A linguo-cognitive analysis of user intentions in interaction with LLMs: a Bloom's taxonomy approach
The article presents the results of automatic classification of 999 natural user requests to the Google Bard AI assistant using the local language model GPT-OSS-20B in accordance with the revised Bloom's taxonomy. The analysis covers two dimensions: categories of cognitive processes and levels of knowledge. The aim of the study is to identify cognitive processes and knowledge types employed in formulating requests to generative AI systems, as well as to evaluate the reliability of the revised Bloom's taxonomy for the automatic classification of such requests. A local open-source language model (GPT-OSS, 20 billion parameters) was used for classification, iteratively processing each prompt with a detailed system prompt containing taxonomy category descriptions, typical action verbs, confidence criteria, and classification examples. The model returned three parameters: cognitive process category, knowledge level, and confidence score (0–100 %). The results reveal the dominance of lower-order cognitive processes: REMEMBER (39.5 %) and CREATE (21.8 %), while ANALYZE accounts for only 2.9 %. In terms of knowledge levels, factual (FACTUAL – 44.4 %) and procedural (PROCEDURAL – 29.9 %) requests prevail. Metacognitive requests represent only 1.7% of the total, indicating a virtually undeveloped niche of human-AI interaction in the domain of reflection and self-knowledge. The study demonstrates the successful applicability of the revised Bloom's taxonomy for the systematic classification of LLM prompts, revealing clear patterns: contemporary users employ large language models as a hybrid tool – simultaneously as a reference source for fact retrieval and as a procedural solution generator – while virtually never using them for metacognitive analysis. Furthermore, the obtained distributions of cognitive processes and knowledge levels are interpreted as an empirical model of typical user tasks when interacting with language models, which enables the formulation of requirements for the architecture and functionality of terminology management systems in the context of generative AI.
While nobody left any comments to this publication.
You can be first.
Дятлова, Е. Н. (2022), «Теоретическое обоснование и разработка уровней сформированности социально-гуманитарных знаний обучающихся общеобразовательной школы», Человек. Наука. Социум, 4-2, 8-32. EDN: RQELLZ
Dyatlova, E. N. (2022), “Theoretical justification and development of levels of attainment in social sciences and humanities knowledge among pupils in general education schools”, Human. Science. Socium, 4-2, 8-32 (in Russ.). EDN: RQELLZ
Коваленко, Е. С. и Кузуб, Н. М. (2020), «Использование таксономии Блума для повышения качества профессиональной подготовки студентов педагогического вуза», Казанский педагогический журнал, 1, 90-96. EDN: GYCAWX
Kovalenko, E. S. and Kuzub, N. M. (2020), “Using Bloom's Taxonomy to Improve the Quality of Professional Training of Pedagogical University Students”, Kazan Pedagogical Journal, 1, 90-96 (in Russ.). EDN: GYCAWX
Кораванец, Н. В. (2025), Потенциал аспектов таксономии Блума–Андерсона как ресурс преобразования образовательной деятельности каждого учителя и Гимназии в целом: методическое пособие, Красноярская Мариинская женская гимназия-интернат, Красноярск.
Koravanets, N. V. (2025), The Potential of Bloom-Anderson Taxonomy Aspects as a Resource for Transforming the Educational Activities of Each Teacher and the Grammar School as a Whole: A Methodological Manual, Krasnoyarsk Mariinsky Women's Boarding School, Krasnoyarsk
Anderson, L. W. and Krathwohl, D. R. (eds.) (2001), A Taxonomy for Learning, Teaching, and Assessing: A Revision of Bloom's Taxonomy of Educational Objectives. Longman, New York, USA.
Bootu, L., Mahato, A., Sriwastav, S. K., Miriyala, R. and Bali, D. (2025), “Automated analysis of exam questions based on Bloom's Taxonomy”, Proc. 2nd Int. Conf. on Intelligent Systems for Cybersecurity (ISCS), Gurugram, India, 1–5. DOI: 10.1109/ISCS69371.2025.11385935
Rivera-Reyes, P., Pérez, L. C. and Delahunty, T. (2018), “Assessing the demand of problems in an undergraduate electrical engineering course”, Proc. IEEE Frontiers in Education Conference (FIE), San Jose, CA, USA, 1–8. DOI: 10.1109/FIE.2018.8659058
Trippas, J. (2026), BIDD-1k: Bard Intelligence and Dialogue Dataset [Online], GitHub, available at: https://github.com/JTrippas/BIDD-1k (Accessed 17.01.2026).
Zhang, Y., Lin, Y., Khan, A. and Wan, H. (2025), “Large Language Model Prompt Datasets: An In-depth Analysis and Insights”, arXiv:2510.09316 [cs.LG] [Online], available at: https://arxiv.org/abs/2510.09316 (Accessed 17.01.2026).