2026-05-19T02:05:34Zhttps://keep.lib.asu.edu/oai/request

oai:keep.lib.asu.edu:node-2012832025-05-23T22:20:13Zoai_pmh:alloai_pmh:repo_items

201283 https://hdl.handle.net/2286/R.2.N.201283 http://rightsstatements.org/vocab/InC/1.0/ http://creativecommons.org/licenses/by-nc-sa/4.0 2025-05 55 pages Rice Nulty, Seth Chavez Echeagaray, Maria Elena Zhu, Qiyun Barrett, The Honors College Department of Psychology Computer Science and Engineering Program While introductory machine learning education emphasizes both theoretical foundations and programming syntax, the ability to confidently navigate data workflows and make informed decisions is typically developed through repeated, project-based experience. Interest in machine learning education is expanding simultaneously with the widespread availability of generative artificial intelligence (GAI) tools, capable of automating data analysis, decision-making, and code generation. While convenient, GAI tools may risk undermining students’ development of essential reasoning skills when used without a solid understanding of the underlying concepts. This issue may be particularly pressing for machine learning students attempting to conduct a machine learning workflow, already balancing the dual demands of learning programming and machine learning theory. DataPylot is an interactive application designed to support early learners in machine learning by fostering strategic intuition around effectively applying machine learning techniques to data. The tool provides a structured interface that enables users to upload a dataset of their choice and apply common supervised machine learning tasks step-by-step. At each step, users make explicit choices through a graphical interface, and the application deterministically maps these selections into formatted Python code ready for execution. DataPylot supports code generation for many typical tasks within a machine learning workflow, including importing a raw dataset, applying exploration and preprocessing techniques, and training and evaluating machine learning models. This approach places the reasoning and decision-making process in the hands of the learner while reducing programming barriers and minimizing common errors. To evaluate DataPylot’s educational value, a user study was conducted with Arizona State University students who had prior experience with machine learning. Participants successfully completed a guided machine learning project using only code generated within DataPylot to explore, preprocess, and model a provided dataset. Afterwards, participants completed a questionnaire assessing perceived challenges in learning machine learning, the ease of use and value of using the tool, and its comparative utility relative to GAI tools. Results indicated that participants found DataPylot useful for applying machine learning and supporting independent reasoning, particularly among lower-experience learners. These findings provide support for the potential of structured, interactive, and deterministic tools like DataPylot to foster conceptual understanding and strategic intuition surrounding applying machine learning to data through guided hands-on engagement. Machine learning Generative AI Education DataPylot Application: Empowering Students to Apply Supervised Machine Learning With Automated Programming