20.500.12556/DKUM-98947 Analiza uporabe velikih jezikovnih modelov in strategij pozivanja pri generativnem oblikovanju uporabniških vmesnikov Analysis of the use of large language models and prompt strategies in generative user interface design Generativno oblikovanje komponent uporabniških vmesnikov z uporabo velikih jezikovnih modelov je naraščajoče področje, ki kljub vse širši praktični uporabi ostaja metodološko slabo empirično raziskano. Magistrska naloga preučuje, kako kakovost pozivov in izbira sistema velikega jezikovnega modela vplivata na zaznano kakovost in prilagodljivost generiranih komponent. V okviru treh eksperimentov so bili primerjani sistemi ChatGPT, Claude Sonnet in Gemini pri treh nivojih kakovosti pozivov in različnih tipih komponent uporabniških vmesnikov. Evalvacija kakovosti je temeljila na okviru šestih metrik, evalvacija prilagodljivosti pa na treh dimenzijah, oblikovanih za namen te raziskave. Rezultati kažejo, da kakovost pozivov in izbira sistema vplivata na zaznano kakovost in prilagodljivost generiranih komponent, pri čemer se njun prispevek razlikuje glede na merjeno dimenzijo. Ugotovitve prispevajo k razvoju evalvacijske metodologije za generativno oblikovanje komponent in imajo neposredne implikacije za uvajanje generativnih orodij v procese oblikovanja programske opreme. Generative design of user interface components using large language models is an emerging field that, despite its increasingly widespread practical adoption, remains methodologically underexplored from an empirical perspective. This master's thesis investigates how prompt quality and the choice of large language model system affect the perceived quality and adaptability of generated components. Across three experiments, the systems ChatGPT, Claude Sonnet, and Gemini were compared at three levels of prompt quality and across different types of user interface components. Quality evaluation was based on a framework of six metrics, while adaptability was evaluated using three dimensions specifically developed for this research. The results indicate that both prompt quality and system choice influence the perceived quality and adaptability of generated components, with their respective contributions varying depending on the dimension being measured. The findings contribute to the development of an evaluation methodology for the generative design of components and have direct implications for the adoption of generative tools in software design processes veliki jezikovni modeli generativno oblikovanje vmesnikov inženiring pozivov evalvacija kakovosti prilagodljivost komponent primerjalna analiza sistemov Nielsenove hevristike large language models generative UI design prompt engineering quality evaluation component adaptability comparative system analysis Nielsen heuristics true true false Slovenski jezik Angleški jezik Magistrsko delo/naloga 2026-07-21 08:19:02 2026-09-24 13:59:44 2026-09-25 03:10:30 0000-00-00 00:00:00 2026 0 Maribor 0 0000-00-00 NiDoloceno NiDoloceno NiDoloceno 0000-00-00 0000-00-00 0000-00-00 MAG_Tomovski_Eftimije_2026.pdf MAG_Tomovski_Eftimije_2026.pdf 1 D953BDA02A7FD6BCF9CA334C16728E90 4edc5664c7ef3ba6d5d4153b9f3ec96cac913db489b457812deb18d8748e5fc8 c0613684-9ae9-11f1-8356-00155d0001be https://dk.um.si/Dokument.php?lang=slv&id=209736 Fakulteta za elektrotehniko, računalništvo in informatiko 0 0 0