TEST DESIGN AND STRUCTURE FOR ASSESSING HUMANS AND LLMs’ LINGUISTIC CREATIVITY

Abstract

In this paper, we describe the design and the methodological framework of a Language
Creativity Test (LCT), whose aim is to assess linguistic creativity across humans and Large
Language Models (LLMs). Based on recent research on human and computational creativity,
this LCT evaluates the capacity to generate novel, meaningful, and contextually appropriate
language. For human participants, the test measures imaginative linguistic performance,
including metaphor generation, neologism creation, and flexible word use, which are
considered key indicators of divergent thinking and linguistic fluency. For LLMs, the same
tasks serve to determine their creative language generation and test the extent to which they
go beyond learned patterns to produce original expressions. Thus, the LCT described here
enables direct comparison of creative language use between biological and artificial systems,
offering insights into how creativity manifests across different types of intelligence.