A study by the Center for Higher Education Development (CHE) shows that artificial intelligence (AI) is increasingly becoming part of everyday student life at universities: Approximately two-thirds of students use AI tools like ChatGPT at least once a week, and a quarter even use them daily. However, this use also has its downsides: Students' ability to write texts independently has declined significantly as a result. This is where the THMresearch project TextAid comes in.
for years at the Department intelligent tutoring system, called the Feedback System, , is designed to support students in writing academic texts with targeted, learning-enhancing feedback. The idea, conceived by Prof. Dr.Prof. Dr.Prof. Dr. Prof. Dr.Prof. Dr.Prof. Dr.Prof. Dr. Prof. Dr.Prof. Dr.Business Information SystemsBusiness Information Systems Business Information SystemsBusiness Information SystemsBusiness Information SystemsBusiness Information Systems Business Information SystemsBusiness Information Systems, describes his observations from teaching: "Students are already using AI and having their texts written by AI – it's better if they receive AI-assisted guidance on how to improve, rather than simply being given finished texts." Unlike traditional AI tools such as ChatGPT, TextAid is intended to accompany the learning process.
The team also places particular emphasis on data protection: "Our servers aren't located somewhere in the USA, but right here at THM," Kammer emphasizes. "TextAid will be a free, data protection-compliant, THMinternal writing skills coach. This coach will be able to respond to students' questions and continue learning independently," adds Prof. Dr. Markus Siepermann, also a project manager and professor of Business Information Systems .
Pia Georgiew, a doctoral candidate and project staff member, explains: “In the early stages of writing, similar, easily recognizable errors often occur, which are then corrected time-consumingly with standard comments from instructors. This is where TextAid comes in: It provides support for these beginner errors and thus relieves the burden on instructors.” TextAid is intended to act as an upstream tool for instructors: “We want to free up more time for the essentials.”
The added value for students: immediate feedback, available around the clock and accessible as often as needed. In addition to checking formal quality aspects such as spelling, punctuation, consistent abbreviations, and instructor-specific guidelines like citations, TextAid aims to support the precise formulation of research questions and titles, provide guidance on improving argumentation structure, and thus help enhance the content and structure of student work. The AI tutor will also offer simple tips on transitions, comprehensibility, and clarity.
But how exactly does the system work technically? "First, the text document is broken down into different chapters – subchapters, paragraphs, and sentences," explains Georgiew. The text is then summarized and tagged. Specialized AI models – each trained in a specific area – then analyze the blocks for relevance, comprehensibility, and language, and identify any errors. The feedback generator summarizes the errors, prioritizes problems, and generates concrete, learning-enhancing feedback, prioritized by importance. This enables adaptive learning. "The goal is also to add a chat function, giving students the opportunity to ask questions," reports Kammer.
TextAid will receive approximately €370,000 in funding over two years from the Foundation for Innovation in Higher Education (StIL) as part of its "Freiraum" funding program. The project is part of the further development of the feedback system that supports students in a wide variety of courses. TextAid builds on its predecessor, Easy Tutor, which provides students with intelligent feedback on SQL and Excel exercises.
One of the biggest challenges is training the AI and gathering enough examples of feedback from academic texts. Therefore, the team is asking for your help: Students and instructors can submit drafts of seminar papers and theses with suggestions for improvement (for example, in the form of Word comments) to further enhance the AI's foundation https://frits.mni.thm.de/s/text-mit-kommentaren: