Posterbeiträge
Autor*innen: Anja Matthiä, Nandi Joubert, Adriana Fernandes Veludo, Aline Veillat, Aurelio di Pasquale, Kristina Pelikan, Julia Bohlius (Swiss Tropical and Public Health Institute, Allschwil, Switzerland and University of Basel, Basel, Switzerland)
The availability of generative artificial intelligence (GenAI) started a new era of how we perform research today and in the near future. This new era comes with uncertainties for researchers on how to apply GenAI in their work. To address these uncertainties, we started a Community of Practice (CoP) to bring together early career and experienced researchers from diverse disciplines to explore the potential of GenAI. We aim to advance our knowledge, share experiences and transfer learnings to our daily research activities.
This CoP functions as a collaborative space for its members to practice applying GenAI to various research processes like acquiring research knowledge, literature review, research planning and study design, data analysis, or scientific writing. The CoP will begin with training sessions on the use of GenAI and will encourage group discussions on related challenges such as ethical considerations, hallucinations, dependencies, and data privacy.
Over time, CoP members will develop guidelines and best practices for the critical use of GenAI in research in a co-creation process that integrates perspectives from all participants, drawing on their diverse backgrounds and expertise. Guidelines and best practices remain flexible to accommodate the rapid evolution of GenAI.
By its activities, CoP members will gain the skills and confidence to use GenAI in their daily research aligned with academic principles and scientific integrity. By effectively communicating CoP outcomes, the community will grow and progressively transfer the established practices into institutional processes, ensuring sustainability.
Autor*innen: Jean Terrier (Digital Literacies/Bildungstechnologien); Barbara Berzel, Anita Kovács, Beatrice Mall-Grob, Stephan Meyer (Language Center); Anja Matthiä (Swiss Tropical and Public Health Institute); Mirjam Weder (Fachbereich Deutsche Sprach- und Literaturwissenschaft)
Academic writing contributes to cognition and knowledge creation. It relies on tools and materials which impact on writing processes and products. This impact has taken on a fundamentally different quality with the amplification of the digital turn through generative AI. Large Language Models call into question established assumptions, necessitating a thorough review of writing development and research, and of the writing competences needed in future.
Wissenschaftliches Schreiben mit digitalen Tools – mehr als ChatGPT, DeepL und Co. is a multidisciplinary, German/English twelve-lecture series with practical exercises at the intersection of academic writing development and digitisation aimed at enhancing the required competences among Bachelor and Master students. This introduction to aspects of general academic writing trialled in HS2024 at the University of Basel, familiarises students with some of the requirements of and processes in academic writing and research and guides them in the practical application the tools that support each writing stage.
Accordingly, it covered topics such as: models and phases of academic writing, searching for and managing literature with bibliographic tools, plotting a structure with phrasebanks and a thesis writing tool, the underlying principles of large language models, relations between authors and tools, effective prompting, formulating research questions, academic argumentation, visualisation, editing and productivity tools, as well as issues related to ethics, the law and academic integrity.
Beyond the instructional setting, this intervention suggests a blueprint for future curricula and contributes to the ongoing multi-stakeholder process in formulating guidelines, designing policies and developing writing competences for current and future higher education.
Autor*innen: Tafadzwa Dhokotera, Nandi Joubert, Aline Veillat, Jan Hattendorf, Julia Bohlius (Swiss Tropical and Public Health Institute, Associated Institute of the University of Basel)
Randomised controlled trials (RCTs) showed that generative Artificial Intelligence (GenAI) can improve but also weaken learning of statistics/mathematics. We conducted a RCT to assess the performance of GenAI when utilised by humans as collaborative or as standalone analytical tool for epidemiological data analyses.
We invited MSc/PhD students and lecturers/ supervisors in the field of epidemiology at Swiss Tropical and Public Health Institute. Participants were randomized to analyse a simulated cross-sectional study dataset with a statistical analysis software (Stata/R) supported by ChatGPT-3.5 (=ChatGPT collaborator arm); or using ChatGPT-4 with integrated data analysis function (=ChatGPT analyst arm).
Of 338 invited individuals 31 participated in the RCT: 16 were randomized to ChatGPT collaborator arm, 15 to ChatGPT analyst arm. There was no evidence for a difference in the mean task scores between the two arms. A higher proportion of participants in the ChatGPT analyst arm correctly presented key descriptive analyses as well as meaningful and correctly labelled visualizations for data/results compared to those in the ChatGPT collaborator arm. Regarding inferential statistics, a greater proportion of participants in the ChatGPT collaborator arm selected the correct model, correctly evaluated categorical and confounding variables, and presented the results accurately compared to the ChatGPT analyst arm.
This RCT highlighted the potential of GenAI as a valuable collaborator in epidemiological data analysis. While advanced AI tools offer significant advantages, their effective use requires balancing of AI capabilities and human expertise. Integrating generative AI with traditional learning methods is essential to ensure it supports, rather than obstructs, the path to expertise.
Autor*innen: Aline Veillat, Karin Gross, Tafadzwa Dhokotera, Nandi Joubert, Jan Hattendorf, Christoph Pimmer, Peter Odermatt, Marco Waser, Julia Bohlius (Swiss Tropical and Public Health Institute (Swiss TPH), Associated Institute at the University of Basel)
Generative Artificial Intelligence (GenAI) is transforming academic research and teaching. We conducted a survey to understand students’ and faculty’s current use, needs and perceptions of risks and benefits for the critical use of GenAI in epidemiology.
We included Swiss TPH students (MSc/PhD) and faculty (lecturers/supervisors). Exploratory focus group discussions with students and faculty informed the survey questionnaire covering themes related to GenAI in general and specifically for epidemiological data analysis.
We invited 347 students and faculty members, 131 (38%) completed the survey. 58% of participants were women, 67% European. The most frequently used GenAI tools were ChatGPT-3.5 (73%), DeepL (70%), and ChatGPT-4 (40%). Students and faculty used ChatGPT-3.5/ChatGPT-4 mainly for information retrieval/learning, text improving and coding. Only 15% of students reported to not use any GenAI tools for assignments. The perception of potential risks and benefits were similar in students and faculty; the top three risks were hallucinations (i.e. nonsensical or inaccurate output), bias reinforcement and mainstream idea dominance; the top three benefits automating repetitive tasks, improving language proficiency/translation quality and assistance/support.
GenAI is widely used by students and faculty in the field of epidemiology at Swiss TPH. Students and faculty reported similar types of use and perceived risks and benefits, and showed high levels of criticality towards GenAI. Additional funding resources are urgently needed to address the paramount needs of students and faculty for training, guidance and policy. Prospective cohort studies including students and faculty should scrutinize changing competences in academic research and teaching in the era of GenAI.
Autor*innen: Gustav Arnold (University of Teacher Education Lucerne, Switzerland), Edina Krompák (University of Teacher Education Lucerne, Switzerland), Stephan Meyer (University of Teacher Education Lucerne, Switzerland), Haley de Korne (University of Teacher Education Lucerne, Switzerland), Kristin Vold Lexander (University of Teacher Education Lucerne, Switzerland), Magdalena Madany-Saá (University of Teacher Education Lucerne, Switzerland)
This poster introduces a Collaborative Online International Learning (COIL) course developed within the SOELE project (Socially and Emotionally Responsible Language Education). The SOELE project leverages COIL as a transformative educational strategy to enhance the competences of future teachers to engage in socially and emotionally responsive language education.
By (i) addressing linguistic justice, (ii) promoting emotional well-being, and (iii) fostering
decolonizing language education through critical thinking and reflective pedagogical practices, this COIL course cultivates inclusive and equitable learning environments.
Central to the project is international co-teaching in teacher education, supported by the Swiss- Norwegian partnership involving the University of Teacher Education Lucerne, the University of Oslo, and the Inland Norway University. Drawing on diverse perspectives, knowledge transfer, and complementary research foci, this collaboration creates transformative potential, advancing linguistic justice as well as fostering overall well-being.
Autor*innen: Valérie Andres (Fachstelle Open Access FHNW), Nadja Böller (Koordination Informations-kompetenz Bibliothek FHNW), Brigitte Schubnell (Leiterin Bibliothek FHNW)
Wir gehen der Frage nach, wie sich die Rolle von Plagiatssoftware im Zeitalter der generativen KI ver-ändert. In unserem Poster beleuchten wir das Spannungsfeld zwischen rechtlichen, technischen und qualitativen Aspekten des wissenschaftlichen Schreibens und diskutieren, warum KI-Detektion keine realistische Lösung darstellt. Stattdessen plädieren wir für einen kompetenzorientierten Ansatz: Anstatt auf technische Tools zu setzen, sollte der Fokus stärker auf der Förderung von Kompetenzen wie kriti-schem Denken, Quellenkritik und der Fähigkeit zum reflektierten Umgang mit KI-Tools liegen. Durch die frühzeitige Auseinandersetzung mit KI lernen Studierende, digitale Technologien verantwortungsvoll zu nutzen. Ziel des Posters ist es, die technischen, inhaltlichen und rechtlichen Herausforderungen von KI-Anwendungen und Plagiatskontrolle aufzuzeigen. Wir beleuchten die Rolle der Hochschulbibliotheken bei der Vermittlung von Kompetenzen und als Partner in transdisziplinären Ansätzen der Hochschulbil-dung.
Autor*innen: Roger Flühler (ZHAW) & Esther Stutz (FHNW)
Offene Bildungsangebote erleichtern den Zugang zur Bildung und fördern eine transdisziplinäre Wissensproduktion, die den Bildungsraum Schweiz nachhaltig stärkt. Switch OER ist die neue Austauschplattform der Schweizer Hochschulen, die einen Beitrag zur digitalen Transformation und Qualitätssicherung leistet, indem sie Open Educational Resources (OER) bereitstellt und so den Dialog zwischen Disziplinen und Institutionen unterstützt.
Autor*innen: Edina Krompák (University of Teacher Education Lucerne, Switzerland),
Irene Althaus (University of Teacher Education Lucerne, Switzerland), Cláudia Hilsdorf
Rocha (State University of Campinas, Brazil), Simone Tiemi Hashiguti (State
University of Campinas, Brazil), Ruberval Franco Maciel (State University of Mato
Grosso do Sul, Brazil)
This poster will present and discuss an innovative transnational project that aims to
contribute to sustainable language education, integrating concepts of linguistic (Stroud,
2018; Lim et al, 2018) and global citizenship (Akkari, A. & Maleq, K., 2020; UNESCO,
2014)). This project is a collaboration between the University of Teacher Education of
Lucerne, Switzerland, and two Brazilian universities: the State University of Campinas
and the State University of Mato Grosso do Sul. It focuses on developing multimodal
teaching materials that combine linguistic and global citizenship, promoting sustainable
linguistic practices and addressing social justice issues (Freire, 1970/2018; Riley et al,
2024). By employing the COIL approach (Rubin, 2022), the project facilitates
intercultural exchange and collaboration between Swiss and Brazilian students and
educators. This experience not only enriches participants› educational experiences, but
also promotes awareness of sustainable language education (UNESCO MGIEP, 2017).
Main results include the creation of multimodal teaching materials, the development of
an ethnographic film and the publication of joint research articles. The project also
contributes to the internationalization of teacher training by incorporating perspectives
from the Global North and Souths, ultimately improving the quality of language teaching
and learning across borders and continents.
Autor*in: Mirjam Weder (Departement Sprach- und Literaturwissenschaften der Universität Basel)
Seit der Veröffentlichung von ChatGPT (2022) hat sich der Diskurs zum Schreiben mit KI im hochschulischen Kontext intensiviert. Dennoch mangelt es noch an evidenzbasierten Erkenntnissen zum KI-gestützten Schreiben von Studierenden. Ziel der Studie ist es, Schreibprozesse und Schreibstrategien von Studierenden beim Schreiben mit textgenerativer KI zu untersuchen. Dazu müssen klassische Schreibprozess-Modelle (vgl. Hayes 2012) neu als «Koaktivität von Mensch und Maschine» (Steinhoff 2023) konzeptualisiert werden.Inhaltlich wird danach gefragt, welche Strategien Studierende beim Schreiben mit KI einsetzen, wie sie Teilprozesse anpassen, etwa das Wechselspiel zwischen Prompt – Output – Überarbeitung des Outputs, und ob unterschiedliche Strategien unterschiedliche Effekte auf die Textprodukte haben.
Methodisch geht es darum, ein Verfahren zu entwickeln, das diese Prozesse der Koaktivität erfassen, beschreiben und in einen Zusammenhang mit den Textprodukten bringen kann. Dazu wurden verschiedene Erhebungsinstrumente eingesetzt: Screen-Capture (OBS Studio) und Keystroke-Logging (Leijten& Waes 2013), um Textproduktionsprozesse und das Zusammenspiel zwischen menschlichem Input und KI-Output zu erfassen; Stimulated Recall (Gass 2000), um metakognitive Prozesse zu erfassen, sowie eine Kurzbefragung zu Schreibstrategien mit KI sowie Selbstwirksamkeitsüberzeugungen.
Die Posterpräsentation berichtet von der Pilotstudie, in der die Methoden getestet und die Prozesse und Strategien explorativ ausgewertet wurden, mit dem Ziel ein Kategoriensystem für die Beschreibung und Analyse der Prozesse und Strategien zu erstellen. Datengrundlage bilden vier Schreibsitzungen von Studierenden der Germanistik mit unterschiedlichen Voraussetzungen im wissenschaftlichen Schreiben und in der Anwendung von KI. Der Fokus der Posterpräsentation liegt auf den Prompting-Strategien und den Revisionshandlungen, die beide – so die erste Erkenntnis dieser Studie – mit den Strategie- und Kompetenzprofilen der Versuchspersonen zusammenhängen.