04045ntm a22005177i 4500 000727707 CZ-PrVSE 20260201122909.0 m d cr n|||||||||| 260201s2025 xr fsbm 000 0 eng d NEZPRACOVANÝ IMPORT ABA006 cze ABA006 ABA006 rda Dervishaj, Trevi ISIS:172664 dis APLIKACE JAZYKOVÝCH MODELŮ (AI) NA AKADEMICKÉ PRÁCE PRO ZLEPŠENÍ VÝKONU eng Application of language models (AI) on Academic works for enhanced performance / Trevi Dervishaj 2025 ?? stran : digital, PDF soubor Vedoucí práce: Antonín Pavlíček Diplomová práce (Ing.)—Vysoká škola ekonomická v Praze. Fakulta informatiky a statistiky, 2026 Obsahuje bibliografii Textový (vysokoškolská kvalifikační práce) Rok obhajoby 2026 The academic community is facing a paradigm shift on the one hand brought about by Large Language Models (LLMs) such as ChatGPT which are generative engines with the ability to generate complex, coherent, and contextually relevant text. Such innovation brings a two-fold dilemma: the potential of unmatched efficiency (e.g., having overcome writer block, drafted faster) is offset by a potentially serious integrity risk due to the fundamental factual and reliability issue of the LLMs, which is the Hallucination Problem. Institutions have mostly been reactive with ineffective strategies, which has led to swamping of Verification Fatigue, the administrative cost of manually verifying every output of AI. The lack of a validated, ethical procedural framework is another critical issue that is dealt with in this dissertation. It uses a Sequential Exploratory Mixed-Methods Design in the initial step which is to determine the starting patterns of LLM adoption, and it will attest that the latter is a student-driven high-velocity content writing support. Then, it proposes and empirically proves Structured AI Integration Workflow: a ten-step, skill-based procedural intervention, which requires human responsibility, strategic encouragement and strict verification at important points of academic development. Findings indicate that the workflow was highly rated in terms of mean utility (4.03), with the greatest effects on Outlining and Structuring (4.12) which confirm AI as an organizational catalyst. Nevertheless, qualitative responses also indicated that Verification Fatigue is the main barrier. The results propose policy shift exceeding the Permissive/ Integrated Level to Transformative/ Curricular Level, where the pedagogy has been reformulated to measure Prompt Mastery and Critical Evaluation. The research finds that the future of academic integrity lies in the definition of human role without being a content maker, but the essential critical verifier of all AI-assisted research. Způsob přístupu: Internet information systems management [obor dipl. práce] diplomové práce fd132022 czenas master's theses eczenas Large Language Models (LLMs) ChatGPT Generative Engines Academic Integrity Pavlíček, Antonín, 1977- vse2008429995 ths Sudzina, František ISIS:137922 opn Vysoká škola ekonomická v Praze. Fakulta informatiky a statistiky kn20010709399 dgg https://insis.vse.cz/zp/86318/podrobnosti VŠKP v InSIS https://insis.vse.cz/zp/86318 Hlavní práce https://insis.vse.cz/zp/86318/posudek/vedouci Hodnocení vedoucího https://insis.vse.cz/zp/86318/posudek/oponent/80575 Oponentura https://insis.vse.cz/zp/86318/podrobnosti dc:identifier NEPOSILAT VSKP vse86318 260116 86318