20.500.12556/RUL-186501 Analiza uporabe pozivov pri delu z generativno umetno inteligenco Analysis of prompting practices when working with generative artificial intelligence Generativna umetna inteligenca je postala vsakdanje orodje pri razvoju programske opreme, osrednji način dela z njo pa je pozivanje. Klasične mere produktivnosti zajamejo izid, ne pa bremena poti do njega. V tem delu to breme poimenujemo breme pozivanja (angl. prompting tax), tj. čas, trud in delo pregledovanja pri pridobivanju delujoče kode iz pomočnika z umetno inteligenco. Razvili smo eksperimentalno platformo z urejevalnikom kode, klepetalnim pomočnikom in samodejnim preverjanjem ter izvedli študijo uporabnikov na treh programerskih nalogah. Analiza 56 nalog, ki jih je rešilo 27 udeležencev, kaže, da ima breme merljiv vedenjski podpis, ki tesno sledi zaznanemu trudu, da ne prinaša večje pravilnosti rešitve, da se zgosti pri najtežji nalogi ter da se z vajo ne zmanjša. Samozavest in izkušnje pa ne napovedujejo, kdo mu je bolj izpostavljen. Odkrivamo tudi skrito breme pregledovanja v obliki zamaknjenega preverjanja predlaganih sprememb in analiziramo strategije pozivanja udeležencev. Generative artificial intelligence has become an everyday tool in software development, with prompting as the primary mode of interaction. Conventional productivity measures capture the outcome but not the cost of reaching it. We call this cost the prompting tax, i.e. the time, effort, and review cost of obtaining working code from an AI assistant. We built an experimental platform combining a code editor, a chat assistant, and automated testing, and ran a user study on three programming tasks. An analysis of 56 tasks solved by 27 participants shows that the tax has a measurable behavioural signature closely tracking perceived effort, that it does not buy higher correctness, that it concentrates on the hardest task, and that it does not ease with practice, while neither confidence nor experience predicts who pays it. We also identify a hidden review tax in the form of deferred verification of proposed code edits, and analyse participants' prompting strategies. Interakcija človek-računalnik generativna umetna inteligenca oblikovanje pozivov študija uporabnikov produktivnost uporabnikov veliki jezikovni modeli human-computer interaction generative artificial intelligence prompting user study user productivity large language models true false false Slovenski jezik Angleški jezik Magistrsko delo/naloga 2026-09-02 13:15:59 2026-09-02 13:16:12 2026-09-04 11:37:19 0000-00-00 00:00:00 2026 0 0 0000-00-00 NiDoloceno NiDoloceno NiDoloceno 0000-00-00 0000-00-00 0000-00-00 38565 290045443 Pahor_Erik_-_Analiza_uporabe_pozivov_pri_delu_z_generativno_umetno_inteligenco.pdf Pahor_Erik_-_Analiza_uporabe_pozivov_pri_delu_z_generativno_umetno_inteligenco.pdf 1 8AB324B0C61A0D8BB4F24F7EE23652AA 950143eae2964e38bba9d3de3d4af33ff43c6779f09fb75a66b548afc95a359b 47b51f6c-a6bf-11f1-9b0d-0050569b8976 https://repozitorij.uni-lj.si/Dokument.php?lang=slv&id=245769 Fakulteta za računalništvo in informatiko 0 0 0