mobile icon
Project

CO-CREATION OF SHORT NARRATIVE TEXTS WITH AI – EFFECTS ON TEXT RECEPTION AND PRODUCTION

WorkgroupMultimodal Interaction Lab
Duration10/2020 - 10/2023
FundingSondertatbestand Data Science
Project description

This research project, which is part of the research network "Human-Agent-Interaction", investigates how laypersons (and literary experts) perceive and evaluate an artificial intelligence (AI) based writing tool that produces creative output such as short narrative texts and poems. How credible, creative, and easy to read are the outputs of such a GPT-3-based writing tool?


In particular, two aspects will be investigated: The generation of texts using an AI-writing-tool and the reception of such texts. On the one hand, it will be examined how AI-texts are evaluated by laypersons (and literary experts) in comparison to exclusively human-written texts with regard to their aesthetics and incorporated creativity. How do they perceive the outputs of such a tool? Do they show different reading comprehension and reading behavior when reading creative narrative AI-texts? Differences in reading behavior between AI-texts and human-written texts are investigated with the use of eye-tracking. Do recipients perceive that the content is partially created by an AI? And how confident are they in distinguishing AI-texts from human-written texts?


In a further step, it will be investigated how an AI-writing-tool can support the writing process. For example, does it foster creativity? Are positive effects of using an AI-writing-tool lasting, meaning do they persist beyond the use of the tool? Furthermore, it will be investigated whether such an AI-writing-tool is accepted and how the topic of AI-co-authorship could be dealt with in the future.


These aspects will be investigated in experimental studies in interdisciplinary collaboration with the German Literature Archive Marbach.  The aim is to gain insights into human-agent processes, especially in creative domains e.g. in the production of literature. These findings will help to show the opportunities and limitations of AI-based tools and to better design human-computer interaction in the context of AI.

Cooperations
  • Prof. Dr. Sandra Richter, Prof. Dr. Heike Gfreies (German Literature Archive Marbach)
  • Prof. Dr. Enkelejda Kasneci (University Tübingen, Department of Computer Science)
  • Prof. Dr. Markus Huff, Dr. Nadia Said (University Tübingen, Applied Cognitive Psychology)
Publications

Gunser, V. E., Gottschling, S., Brucker, B., Richter, S., & Gerjets, P. (2021). Can users distinguish narrative texts written by an artificial intelligence writing tool from purely human text? In C. Stephanidis, M. Antona, & S. Ntoa (Eds.), HCI International 2021 - Posters. HCII 2021. Communications in Computer and Information Science: Vol. 1419 (pp. 520-527). Springer. https://dx.doi.org/10.1007/978-3-030-78635-9_67