Research Institute

AI Research Institute der IU

AI systems are transforming how people learn, work, and make decisions. Yet the most consequential questions are not about the technology itself — they are about the humans using it. How does a nurse, a marketing analyst, or a student decide when to involve AI, how far to trust its output, and when to take back control? How do expert judgement, problem-solving ability, and professional identity develop, or erode, as AI use becomes sustained and routine?

This long-term operational dimension of human-AI collaboration remains underresearched. The AI Research Institute at IU exists to change that.

We research at the intersection of human and AI with the people who practice it daily.

As a university of applied sciences, our approach is by design: our research is born in real learning and working environments, not in controlled laboratory conditions. What positions us uniquely is direct access to over 100,000 students already operating in an AI-saturated academic environment. This enables what few institutes can offer: longitudinal study of competence development and human-AI collaboration under real conditions — producing evidence that is relevant to both education and professional practice.

The decisive factor in successful human-AI collaboration is not the technology, it is the human capacity to decide, situationally, when and how to involve AI, and how to orchestrate that collaboration well. This capacity can be taught, measured, and developed. That is our research mandate.

The project investigates the use of a self-developed AI chatbot to encourage professional self-reflection among distance learning students at IU International University. It focuses both on the scientific analysis of user experience and interaction patterns with the chatbot, as well as the didactic integration of the chatbot into teaching and learning in distance learning settings at university level. The project builds on existing research about AI in higher education and the importance of self-reflective skills for professional pedagogical development. Empirical findings will be collected to help improve the design of reflection processes supported by chatbots.

Fairgrade investigates whether AI can support the evaluation of full academic theses — a task that resists automation because it depends on judgment across many interacting dimensions. Rather than scoring isolated features, the project treats thesis assessment as an integrated scholarly judgment that AI can make more explicit and consistent, but never replace. The examiner remains the authority at every stage; the system's role is to surface evidence, not to decide.

This project investigates how individual personality traits influence interactions with AI-based learning environments and how these insights can be used to improve learning outcomes. Building on the Big Five personality model, the project systematically analyzes how different personality dimensions affect engagement, satisfaction, and learning success when interacting with LLM-based systems. Based on these insights, the project explores the design of personality-adaptive chatbot interactions that dynamically adjust to users’ interaction styles. The goal is to develop more personalized and effective AI-supported learning experiences that enhance both user experience and educational outcomes.

The project examines which learning strategies in distance learning actually contribute to learning success—and how strongly this depends on the personality structure and life situation of students. It is based on a longitudinal, psychometrically validated recording of learning behaviour: which offerings are used, how do they affect learning success, and which characteristics of learners influence both? Particular attention is paid to recurring highs and lows over the course of study, such as drops in motivation or changes in learning strategies. From these patterns, the project derives intervention strategies to prevent student dropout. Methodologically, it relies on theory-guided index construction, factor and reliability analyses, as well as causal analyses using a wait-list control group design.