Künstliche Intelligenz hat Eingang ins studentische Schreiben gefunden

Nora Hoffmann, Leiterin des Schreibzentrums, über die bundesweite Studierendenbefragung »Die Zukunft des akademischen Schreibens mit KI gestalten«

UniReport: Ms. Hoffmann, the results of the student survey are now available. ChatGPT & Co have clearly arrived among today's students, with 66 percent of those surveyed stating that they have generally used generative AI writing tools. Were you surprised by the results?

Nora Hoffmann: We were actually unsure how intensively students already use generative AI, as they do not necessarily communicate this openly for fear of negative consequences, especially to lecturers. In the protected setting of writing consultations or workshops at the Writing Center, on the other hand, it was already clear in the run-up to the survey that AI has found its way into student writing. In this respect, the result was not entirely unexpected, but I still found the very large number and breadth of student users astonishing, as the topic did not seem to be equally present among all student groups beforehand. According to the survey results, students from all subject areas and semesters already use AI for numerous sub-processes of academic writing, a quarter of them at least once a week.

Participants in the study were asked about their self-assessment of writing skills. What correlations were identified with the use of AI tools?

Slight correlations can be observed across many questions: Students who rated their writing skills higher used AI slightly less often and, above all, had fewer complete texts generated by AI. They also used it for a narrower range of tasks. In the free texts, there are also repeated statements that students' own individual style is important when writing, or that they learn and develop thoughts while writing and therefore do not want to have their texts generated by AI.

Can statements be made about how students themselves assess the quality of AI-generated texts? Where do they see possible shortcomings, when would they not use AI?

According to our data, students rate the linguistic quality of the outputs highly, meaning that the use of AI to improve the language of their own texts is widespread. In terms of content quality, around half of the students perceive that AI produces incorrect statements and invents sources. They see the biggest disadvantages in the lack of transparency and lack of precision. However, this awareness is countered by the fact that the most frequent use is to have literature summarized. The free texts provide a possible explanation for this contradiction: In them, some students explain that they only use these summaries as an initial starting point for orientation and then read the texts themselves, i.e. they do not rely exclusively on AI.

What can be said about those students who are not yet using AI?

Most of these students seem to make a conscious decision against AI, as only a few state that they do not know or cannot use the tools. The reasons for rejection vary: many see no benefit in using AI, for example because they appreciate the added value of critical thinking in independent text development. Some also have reservations about using the tools, especially since at the time of the survey, most universities (including GU) did not yet have clear regulations on this. A further proportion were unsure how the tools could be integrated into the writing process. In this respect, the willingness of students to use AI is likely to increase further in the future if previous obstacles are removed by clear regulations and information provided by the universities.

Where do you still see a need for research, what in particular should writing research investigate (or have investigated) in more detail?

For the writing didactics community, teaching the reflective and targeted use of AI tools in the writing process is a (new) part of their task of teaching writing skills, as the use of AI will presumably be part of both professional and academic writing in the future (or already is). However, it is not yet possible to predict exactly what the effects of AI will be on text production processes and products, writing skills and reading habits, what support measures will be needed accordingly and how they will work, so I see a need for research here. As a first step, our survey recorded how and with what attitudes students used AI before clear regulations, information or broad didactic support were available. This makes the current starting point and need for support more tangible, although many individual questions remain unanswered, e.g: Does the joint development of ideas with AI inhibit one's own creativity or, on the contrary, does it stimulate it? Does the use of AI and the incorporation of generated formulations into one's own repertoire lead to further linguistic development or does one lose one's own skills? Is it possible to critically evaluate and revise AI-generated texts without personal writing experience? On the one hand, research should investigate such questions. On the other hand, the next step in writing didactics is to (further) develop support services based on the knowledge currently available and to research their effects. Studies on the use of AI by researchers would also be informative in order to identify (possibly subject-specific) professional usage methods that could be taught to students.

What suggestions and conclusions do you as a writing center draw from the results?

Das Schreibzentrum hat zahlreiche Informations- und Unterstützungsangebote zum Schreiben mit KI für Studierende und Lehrende entwickelt, siehe tinygu.de/ki-schreiben. Die Umfrageergebnisse nehmen einerseits etwas den Druck, das Angebot zu KI darüber hinaus stark auszuweiten, da sie zeigen, dass Studierende KI größtenteils mit ansatzweise kritischem Bewusstsein für deren Stärken, Schwächen und Gefahren nutzen. Andererseits besteht durchaus Luft nach oben: Der kritische Umgang ist noch nicht in vollumfänglichem Maße und zudem längst nicht bei allen in der Breite vorhanden. So macht die Umfrage auch Minderheiten sichtbar, die entweder KI zum Generieren vollständiger Text einsetzen oder aber die Nutzung kategorisch ablehnen. Zudem äußert die Mehrheit der Befragten hohen Bedarf nach Information und klaren Regelungen. Insofern ziehen wir als Schreibzentrum aus den Ergebnissen den Auftrag, weiterhin die technischen Entwicklungen zu verfolgen, uns in die regen schreibdidaktischen Diskurse einzubringen, Studierenden wie Lehrenden Unterstützung bei der Einbindung von KI ins akademische Schreiben anzubieten und didaktische Maßnahmen hierzu weiterzuentwickeln und zu beforschen.

Questions: Dirk Frank