ScholarWorks@성균관대학교

초록

Rapid advancements in generative artificial intelligence are driving innovation across diverse industries, with large language models (LLMs) expanding their application scope as they acquire near-human language-processing capabilities. However, existing research has primarily focused on qualitative evaluations and performance comparisons of LLM models, limiting our objective understanding of how consumers evaluate LLM services and assign them economic value. To address this gap, this study quantitatively evaluates the user experiences of LLM services and analyzes their economic value. Using a discrete choice experiment, we systematically examine consumer preferences for key attributes, including price, response accuracy, response speed, maximum response length, content type, and lexical comprehension level. The results reveal that response accuracy is the most important factor, followed by price, language comprehension, and content type. Particularly, users demonstrate a significantly higher willingness to pay for image-generation functions than for text-generation ones. Simulation outcomes further indicate that depending on pricing and functionality strategies, on-device models have the distinct potential to compete with cloud-based models. By classifying LLM service attributes into industry-driven and user-centered factors, this study provides actionable insights for firms seeking to design user-centric and sustainable business models.

키워드