CLASS: A Design Framework for Building Intelligent Tutoring Systems Based on Learning Science principles
Open accessAbstract
We present a design framework called Conversational Learning with Analytical Step-by-Step Strategies (CLASS) for building advanced Intelligent Tutoring Systems (ITS) powered by high-performance Large Language Models (LLMs).The CLASS framework empowers ITS with two key capabilities.First, through a carefully curated scaffolding dataset, CLASS equips ITS with essential problemsolving strategies, enabling it to provide tutorlike, step-by-step guidance to students.Second, by using a dynamic conversational dataset, CLASS assists ITS in facilitating natural language interactions, fostering engaging studenttutor conversations.The CLASS framework also provides valuable insights into ITS's internal decision-making process which allows seamless integration of user feedback, thus enabling continuous refinement and improvement.We also present a proof-of-concept ITS, referred to as SPOCK, which is trained using the CLASS framework with a focus on introductory college-level biology content.A carefully constructed protocol was developed for SPOCK's preliminary evaluation, examining aspects such as the factual accuracy and relevance of its responses.Experts in the field of biology offered favorable remarks, particularly highlighting SPOCK's capability to break down questions into manageable subproblems and provide encouraging responses to students.
The claims and statistics of this paper haven’t been checked yet.
Run the check — free, no account needed. It reads the open-access text and checks the sources it cites.
Integrity signals
Registry
Cites
Statistics
Claims the paper makes
The claims and statistics of this paper haven’t been checked yet.
Run the check — free, no account needed. It reads the open-access text and checks the sources it cites.
Claims other papers make about this one
Each is checked against this paper’s own text: does it say what the citing paper says it does?
The method of using simulated students, known as interactive tutoring, is used in the literature.
“This method of using simulated students, known as interactive tutoring, is used in the literature (Macina et al., 2023a; Sonkar et al., 2023).”
From Towards the Pedagogical Steering of Large Language Models for Tutoring: A Case Study with Modeling Productive Failure (2025) · cites this paper as Sonkar et al., 2023
The set of labels used in Student State Tracing was discussed with learning scientists, and the prompt structure follows the CLASS framework.
“The set S (as per the notation used in Section 3.1) of labels (also called features) we choose from during the Student State Tracing procedure have been discussed with learning scientists and the prompt structure follows one used by the CLASS framework (Sonkar et al., 2023).”
From Towards the Pedagogical Steering of Large Language Models for Tutoring: A Case Study with Modeling Productive Failure (2025) · cites this paper as Sonkar et al., 2023
Self-check
Percentages, sums and reported differences re-computed from the paper’s own numbers.
The self-check needs the paper’s full text.
Cites
Transformers: State-of-the-Art Natural Language Processing
Wolf, T., Debut, L., Sanh, V., Chaumond, J., Delangue, C., Moi, A., Cistac, P., Rault, T., Louf, R., Funtowicz, M., et al. (2020) · Proceedings of the 2020 Conference on Empirical Methods in Natural Language Processing: System Demonstrations · doi:10.18653/v1/2020.emnlp-demos.6 · Cited by 8,741 · see citing works
Training language models to follow instructions with human feedback
Ouyang, L., Wu, J., Jiang, X., Almeida, D., Wainwright, C. L., Mishkin, P., Zhang, C., Agarwal, S., Slama, K., Ray, A., et al. (2022) · arXiv · doi:10.48550/arxiv.2203.02155 · Cited by 4,208 · see citing works
PaLM: Scaling Language Modeling with Pathways
Chowdhery, A., Narang, S., Devlin, J., Bosma, M., Mishra, G., Roberts, A., Barham, P., Chung, H. W., Sutton, C., Gehrmann, S., et al. (2022) · arXiv · doi:10.48550/arxiv.2204.02311 · Cited by 2,590 · see citing works
Judging LLM-as-a-Judge with MT-Bench and Chatbot Arena
Zheng, L., Chiang, W.-L., Sheng, Y., Zhuang, S., Wu, Z., Zhuang, Y., Lin, Z., Li, Z., Li, D., Xing, E. P., et al. (2023) · arXiv · doi:10.48550/arxiv.2306.05685 · Cited by 1,888 · see citing works
Question Decomposition Improves the Faithfulness of Model-Generated Reasoning
Radhakrishnan, A., Nguyen, K., Chen, A., Chen, C., Denison, C., Hernandez, D., Durmus, E., Hubinger, E., Kernion, J., Lukošiūtė, K., et al. (2023) · arXiv · doi:10.48550/arxiv.2307.11768 · Cited by 8 · see citing works
Flagged first, then by citations
Research works printed but not linked: 25
- Abubakar Abid, Ali Abdalla, Ali Abid, Dawood Khan, Abdulrahman Alfozan, and James Zou. 2019. Gradio: Hassle-Free Sharing and Testing of ML Models in the Wild. arXiv preprint arXiv:1906.02569. · arXiv
- John R Anderson, Albert T Corbett, Kenneth R Koedinger, and Ray Pelletier. 1995. Cognitive tu- tors: Lessons learned. The journal of the learning sciences, 4(2):167-207.
- Maria Bannert. 2003. Effekte metakognitiver Lernhil- fen auf den Wissenserwerb in vernetzten Lernumge- bungen. Zeitschrift für Pädagogische Psychologie, 17(1):13-25.
- Sébastien Bubeck, Varun Chandrasekaran, Ronen El- dan, Johannes Gehrke, Eric Horvitz, Ece Kamar, Peter Lee, Yin Tat Lee, Yuanzhi Li, Scott Lund- berg, et al. 2023. Sparks of Artificial General In- telligence: Early experiments with GPT-4. arXiv preprint arXiv:2303.12712. · arXiv
- Jack Conklin. 2005. A taxonomy for learning, teaching, and assessing: A revision of Bloom's taxonomy of educational objectives complete edition.
- K Anders Ericsson. 2017. Protocol analysis. A compan- ion to cognitive science, pages 425-432.
- Shi Feng, Alejandra J Magana, and Dominic Kao. 2021. A systematic review of literature on the effectiveness of intelligent tutoring systems in STEM. In 2021 IEEE Frontiers in Education Conference (FIE), pages 1-9. IEEE.
- Dan Hendrycks, Collin Burns, Steven Basart, Andy Zou, Mantas Mazeika, Dawn Song, and Jacob Steinhardt. 2020. Measuring Massive Multitask Language Un- derstanding. CoRR, abs/2009.03300. · arXiv
- Ted Kesler, Pablo PL Tinio, and Brian T Nolan. 2016. What's our position? A critical media literacy study of popular culture websites with eighth-grade special education students. Reading & Writing Quarterly, 32(1):1-26.
- Naiming Liu, Shashank Sonkar, Zichao Wang, Simon Woodhead, and Richard G Baraniuk. 2023. Novice Learner and Expert Tutor: Evaluating Math Reason- ing Abilities of Large Language Models with Mis- conceptions. arXiv preprint arXiv:2310.02439. · arXiv
- Naiming Liu, Zichao Wang, Richard Baraniuk, and An- drew Lan. 2022. Open-ended knowledge tracing for computer science education. In Proceedings of the 2022 Conference on Empirical Methods in Natural Language Processing, pages 3849-3862.
- Jakub Macina, Nico Daheim, Sankalan Pal Chowd- hury, Tanmay Sinha, Manu Kapur, Iryna Gurevych, and Mrinmaya Sachan. 2023. MathDial: A Dia- logue Tutoring Dataset with Rich Pedagogical Prop- erties Grounded in Math Reasoning Problems. arXiv preprint arXiv:2305.14536. · arXiv
- Antonija Mitrovic, Stellan Ohlsson, and Devon K Bar- row. 2013. The effect of positive feedback in a constraint-based intelligent tutoring system. Comput- ers & Education, 60(1):264-272.
- Jeff Rasley, Samyam Rajbhandari, Olatunji Ruwase, and Yuxiong He. 2020. Deepspeed: System optimiza- tions enable training deep learning models with over 100 billion parameters. In Proceedings of the 26th ACM SIGKDD International Conference on Knowl- edge Discovery & Data Mining, pages 3505-3506.
- Valerie J Shute, Chen Sun, and Jodi Asbell-Clarke. 2017. Demystifying computational thinking. Educational research review, 22:142-158.
- Shashank Sonkar and Richard G Baraniuk. 2023. De- duction under Perturbed Evidence: Probing Student Simulation Capabilities of Large Language Models. arXiv preprint arXiv:2305.14507. · arXiv
- Shashank Sonkar, MyCo Le, Xinghe Chen, Naiming Liu, Debshila Basu Mallick, and Richard G Bara- niuk. 2023. Code Soliloquies for Accurate Calcu- lations in Large Language Models. arXiv preprint arXiv:2309.12161. · arXiv
- Shashank Sonkar, Andrew E Waters, Andrew S Lan, Phillip J Grimaldi, and Richard G Baraniuk. 2020. qdkt: Question-centric Deep Knowledge Tracing. arXiv preprint arXiv:2005.12442. · arXiv
- C Addison Stone. 1998. The metaphor of scaffolding: Its utility for the field of learning disabilities. Journal of learning disabilities, 31(4):344-364.
- Pramuditha Suraweera and Antonija Mitrovic. 2002. KERMIT: A constraint-based tutor for database mod- eling. In Intelligent Tutoring Systems: 6th Interna- tional Conference, ITS 2002 Biarritz, France and San Sebastian, Spain, June 2-7, 2002 Proceedings 6, pages 377-387. Springer.
- Danielle Thomas, Xinyu Yang, Shivang Gupta, Ade- tunji Adeniran, Elizabeth Mclaughlin, and Kenneth Koedinger. 2023. When the Tutor Becomes the Stu- dent: Design and Evaluation of Efficient Scenario- Based Lessons for Tutors. In LAK23: 13th Inter- national Learning Analytics and Knowledge Confer- ence, LAK2023, page 250-261, New York, NY, USA. Association for Computing Machinery. · DOI
- Hugo Touvron, Thibaut Lavril, Gautier Izacard, Xavier Martinet, Marie-Anne Lachaux, Timothée Lacroix, Baptiste Rozière, Naman Goyal, Eric Hambro, Faisal Azhar, et al. 2023. Llama: Open and effi- cient foundation language models. arXiv preprint arXiv:2302.13971. · arXiv
- Fons JR Van de Vijver and Kwok Leung. 2021. Methods and data analysis for cross-cultural research, volume 116. Cambridge University Press.
- Lev Semenovich Vygotsky and Michael Cole. 1978. Mind in society: Development of higher psychologi- cal processes. Harvard university press.
- Jeannette M Wing. 2006. Computational thinking. Communications of the ACM, 49(3):33-35.
Cited by
Sorted by citations
History, development, and principles of large language models: an introductory survey
Wang, Z., Chu, Z., Doan, T. V., Ni, S., Yang, M., & Zhang, W. (2025) · AI and Ethics · doi:10.1007/s43681-024-00583-7 · Cited by 109 · see citing works
Puech, R., Macina, J., Chatain, J., Sachan, M., & Kapur, M. (2025) · Findings of the Association for Computational Linguistics: ACL 2025 · doi:10.18653/v1/2025.findings-acl.1348 · Cited by 11 · see citing works
LLM-Powered Virtual Patient Agents for Interactive Clinical Skills Training with Automated Feedback
Voigt, H., Sugamiya, Y., Lawonn, K., Zarrieß, S., & Takanishi, A. (2025) · arXiv · doi:10.48550/arxiv.2508.13943
- Authors
- Sonkar, S., Liu, N., Mallick, D., & Baraniuk, R.
- Affiliations
- Rice University (US)
- Venue
- Findings of the Association for Computational Linguistics: EMNLP 2023
- Published
- 2023
- Pages
- 1941-1961
- OpenAlex
- W4389523742
- Cited by
- 35
- References
- 5
- Citations by year
- 2026: 11 · 2025: 17 · 2024: 7
- FWCI
- 6.21
- Type
- conference paper
- Language
- en
- Open access
- hybrid
- doi.org
- Licence
- cc-by
- Locations
- Findings of the Association for Computational Linguistics: EMNLP (Findings of EMNLP) (publishedVersion, cc-by)
- arXiv (Cornell University) (submittedVersion, cc-by)
- arXiv (Cornell University) (cc-by)
- Funders
- National Science Foundation (1842378)
- Office of Naval Research (FA9550, N00014-18-1-2047, N00014)
- Air Force Office of Scientific Research (FA9550-, FA9550-22-1-0060)
- Topics
- Physical Sciences › Computer Science › Artificial Intelligence › Intelligent Tutoring Systems and Adaptive Learning
- Physical Sciences › Computer Science › Information Systems › Artificial Intelligence in Education
- Physical Sciences › Computer Science › Artificial Intelligence › Edcuational Technology Systems
- Keywords
- intelligent tutoring systems, large language models, step-by-step problem solving, conversational datasets, student-tutor interaction, user feedback, biology education, subproblem decomposition, factual accuracy
About these checks
Registry, retraction, Cites and statistics results are computed, not judged, with one exception: a printed reference that no index lists is read once by a model before it counts as a citation error. Claims are extracted by a model and kept only when their quote appears verbatim in the text; each cited source is then checked, with a model’s help, for whether it says what the claim relies on. Those verdicts can be wrong — read the claim’s quote and the cited source before relying on one.
Statistics and arithmetic are recomputed, not judged.
Confidence bands are shown only for fields calibrated on our annotated test set. None is calibrated yet, so no band is shown.
What counts as a claim ›How we measure confidence ›How we check this ›
Are you an author, or do you see a problem with this page? Request a review
