Schlagwort: AI

Veröffentlicht am

[publication] Prompting.School: A Design-Based Research Approach to Teaching AI Literacy Through Guided Prompt Engineering Practice #AIinEducation #tugraz #AI

Our publication (presented at least year ICL conference) about „Prompting.School: A Design-Based Research Approach to Teaching AI Literacy Through Guided Prompt Engineering Practice“ is published now.

Abstract:
We introduce prompting.school, a browser-based learning platform that teaches foundational prompt engineering skills to support AI literacy. Developed using a design-based research approach, the platform was evaluated through four iterative cycles with pre-service and in-service teachers, as well as vocational learners. Aligned with the UNESCO AI Competency Framework and DigComp 2.3 AT, it combines scaffolded lessons, interactive feedback, and real-time AI responses to promote self-regulated learning. Findings show that structured, hands-on prompting practice enhances learners’ understanding, with notable differences in engagement and performance across educational contexts.

[publication @ publisher’s homepage]
[draft @ ResearchGate]

Reference: Brünner, B., Jahic, I., Ebner, M. (2026). Prompting.School: A Design-Based Research Approach to Teaching AI Literacy Through Guided Prompt Engineering Practice. In: Auer, M.E., Toth, P. (eds) Innovation via Collaborative Learning in Engineering Education. ICL 2025. Lecture Notes in Networks and Systems, vol 1848. Springer, Cham. [https://doi.org/10.1007/978-3-032-20381-6_62]

This is an impactful contributions, methodological rigor, and exceptional novelty in the research field of AI in education.

Veröffentlicht am

[publication] Development and Evaluation of an AI-Supported Teacher Assistant Tool (AI-TAT) to Foster Self-Regulated Learning #tugraz #AIinEducationa

Our publication titled „Development and Evaluation of an AI-Supported Teacher Assistant Tool (AI-TAT) to Foster Self-Regulated Learning“ has been published.

Abstract:
Self-regulated learning (SRL) is increasingly recognised as a key competence for students, yet many teachers struggle to integrate SRL concepts into concrete classroom materials like worksheets. In this paper, SRLly is presented, an AI-supported teacher assistance tool (AI-TAT) that helps teachers revise existing worksheets to better scaffold SRL processes. Rather than submitting entire documents to a large language model (LLM), SRLly performs local preprocessing of uploaded PDFs to extract text with positional metadata, segments the worksheet into meaningful blocks, and sends only relevant blocks for analysis. Based on this structured representation, the system generates actionable recommendations aligned with four SRL dimensions (planning, strategy, monitoring, reflection). Recommendations are presented in an easy-to-use interface that supports teachers’ workflows by providing short rationales and concrete recommendations with ready-to-use text snippets that can be copied. We conducted an initial evaluation during a teacher professional development training with 30 teachers at an Austrian school (student age range 10–18). Overall perceptions were positive: 76% of participants rated SRLly as good or very good. Likert-scale results indicated high perceived usefulness and feasibility, while trustworthiness ratings were more mixed, highlighting reliability of AI output as a key area for improvement. Our findings suggest that structured, material-centred AI feedback can support teachers in integrating SRL into everyday worksheet design, while underscoring the need for stronger reliability and real-world validation.

[full article @ publisher’s homepage]
[draft @ ResearchGate]

Reference: Geier, G., Ebner, M., Burgsteiner, H. (2026). Development and Evaluation of an AI-Supported Teacher Assistant Tool (AI-TAT) to Foster Self-Regulated Learning. In: Smith, B.K., Borge, M. (eds) Learning and Collaboration Technologies. HCII 2026. Lecture Notes in Computer Science, vol 16733. Springer, Cham. https://doi.org/10.1007/978-3-032-30784-2_12

Veröffentlicht am

[publication] On Using an LLM-Based Approach for Automatic Evaluation of Programming Tasks in MOOC Videos #tugraz #AIinEducation

Our publication about „On Using an LLM-Based Approach for Automatic Evaluation of Programming Tasks in MOOC Videos“ has been published.

Abstract:
Interactive videos promote learner engagement by transforming the passive activity of watching videos into an active learning process. LIVE, a web platform developed in previous research that allows for the integration of interactive elements into videos, also supports embedding programming tasks. However, at the beginning of our research, the output of the submitted solution was automatically compared to the reference output, but the code itself still had to be checked manually, as no automatic process was available to verify the solution and provide feedback. To address this, a stand-alone application was developed that automatically evaluates students’ solutions to Python programming tasks using a self-hosted large language model. This application was then integrated into the existing platform. To test this approach, a massive open online course containing four interactive videos with several programming tasks and quizzes was created. Students’ performance was analyzed using quiz results, task submissions, self-assessments, and user statistics provided by the course platform and LIVE. The results indicate that utilizing a large language model for automated feedback can positively impact learning outcomes. However, some limitations must be taken into account, such as the need to correct some of the model’s generated outputs manually due to insufficient feedback quality for certain tasks. This suggests that the combination of automated and manual evaluation seems to be the most effective approach for now.

[full article @ publisher’s homepage]
[draft @ ResearchGate]

Reference: Kindlhofer, C., Wachtler, J., Ebner, M. (2026). On Using an LLM-Based Approach for Automatic Evaluation of Programming Tasks in MOOC Videos. In: Smith, B.K., Borge, M. (eds) Learning and Collaboration Technologies. HCII 2026. Lecture Notes in Computer Science, vol 16733. Springer, Cham. https://doi.org/10.1007/978-3-032-30784-2_13

Veröffentlicht am

[publication] How Embodiment Changes AI Tutoring: Comparing Student Perceptions of Text and AI Human Avatar-Based Chatbot Systems #AIinEducation #chatbot #tugraz

Our publication titled „How Embodiment Changes AI Tutoring: Comparing Student Perceptions of Text and AI Human Avatar-Based Chatbot Systems“ is now published.

Abstract:
Embodied conversational interfaces are able to deliver generative AI tutoring via synthesised speech and human-like video avatars. However, there is limited empirical evidence on how embodiment shapes the learner experience and self-regulated learning (SRL). This study compares the perceptions of students interacting with a text-based Retrieval-Augmented Generation (RAG) tutor and an avatar-based RAG tutor in two authentic higher education contexts (introductory computer science and Python programming) across two European universities. Adopting a Design-Based Research approach, we collected (1) questionnaire data from users and non-users, and (2) anonymised interaction logs comprising 341 user messages, which were coded using an SRL process–action framework. The results show that the perceived level of learning support is largely consistent with previous text-only deployments: the tutor is primarily valued for providing explanations and support related to practice, while motivational support is still comparatively weak. However, embodiment notably improves perceptions of the system as a communication partner, with active users rating this dimension more positively than non-users. Modality ratings suggest that audio output is perceived as more helpful than the visual avatar. Log analyses show that interactions are still dominated by information-seeking (Seeking–Search) and that metacognitive SRL actions (e.g. goal setting and self-evaluation) are rare. Embodiment also introduces additional peripheral interactions (e.g. character-focused prompts and human-likeness checks), thereby increasing the proportion of non-learning dialogue. Overall, the findings suggest that, while embodiment enhances social presence, it does not automatically foster deeper SRL. Explicit pedagogical scaffolding is needed to encourage learners to engage in metacognitive regulation.

[full paper @ publisher’s homepage]
[draft @ ResearchGate]

Reference: Brünner, B., Nestler, M., Pucher, L., Ebner, M. (2026). How Embodiment Changes AI Tutoring: Comparing Student Perceptions of Text and AI Human Avatar-Based Chatbot Systems. In: Smith, B.K., Borge, M. (eds) Learning and Collaboration Technologies. HCII 2026. Lecture Notes in Computer Science, vol 16731. Springer, Cham. https://doi.org/10.1007/978-3-032-30542-8_25

This is an impactful contributions, methodological rigor, and exceptional novelty in the research field of AI in education, Chatbots and MOOCs.

Veröffentlicht am

[blogpost] Generative KI in der Hochschullehre – Wie wir Innovation verantwortungsvoll gestalten #hfd #tugraz #AI

Wir haben einen kurzen Blogpost über „Generative KI in der Hochschullehre – Wie wir Innovation verantwortungsvoll gestalten,“ beim Hochschulforum Digitalisierung verfasst um auch auf unsere Whitepaper hinzuweisen:

Dieser Blogbeitrag von Philipp Leitner, Benedikt Brünner, Martin Ebner und Sandra Schön bietet Hochschulen eine kompakte Anleitung, wie sie generative KI überlegt in der Lehre einsetzen können. Die Tipps wurden aus der Praxis an der TU Graz entwickelt und getestet und werden ergänzt mit Beispielen sowie Hinweisen auf nützliche Tools.

[Link zum Blogpost]

Whitepaper: Schön, S., Brünner, B., Ebner, M., & Leitner, P. (2026). Evaluating GenAI Innovation in Higher Education. A Whitepaper (Version 1.0 – May 22, 2026). Graz University of Technology. https://doi.org/10.3217/1tfkj-ntj85

Veröffentlicht am

[publication] Empowering Preservice Biology Teachers as Designers of Pedagogical AI Agents: Moving Beyond Data Feeding toward Pedagogical Design #research

Our article about „Empowering Preservice Biology Teachers as Designers of Pedagogical AI Agents: Moving Beyond Data Feeding toward Pedagogical Design“ is published and available online.

Abstract:
Artificial intelligence (AI) influences learning processes, student activities, and the organization of teaching, both in general and specifically in science education. Research indicates that the way AI will affect science learning primarily depends on how it is implemented in teaching. This study examines the ability of preservice biology teachers (PSBTs) to independently develop pedagogical AI agents (TDP-AI agents), most often in the form of chatbots, tailored to teaching goals, content, and students‘ needs. This empirical study, based on a mixed-methods research approach and applying the ICAP theoretical framework, analyzed 54 lesson plans. In addition, the study explored PSBTs‘ perceptions regarding the development and contribution of TDP-AI to biology teaching. The results reveal three patterns of AI agent application in lesson plans: (a) intensive use of TDP-AI agents across all lesson phases, promoting students‘ cognitive engagement, (b) moderate and selective use focused on core activities, and (c) limited use mainly in introductory segments. Five thematic areas reflect PSBTs‘ perspectives: (1) personalization and flexibility of learning, (2) enhancement of student motivation and engagement, (3) development of teachers‘ digital and pedagogical competencies, (4) technical and resource-related challenges in agent development, and (5) pedagogical-methodological barriers in designing student-AI interactions. The findings emphasize the need for systematic support for pre-service teachers in developing digital tools and building pedagogical approaches to ensure AI is used ethically and effectively in science education.

[full article @ publisher’s homepage]
[full article @ ResearchGate]

Reference: Anđić, B., Helm, C., Weinhandl, R. et al. Empowering Preservice Biology Teachers as Designers of Pedagogical AI Agents: Moving Beyond Data Feeding toward Pedagogical Design. Res Sci Educ (2026). https://doi.org/10.1007/s11165-026-10359-0

This is an impactful contributions, methodological rigor, and exceptional novelty in the research field of AI in education.

Veröffentlicht am

[publication] Whitepaper on Evaluating GenAI Innovation in Higher Education #tugraz #research #AIinEducation

Our publication, „Whitepaper on Evaluating GenAI Innovation in Higher Education,“ is included in the proceedings of the EDMedia 2026 conference.

Abstract:
Generative artificial intelligence (genAI) is increasingly shaping higher education by enabling new forms of content creation, assessment, learner support, personalization, and synthetic media. However, its value cannot be determined solely by technical performance, novelty, or efficiency. This paper presents an evaluation framework for quality genAI applications in higher education. This paper presents a practice-derived framework developed through a cross-case synthesis of nine diverse GenAI implementations at Graz University of Technology. Analyzing projects ranging from AI-generated content to RAG-based chatbots, we identified recurring decision points and risk patterns to formulate a five-phase, non-linear evaluation model. The framework guides institutions through specifying context, assessing feasibility, selecting implementation strategies, conducting multi-layered pilots, and performing data-informed analysis. A defining feature is the integration of explicit „Sustain-or-Discontinue“ decision gates at each stage, ensuring resources are committed only to viable applications while providing a structured pathway to terminate initiatives that fail to meet pedagogical or ethical standards. We argue that the core institutional capability lies not in rapid adoption, but in the discipline to evaluate responsibly, balancing experimentation with the rigor to govern and discontinue when necessary.

[publication @ conference’s homepage]
[preview @ ResearchGate]

Reference: Schön, S., Brünner, B., Ebner, M. & Leitner, P. (2026). Whitepaper on Evaluating GenAI Innovation in Higher Education. In Proceedings of EdMedia 2026 Edinburgh (pp. 697-703). Waynesville, NC: Association for the Advancement of Computing in Education (AACE). Retrieved June 15, 2026 from https://www.learntechlib.org/primary/p/2129693/.

This is an impactful contributions, methodological rigor, and exceptional novelty in the research field of AI in education.

Veröffentlicht am

[mooc] Meinungsbildung & Desinformation #imoox

Ein sehr spannender MOOC geht heute auf iMooX.at online, zum Thema „Meinungsbildung & Desinformation“ von der MCI:

Desinformation ist aktueller denn je – von polarisierenden Informationen zu Wahlkämpfen bis zu KI-generierten Deepfakes. Unser neuer Selbstlernkurs „Meinungsbildung und Desinformation“ stärkt Informationskompetenz und unterstützt dich dabei, eine reflektiert-fragende Haltung gegenüber Medieninhalten zu entwickeln. Teilnehmende lernen, Informationen kritisch einzuordnen, Quellen zu prüfen und manipulative Inhalte zu erkennen – eine Schlüsselkompetenz für demokratische Teilhabe und Verantwortung.

Natürlich ist der MOOC kostenslos durchführbar: [Link zur kostenlosen Anmeldung]

Veröffentlicht am

[publication] Sieben Mythen der KI-Nutzung #tugraz

Unser Beitrag zu „Sieben Mythen der KI-Nutzung“ hat viele Reaktionen hervorgerufen und nun wurde er auch in die Zeitschrift „Die Österreichische Volkshochschule“ aufgenommen.

Abstract:
Wer das Internet nutzt, kommt im Frühjahr 2026 nicht um Anwendungen generativer Künstlicher Intelligenz (kurz KI) herum. Suchmaschinen bieten neben Links standardmäßig KI-generierte Antworten an, Chatbots unterstützen bei der Buchung von Websites, Schüler:innen lassen sich Tests passend zu den Arbeitsblättern der Lehrer:innen generieren usw. – Doch nicht alles, was uns die KI-Anwendungen liefern, wie wir sie nutzen und ihre Ergebnisse verstehen, ist zutreffend und unproblematisch. Das liegt auch an Missverständnissen darüber, wie KI-Anwendungen funktionieren. Aus unserer Sicht – es gibt dazu noch keine empirische Evidenz – verdienen folgende sieben Aussagen besondere Aufmerksamkeit, insbesondere auch im Kontext von Bildung, Schule und Hochschule:

  1.  KI-Anwendungen sind neutral, objektiv und vorurteilsfrei
  2.  KI-Anwendungen arbeiten logisch
  3. KI-Anwendungen denken und lernen wie Menschen
  4. KI-Anwendungen sind empathisch
  5. KI-Anwendungen sind ökologisch und sozial problemlos
  6. KI-Nutzung ist rechtlich einwandfrei
  7. KI-Anwendungen machen Wissen und Kompetenzentwicklung überflüssig

In diesem Beitrag möchten wir diese als „Mythen“ bezeichneten Aussagen beschreiben und aufzeigen, dass und warum sie nicht zutreffend sind. Damit möchten wir einen zukünftig fundierten Umgang und durch die Beschreibung von KI-Mythen Forschung dazu initiieren und unterstützen. 

Referenz: Schön, S., Brünner, B., Ebner, M., Diesenreither, S., Hanfstingl, B., Krammer, G (2026) Sieben Mythen der KI-Nutzung. Die Österreichische Volkshochschule. Jg. 2026 / 286. [Link]

This is an impactful contributions, methodological rigor, and exceptional novelty in the research field of AI in education.

Veröffentlicht am

[publication] Generative AI Chatbots in Secondary Mathematics Education: Development and Implementation of a Dynamic Large Language Model-Based Learning Assistant for Quadrilaterals #tugraz

Our contribution titled „Generative AI Chatbots in Secondary Mathematics Education: Development and Implementation of a Dynamic Large Language Model-Based Learning Assistant for Quadrilaterals“ is now published.

Abstract:
As artificial intelligence becomes more and more a part of education, the challenge is not about having access to generative tools, but about connecting them with the goals of the curriculum and the needs of the classroom. This chapter presents the design and evaluation of a large language model–based chatbot developed specifically for teaching quadrilaterals in lower secondary mathematics. The chatbot integrates fine-tuning with retrieval-augmented generation (RAG), combining accurate, curriculum-aligned content with flexible, conversational support. The chatbot allows learners to ask conceptual questions, solve problems step by step, receive guided hints, and generate flashcards or exercises of varying difficulty. A hybrid routing mechanism selects the most appropriate response strategy based on user intent. Evaluations using both isolated prompts and multi-turn dialogues demonstrate that the hybrid system significantly outperforms standard LLM baselines in terms of accuracy, consistency, and pedagogical suitability. A classroom trial with 20 students confirmed the tool’s usability and effectiveness; students reported high satisfaction and meaningful engagement. This study demonstrates that, with careful content and architectural structuring, generative AI can enhance student learning while supporting differentiated instruction. Future directions include scaling the approach to other topics and incorporating multimodal capabilities.

[full article @ publisher’s homepage]
[draft @ ResearchGate]

Reference: Mallweger, M., Brünner, B., Ebner, M. (2026). Generative AI Chatbots in Secondary Mathematics Education: Development and Implementation of a Dynamic Large Language Model-Based Learning Assistant for Quadrilaterals. In: Auer, M.E., Nikou, S.A. (eds) GenAI in Novel Educational Applications. Studies in Computational Intelligence, vol 1260. Springer, Cham. https://doi.org/10.1007/978-3-032-16153-6_7

This is an impactful contributions, methodological rigor, and exceptional novelty in the research field of AI in education.