Sebastian Schorcht

Sebastian Schorcht

Auto-generated profile, based on publicly available information

Sebastian Schorcht

TUD Dresden University of Technology, Dresden, Germany

Research Focus & Metrics

Academic Career Length

14Years (2013-Present)

(Calculated once Scilit publication data is synced: current year − year of first publication + 1)

Representative Publications

View All Publications
9 May 2024 in Frontiers in Education
Verified
Journal Article
Open access
SSSebastian Schorcht
Sebastian Schorcht
NBNils Buchholtz
Nils Buchholtz
LBLukas Baumanns
Lukas Baumanns
The use of and research on the large language model (LLM) Generative Pretrained Transformer (GPT) is growing steadily, especially in mathematics education. As students and teachers worldwide increasingly use this AI model for teaching and learning mathematics, the question of the quality of the generated output becomes important. Consequently, this study evaluates AI-supported mathematical problem solving with different GPT versions when the LLM is subjected to prompt techniques. To assess the mathematics educational quality (content related and process related) of the LLM’s output, we facilitated four prompt techniques and investigated their effects in model validations (N = 1,080) using three mathematical problem-based tasks. Subsequently, human raters scored the mathematics educational quality of AI output. The results showed that the content-related quality of AI-supported problem solving was not significantly affected by using various prompt techniques across GPT versions. However, certain prompt techniques, particular Chain-of-Thought and Ask-me-Anything, notably improved process-related quality.

24 October 2024 in Digital Experiences in Mathematics Education
Verified
Journal Article
Open access
PDF file
SSSebastian Schorcht
Sebastian Schorcht
FPFranziska Peters
Franziska Peters
JKJulian Kriegel
Julian Kriegel
This study explores the application of communicative AI agents, specifically a network of customized generative pretrained transformer agents, in designing mathematical tasks. It focuses on how these AI agents, functioning as a multi-professional team, can perform mathematical task design (concerning a collection of task activities and not curriculum materials/textbooks) through collaborative and context-aware communication. Concentrating on four perspectives—mathematical depth, language sensitivity, natural differentiation, and competence orientation—four different AI agents were instructed to evaluate and modify six mathematical tasks based on individual research knowledge bases. In a consensus-seeking process, the AI agents were connected via a chat chain, prompting multiple iterations to modify the tasks. The output (six AI-modified tasks) was then evaluated by six in-service teachers as human experts by making them choose blindly between the original and the AI-modified tasks and by then analyzing the additional comments to their decisions in qualitative content analysis. Furthermore, the AI-modified tasks were rated on a multidimensional Likert scale. The results indicate that for the AI-modified tasks, achieving a balance between substantial text generation and precise task formulation is crucial and was not always found in the GPT network output. At the same time, the combination of the four AI agents was able to enrich the tasks with potential solution approaches and specific calls to action.

7 April 2026 in ZDM – Mathematics Education
Verified
Journal Article
Open access
SSSebastian Schorcht
Sebastian Schorcht
FMFabian Anton Müller
Fabian Anton Müller
NBNils Buchholtz
Nils Buchholtz
Large Language Models (LLMs) are increasingly discussed as tools for supporting mathematical problem solving. However, existing research predominantly evaluates LLM performance in terms of correctness, while the mathematics educational quality of AI-generated worked-out solutions remains largely unexplored. This study investigates how the mathematics educational quality (a set of quality criteria relating to teaching and learning mathematics) of AI-generated problem solutions can be influenced by model choice and prompt engineering. Four LLMs (Gemini 1.5 Pro (Advanced), Claude 3.5 Sonnet, ChatGPT-o3 mini, and DeepSeek-R1) were tested on six problem-solving tasks from the domains of number and algebra using four prompt techniques (Zero Shot, Chain of Thought, Persona, and Retrieval-Augmented Generation). In total, 2880 solutions were analyzed, combining human expert coding of content-related, process-related, and pedagogical-contextual quality with binary logistic regression. Results show that content-related quality is mainly driven by model type and task characteristics, whereas process-related and pedagogical-contextual quality depend strongly on prompt design, particularly Persona prompting. Overall, no single model or prompt technique performs optimally across all dimensions, indicating that effective educational use of LLMs requires context-sensitive combinations of models, prompts, and tasks.

26 September 2024 in Journal of Mathematics and the Arts
Verified
Journal Article
Open access
SSSebastian Schorcht
Sebastian Schorcht
LBLukas Baumanns
Lukas Baumanns
NBNils Buchholtz
Nils Buchholtz
JHJudith Huget
Judith Huget
FPFranziska Peters
Franziska Peters
MPMaximilian Pohl
Maximilian Pohl

Verified
Conference Paper
Closed access
PDF file
SSSebastian Schorcht
Sebastian Schorcht

6 July 2013
Verified
Book Chapter
Closed access
PDF file
SSSebastian Schorcht
Sebastian Schorcht