Abstract:

Artificial intelligence (AI) is finding its way into universities, whether planned by university management or unplanned through individual initiatives, whether in learning support systems, student counselling or administrative processes in campus management. While companies can follow a technology-first approach, in which new workflows are developed first and then evaluated within regulatory frameworks, public institutions such as European universities face the opposite challenge: they must consider AI systems in the context of institutional responsibility from the very beginning, without knowing in advance what technological capabilities will emerge. This is made more difficult by the fact that AI systems are increasingly evolving from isolated assistance systems to generalised and partially autonomous applications. Despite this challenge, there is currently no generally established conceptual framework that systematically classifies different AI systems in the university context and makes them comparable.

This article addresses this research gap by developing the RAGI model, a classification framework for AI systems based on the performance dimensions of autonomy (A), generality (G) and intelligence (I), as well as the meta-dimension of responsibility (R). Building on an analysis of existing classifications, it is shown that autonomy, generality and intelligence are recurring dimensions for describing the capabilities of AI systems and – supplemented by the perspective of responsibility – are suitable for systematically classifying AI systems in a higher education context.

On this basis, the possible combinations of the four dimensions are systematically discussed and illustrated using exemplary applications in higher education and campus management, such as those from the perspective of applicants, students, lecturers, administrative staff or management. For example, some AI systems combine autonomy and intelligence without being general - such as room optimisation systems - while others combine generality and intelligence but lack autonomy, such as learning chatbots. This reveals how different AI applications differ in terms of their capabilities and benefits, but also in terms of their risk profiles. The paper therefore examines whether for reasons of responsibility certain combinations of the three capability dimensions should be limited, particularly those leading to highly autonomous and general intelligent systems (AGI).

The article shows that the RAGI model enables a structured classification of AI systems and thus creates a basis for comparing different AI applications in a higher education context. The central scientific contribution lies in the introduction of a simple but extendable dimension model that links the technical capabilities of AI systems with their institutional application context, thus enabling a systematic analysis of AI applications, even outside higher education contexts.

Keywords:

Artificial Intelligence, AI Classification, Responsible AI, AI Capability Framework, RAGI Model, Higher Education, Campus Management, AI Systems, Institutional Responsibility, AI Governance.