Cognitive Science Student Makes the Cut for Main Conference
Over the weekend, Halfdan Nordahl Fundal and his research colleagues received the news that their paper has been accepted as a Main Conference Paper at the Empirical Methods in Natural Language Processing (EMNLP), which will be held in Budapest in late October. The paper examines turn-based collaborative storytelling across three matched conditions: Human-Human, Human-AI, AI-AI.
Halfdan Nordahl Fundal recently completed his bachelor’s degree in Cognitive Science at School of Communication and Culture at Aarhus University, where he is now pursuing a master’s degree. Alongside his studies, he works as a Student Assistant at TEXT: Center for Contemporary Cultures of Text. At TEXT he has had the opportunity to investigate how creative co-creation processes unfold in different configurations involving humans and AI.
In July 2026, Halfdan Nordahl Fundal and Yuri Bizzoni, PhD and Senior Researcher at the Center for Humanities Computing at Aarhus University, presented their research project in San Diego, USA, at a workshop held as part of the Annual Meeting of the Association for Computational Linguistics (ACL). Their ACL paper focused exclusively on interactions between humans and language models, and the results reveal a clear asymmetry in the roles played by the two.
- Our results suggest that humans write more negatively and also appear to be more persistent in their emotional starting point for the story. Language models, on the other hand, adapt to humant input – for example, when a person changes the tone from negative to positive, says Halfdan Nordahl Fundal.
Humans are more creative than AI
Their research also show that humans tend to drive the creative process, introducing new ideas and shaping the direction of the text, while language models primarily build on these contributions by elaborating, creating coherence and adapting to the human input.
- Humans introduce more surprising – and perhaps also more creative – elements into the text. These surprising elements also have a greater influence on what happens next than similarly surprising elements introduced by the language model. It is almost as if humans have a special privilege: the surprising elements they contribute are given a special status that allows them to change the story in a way that comparable contributions from the language model do not, he says.
In their Main Conference Paper for the EMNLP, Halfdan Nordahl Fundal and his colleagues expand the analysis by comparing human–AI interactions with both human–human and AI-AI interactions. This provides a more complete picture of how the observed differences play out across collaborations between different types of actors.
- From a broader perspective, the project is about fundamental dynamics in collaboration between humans and AI. We hope that the methods can also be generalised to other types of linguistic exchange, such as conversations and political debate, as well as to broader research into human creativity and text, says Halfdan Nordahl Fundal.
The idea dates back to his second semester
The roots of the project can be traced all the way back to a project in the second semester of his bachelor’s degree, in the course Applied Cognitive Science. Together with his fellow students, he developed a close collaboration with Senior Researcher Karsten Olsen at the Interacting Minds Centre (IMC) at Aarhus University.
The examination project has since evolved into a science communication platform that has toured venues across Denmark as part of Videnskaber. It also led to the spinout company Tinkestudio and, not least, provided the initial foundation for the research project at TEXT from which the two papers have emerged.
Facts
Halfdan Nordahl Fundal is the lead author of the newly accepted paper. Yuri Bizzoni, PhD and Senior Researcher, is second author. Two former interns and student assistants, Charlotte Bilde and Ida Bække Johannesen, contributed to the data collection and are third and fourth authors, respectively. Rebekah Baglini, Co-Director of TEXT, is fifth author and supervisor.
Learn more
Studerende og forsker fra Arts formidler kunstig intelligens på landsækkende tourné
TEXT: Center for Contemporary Cultures of Text
Interacting Minds Centre
New Spinout company launched: Tinke Studio
Contact
Halfdan Nordahl Fundal
Stud. MSc Cognitive Science
TEXT: Center for Contemporary Cultures of Text
E-mail: halfi@cc.au.dk
Phone: 51324169