Cute or Not: Validating Lorenz's Kindchenschema With Text-To-Image Generation
dc.contributor.author | Nischwitz, Lena Marcella | |
dc.contributor.author | Haupt, Josef | |
dc.contributor.author | Weigelt, Aurora-Zoe | |
dc.contributor.author | Chuang, Lewis L | |
dc.date.accessioned | 2024-10-08T15:13:01Z | |
dc.date.available | 2024-10-08T15:13:01Z | |
dc.date.issued | 2024 | |
dc.description.abstract | This study validates Lorenz’s Kindchenschema features by using a commercial text-to-image generator (Midjourney v.5.2) to generate stimulus material that depicted either animals or objects with or without the text prompts for “cute” and “anthropomorphic”. In an online study human participants evaluated the presence of Kindchenschema features in said images. Our results show their increased presence and perception in AI-generated images created with the prompt “cute” compared to those explicitly excluding it, as well as the systematically prioritisation of features such as “large, deep-set eye” and “rounded body shapes” in cute images. Whereas “round, protruding cheeks”, traditionally considered essential by Lorenz, ranks lower. Furthermore, the anthropomorphisation of subjects reduces the likelihood of Kindchenschema features, highlighting constraints in AI-generated stimuli. This research demonstrates the potential of AI-tools to objectively quantify and test theoretical concepts as represented in popular culture. | en |
dc.identifier.doi | 10.1145/3670653.3677512 | |
dc.identifier.uri | https://dl.gi.de/handle/20.500.12116/44883 | |
dc.language.iso | en | |
dc.pubPlace | New York, NY, USA | |
dc.publisher | Association for Computing Machinery | |
dc.relation.ispartof | Proceedings of Mensch und Computer 2024 | |
dc.subject | Anthropomorphism | |
dc.subject | Baby Schema | |
dc.subject | Cuteness | |
dc.subject | Cuteness Aesthetic | |
dc.subject | Diffusion Model | |
dc.subject | Kindchenschema | |
dc.subject | Large Language Model | |
dc.subject | Midjourney | |
dc.subject | Text-To-Image Generation | |
dc.title | Cute or Not: Validating Lorenz's Kindchenschema With Text-To-Image Generation | en |
dc.type | Text/Conference Paper | |
gi.citation.startPage | 594–598 | |
gi.conference.location | Karlsruhe, Germany |