoming students. Such a system assesses a student’s abilities to formulate recommendations addressed to educators whether the student should be in beginner, intermediate, or advanced classes, or decides if he or she should move to higher education or vocational training based on their skills.
- Vocational training level assessment system s intended to be used by a vocational training institution to assess the level of education that a learner will be able to access in a specific vocational program. Such a system may use a combination of placement tests and skill assessments to evaluate the learner’s knowledge and skills in relevant areas, such as mathematics, physics, and engineering principles, and determine their eligibility for a particular level of study or course.
- AI classifier s for special education intended to be used to classify students with special educational needs into appropriate educational programs and recommend the most suitable educational placement for each student, including the level of support required . Such a system may use a machine learning algorithm to analyse a range of data, including student assessment data, such as results from standardized tests, including IQ tests, achievement tests, and behavioural assessments. It may also consider teacher’s feedback on student’s behaviour, social skills, and academic performance, as well as psychological evaluations, including reports from psychologists on student’s cognitive abilities, emotional intelligence, and learning styles, and the student’smedical history, including information on any diagnosed conditions or disabilities.
b) Practical examples of AI systems falling outside the high-risk use case of point 3(c)
- Personalised education recommendations for students: An AI system intended to be used to provide personali sed educational recommendations to students for the appropriate level of education, such as courses, programs, or certifications, that they may be interested in applying for the following academic year on the basis of information provided by students themselves, such as interests and career goals. Since the system is not intended to be used by an educational institution, but is only intended to help students make informed decisions about their educational path, of the system does not fall within the use case listed in point 3(c) of Annex III.
- Trend Analysis of Educational Attainment: An AI system intended to be used by educational institutions, policymakers, and researchers to analyse from various sources, including educational institutions and government databases, trends in the levels of education accessed after completion of secondary education by students, and to identify patterns in the types of educational programs and levels of education that students are pursuing. Since the system’s output is not used to make decisions about individual students’ educational placement or opportunities, but rather to provide a broader understanding of the educational landscape, it does not fall within the use case listed in point 3(c) of Annex III.
c) Practical examples ofAI systems falling within theuse cases but exempted by the filter mechanism of Article 6(3)
- AI-enabled p re-assessment tool intended to be used to help educators prepare for the assessment of a student’s readiness for a particular level of education or trainingbased on data , about the student’s prior education and work experienceand generating , a set of questions and topics that the educator can use to assess the student’s readiness should not be considered high-risk. Since the system’s output is used by educators to inform their assessment methodology, but does not itself make any decisions about the appropriate level of education that an individual will receive or will be able to access, it should therefore be considered to fall under the exception for systems intended to perform a preparatory task listed in Article 6(3)(d) AI Act.
- Improving recommendations for advanced course placement: An AI system intended to be used by an educational or vocational training institution to analyse the results of tutors’ recommendation to apprentices to pursue certain advanced courses , based on data from the apprentice's performance records and learning outcomes, and providing suggestions for improving the recommendation methodology , such as identifying additional factors that could be considered, or suggesting alternative courses of action that could be taken. Since the system’s output is used by the institution to refine its recommendations and to improve the decision-making process, but it does not take any decisions on the apprentice’s educational placement or opportunities, it should be considered to fall under the exception for systems intended to improve the result of a previously completed human activity listed in Article 6(3)(b) AI Act.