Point 3 (a) of Annex III classifies as high-risk AI systems intended to be used to determine access or admission, or to assign natural persons, to educational and vocational training institutions at all levels. In this context, ‘access, admission or assignment ’ refers to the process by which it is determined whether a person may enrol, gain entry, or be placed into a specific course, programme or learning pathway. The output of these systems c ould affect one’s learning and educational pathway and right to education.
AI systems typically falling within this use case are those intended to be used to assess applications to determine the access of specific natural persons to specific educational institutions at various levels, to process the evaluation of their applications, or to assist in such admission decisions, such as student placement and assignment. Similarly, vocational training assignment systems, which match job seekers or employees with suitable vocational training programs, should also be considered to determine access or admission to vocational training institutions, including access or admission to various programs within those institutions at all levels, and therefore be classified as high-risk pursuant to point 3(a) of Annex III. Considering that the evaluation of a learner’s performance or progress can occur independently of the admission and assignment processes, AI systems intended to be used for such evaluations should be considered to fall outside this use case, although they may fall within the use case listed in point 3(b) of Annex III.
The use case listed in point 3(a) of Annex III should also be understood to include AI systems intended to be used to assign students or apprentices to specific programs or courses within an educational or vocational training institution (or to a subunit of it) at all levels, since their output directly affects a person’seffective exercise of their right to study, train, or enrol in that institution.
a) Practical examples of AI systems falling within the high-risk use case of point 3(a)
- Automated admissions systemsintended to beused to review and evaluate student applications, transcripts, and test scores to determine eligibility for admission to an educational institution. This includes systems that evaluate applications from prospective students and determin e their eligibility for admission, since the system’s output will directly affect the applicant’sright to study, train, or enrol at the institution . That will be the case, for instance, where the system's evaluation and recommendation have the potential to influence the decision-making process, in that it determines the outcome of an admissions decision. In cases where the system does not perform profiling, the filter mechanism in Article 6(3) AI Act could apply , for example, if the system is instead intended only to perform a preparatory task to the assessment of eligibility done by the educational institution (Article 6(3), letter (d)).
- School assignment system s which take their decisions based on location , availability and other personal characteristics of the applicant that is intended to be used by municipalities or regional authorities to assign students-often in primary or secondary education-to public schools in an automated manner. Such systems function by collecting structured data, such as the student’s home address, the geographical boundaries of school catchment areas, and the available capacity at each school. They also factor in sibling attendance and parental status to keep families together or to optimise for logistical efficiency, such as minimising commuting distance. Such systems process such data to assign students to schools, ensuring compliance with zoning rules and balancing enr olment across institutions. Since the system’s assessment of data constitutes profiling, the filter mechanism in Article 6(3) AI Act cannot be applied to it.
- Vocational training assignment AI tools intended to be used by a regional employment agency to process applicants’ prior education records, completed certifications, and standardized aptitude test results to match individuals to available apprenticeship programs should be classified as high-risk. Such systems apply predefined eligibility criteria , such as minimum qualifications required for each program and capacity limits, to assign candidates to appropriate training slots. Since the system’s assessment of data constitutes profiling, the filter mechanism in Article 6(3) AI Act cannot be applied to it.
- Automated scholarship eligibility assessment system intended to be used to automate the process of determining an individual’s eligibility for a scholarship programme based on data about the individual’s financial situation, including their income, expenses, and family size. Such systems assess the eligibility of an individual’s eligibility for financial assistance , thus determining their access to an educational institutional.
b) Practical examples of AI systems falling outside the high-risk use case of point 3(a)
- Educational programme matching platform using AI to provide secondary school students with recommendations ontertiary education programmes andthe most suitable institutions based on their indicated preferences and previously selected topics of interest The output of such systems does not determine access or admission to educational programs, but rather serves as a tool to inform prospective students to support their own decision-making process on whether to apply to a specific institution or programme.
- Chatbot for admission information provided by a university to answer prospective students’ questions and to provide them with information about its admission requirements, application process, and available programs. While such chatbots are linked to the admission process, they do not make decisions about student admission s or provide personalised recommendations that could influence an admissions decision. Such a chatbot’s output is limited to providing general information and guidance, and it does not have the potential to materially influence the decision-making process.
c) Practical examples of AI systems falling within the use case but exempted by the filter mechanism in Article 6(3)
- AI systems supporting the processing of application s to be used by an education institution for file handling, such as indexing, searching, text and speech processing, as well as for the translation of documents provided along with applications, and for extracting, transforming, and organising the data collected into a meaningful and usable format . Although s uch systems support the assessment process , they do not have a relevant impact on the decision-making process, since they are only intended to facilitate the organisation and review of applicant files. Such systems therefore fall under the exception for AI systems intended to perform a preparatory task listed in Article 6(3)(d) AI Act.
- AI-enabled admissions decision review tools intended to be used by an educational institution to review and analyse ex post admissions decisions made by an admissions committee. Such systems are not intended to be used to replace or influence human assessment, but to provide a quality control check to ensure that the admissions process is functioning as intended and to support and improve future decision-making processes, rather than to make decisions. Since such systems are intended to detect decision-making patterns and deviations, and are not meant to replace or influence the previously completed human assessment, they can be considered to fall under the exception for AI systems intended to detect decision-making patterns listed in Article 6(3)(c) AI Act.
- AI-enabled admissions data organisers intended to be used by an educational institution to automatically categorise and organise incoming applications prior to the admissions process . Such systems are intended to be used to extract relevant information from unstructured application documents, such as resumes, cover letters, transcripts, and convert it into a structured format, to classify each application into one of several predefined categories, such as undergraduate, graduate, international student, and to identify and flag duplicate applications to prevent unnecessary processing. Provided the system does not take any decision on admission or access to the educational institution, it can be considered to fall within the exception for systems intended to perform a narrow procedural task of data processing and organisation listed in Article 6(3)(a) AI Act.