AI evidence extraction

Understanding the Effects of GenAI as No-Code Alternative for Teaching Machine Learning Workflows

Martin Strobel · Proceedings of the AAAI Conference on Artificial Intelligence · 2026

AI-generated evidence extraction, verified across multiple analytical personas. Not a substitute for the peer-reviewed original.

8/10

Relevance

0/4

Quality (LMQS)

Evidence

Citations

This is an AI analysis. Read the peer-reviewed original at the publisher: https://doi.org/10.1609/aaai.v40i47.41516

Methodology & findings

Study design

Mixed methods case study combining survey responses, informal interviews, and classroom observations in a polytechnic-wide elective course

Primary method

Descriptive analysis of survey responses, qualitative analysis of informal interviews, and thematic analysis of classroom observations. No inferential statistical testing reported.

Main result

The study found that "both tools supported conceptual learning, but students' experiences diverged: KNIME provided predictability and structured guidance, while GenAI offered speed and flexibility yet posed setup challenges and required coding familiarity." Students valued having a choice between the two platforms, though this complicated teaching logistics.

Reports effect sizes.

Research paradigm

Mixed methods (qualitative and quantitative descriptive)

Author conclusions

The authors conclude that "GenAI can complement—but not yet replace—traditional no-code platforms, and that the design of introductory activities is critical for adoption." They emphasize the importance of sharing "lessons learned for educators considering GenAI as an alternative in workflow-based ML education."

Risk of bias

Selection bias: students self-selected which tool to use rather than random assignment; Confounding: prior coding experience and familiarity with Python varied among students; Observer bias: classroom observations by educators may influence student behavior; Attrition: no mention of dropout rates or completion percentages; Self-reporting bias in survey responses and interviews; Selection bias: Students self-selected which tool to use (KNIME or GenAI); No control group for comparison; Lack of randomization; Potential instructor bias in classroom observations; Small-scale, single-institution study may limit generalizability; Single institution study may limit generalizability; Potential sampling bias in survey and interview respondents

Limitations

The authors note that "Our experience suggests that GenAI can complement—but not yet replace—traditional no-code platforms" and emphasize that "the design of introductory activities is critical for adoption," suggesting limitations in GenAI's current readiness as a standalone replacement and the context-dependent nature of tool effectiveness.

Open questions raised

Further investigation into optimal design of introductory activities for GenAI adoption in ML education
Exploration of how GenAI and traditional no-code platforms can be integrated effectively
Study of long-term retention and deeper learning outcomes with GenAI tools
Investigation of how to address setup challenges and coding familiarity barriers when using GenAI
Need for further research on GenAI's role in ML education beyond current no-code platforms
Investigation of optimal introductory activity design for GenAI adoption

Data: not_statedCode: not_statedExtracted from: pdfAgreement 66%

Explore related topics

AI Literacy Training Doctoral Training Mixed Methods