details
| Title |
AI-based STEM education. Generating individualized exercises in mathematics |
|---|---|
| Authors |
Lange-Hegermann, Markus   ;
Schmohl, Tobias ;
Watanabe, Alice ;
Heiss, Stefan ;
Rubart, Jessica |
| Source | Pixel [Hrsg.]: Conference proceedings. International conference "New perspectives in science education", 10th edition. Virtual event, 18-19 march 2021. Bologna : Filodiritto Editore 2021, 4 S. |
| Document | full text (329 KB) |
| License of the document |
|
| Keywords (German) | Künstliche Intelligenz; Aufgabenstellung; STEMS <Science, Technology, Environment in Modern Society>; Technologieunterstütztes Lernen; Selbst gesteuertes Lernen; Hochschule; Studium; Algorithmus; Bayes-Formel; Mathematik; Lernsystem; Learning Analytics |
| sub-discipline | Media Education Higher Education |
| Document type | Article (from a serial) |
| Language | English |
| Year of creation | 2021 |
| review status | Peer-Reviewed |
| Abstract (English): | During the COVID-19 pandemic, the trend towards self-directed learning at universities received a strong boost. However, some students show considerable deficits regarding their self-learning competences. These become especially apparent in the first semesters, creating gaps in the studentsʼ knowledge which will not only slow down their progress in later semesters but may even lead to their dropping out of university altogether. For this reason, several approaches in the field of mathematics teaching attempt to prevent knowledge gaps from the very first week of studies, usually by employing educational instruments such as peer feedback or corrected homework. Despite these efforts, dropout rates in STEM subjects remain high. We propose to address this problem with an instructional design based on AI algorithms which create mathematical exercises, tailoring their degree of difficulty individually to fit each studentʼs skills and speed. Our hypothesis is that this individualized training will keep students from feeling overwhelmed and increase their motivation to study. As the exercises depend on many parameters to determine the appropriate degree of difficulty, they are adjusted iteratively, based on final or intermediate results of previously processed tasks and Learning Analytics data through Bayesian optimization. (DIPF/Orig.) |
| Abstract (German): | Die Tendenz zum selbstgesteuerten Lernen an den Hochschulen wird durch die COVID-19-Pandemie noch verstärkt. Allerdings weisen gerade in der Studieneingangsphase einige Studierende erhebliche Defizite in Bezug auf die Selbstlernkompetenzen auf. Deshalb wird in der Mathematiklehre ab der ersten Studienwoche versucht, Wissenslücken durch Feedback in Kleingruppen oder durch die Korrektur von Hausaufgaben zu verhindern. Die dennoch hohe Zahl der Studienabbrecher in den MINT-Fächern steht im Gegensatz zum wachsenden Bedarf der Wirtschaft an qualifizierten Absolventen. Wir schlagen vor, diese Abbrecherquote durch ein auf KI-Algorithmen basierendes Unterrichtsdesign anzugehen, das mathematische Aufgaben mit einem maßgeschneiderten, individuellen Schwierigkeitsgrad für Studierende erstellt. Unsere Hypothese ist, dass diese Intervention dem selbst eingeschätzten Gefühl der Überforderung entgegenwirkt und die individuelle Studienmotivation erhöht. Die Aufgaben hängen von vielen Parametern ab, die den Schwierigkeitsgrad bestimmen. Diese werden iterativ angepasst, basierend auf End- oder Zwischenergebnissen von zuvor bearbeiteten Aufgaben und Learning Analytics-Daten durch Bayes'sche Optimierung. (Autor) |
| Statistics |  Number of document requests
|
| Checksums | checksum comparison as proof of integrity |
| Date of publication | 10.11.2023 |
| Citation | Lange-Hegermann, Markus; Schmohl, Tobias; Watanabe, Alice; Heiss, Stefan; Rubart, Jessica: AI-based STEM education. Generating individualized exercises in mathematics - In: Pixel [Hrsg.]: Conference proceedings. International conference "New perspectives in science education", 10th edition. Virtual event, 18-19 march 2021. Bologna : Filodiritto Editore 2021, 4 S. - URN: urn:nbn:de:0111-pedocs-279445 - DOI: 10.25656/01:27944 |
share content at social platforms
Teile diese Seite:
