Methodology of introducing artificial intelligence into practice-oriented disciplines for modeling professional cases and developing decision-making competencies

Abstract

The article examines a methodology for the systematic integration of artificial intelligence technologies into practice-oriented disciplines in higher education for generating variable professional cases and forming decision-making competencies under conditions of uncertainty. It substantiates that traditional static training scenarios reproduce predictable decision trajectories and thereby foster «tunnel thinking», whereas generative models enable the construction of context-dependent situations with incomplete and contradictory data, introduction of sudden constraints (changes in budget, requirements, regulations), and provision of prompt feedback in a dialogic format. The methodology relies on a systems-activity logic of designing sessions, delineation of functions between the instructor and the intelligent agent, and management of cognitive load to prevent uncritical trust in machine responses. During approbation in three universities with a sample of 842 senior undergraduate and graduate students, large language models and industry simulators integrated into LMS were used; digital traces of interactions, expert evaluations of solutions based on criteria of substantiation, originality, ethics, and economic efficiency, as well as psychometric and survey indicators with statistical testing of differences were analyzed. It is shown that working in an adaptive environment reduces the time for initial analysis of conditions while simultaneously increasing the depth of elaboration, increases the number of considered alternatives (on average more than twice), improves risk forecast accuracy and substantiation index, and levels results within the group due to the scaffolding effect. A significant increase in resilience to informational noise, ability to revise strategy, and detection of logical errors in generated content has been recorded, indicating enhancement of critical verification rather than its degradation. Survey indices reflect high perceived usefulness of the technology amid heterogeneous «ease of use» for underprepared students and increased anxiety, while engagement remains elevated. External employer evaluation after practice demonstrates reduction of errors in non-standard situations and acceleration of social integration, confirming transfer of formed strategies to real activity; conditions for scaling, necessity of personalizing complexity, and ethical-methodological limitations of AI application in education are discussed.