A Case Study of Integrating AI Literacy Education in a Biology Class

Integrating AI literacy into an advanced biology course — Zha et al. (2025). A concurrent-triangulation case study in which 37 high-school students learned machine learning, artificial neural network, and convolutional neural network concepts embedded in four biology lessons. Students' overall AI knowledge improved significantly; overall biology knowledge rose slightly but not significantly. Biology knowledge significantly predicted AI learning overall and in the two lessons where AI concepts were concretized in familiar biology.

Key Findings

This study addresses a persistent gap in K-12 AI education: whereas most US AI literacy instruction has been delivered through extracurricular activities, this work integrates AI learning directly into a standard academic course — a high-school honors biology class. The authors argue that integrating AI into a disciplinary context both broadens access and tests whether contextual subject knowledge supports AI learning, an under-examined question in the field.

Design. The researchers used a concurrent triangulation (mixed-method) design. Quantitatively, they measured conceptual understanding via pre/post-tests and fill-in-blank worksheet questions; qualitatively, they analyzed students' interdisciplinary reasoning in open-ended worksheet answers using an adapted version of Shen et al.'s (2015) interdisciplinary reasoning and communication framework, with strong inter-rater agreement (κ=0.88).

Four biology-embedded AI lessons. Students learned machine learning, artificial neural networks, and convolutional neural networks across lessons on cell classification with Teachable Machine, comparing human neurons with ANN structure, CNN in plant science, and AI in colorectal cancer detection. The Instructional Design deliberately selected biologically inspired AI topics so students could see the connection between the two domains.

Outcomes. Paired t-tests showed significant growth in overall AI knowledge (M 2.81 → 4.12, p<0.001), while overall biology knowledge increased only slightly and non-significantly (p=0.21). A regression (R²=0.64, p<0.01) found that both biology knowledge and prior AI knowledge significantly predicted post-test AI scores.

The role of context. Per-lesson regressions revealed that prior biology knowledge significantly predicted new AI learning in Lesson 1 (R²=0.24) and Lesson 2 (R²=0.52), but not in Lessons 3 and 4, where AI concepts were presented through abstract academic articles. The authors interpret this through semantic wave theory: hands-on or familiar contexts yielded high semantic gravity and concretized AI, whereas article-based lessons had lower gravity and higher semantic density, hindering learning.

Interdisciplinary reasoning. Qualitative analysis found that students who transferred source-domain knowledge generally produced better explanations in the target domain, while non-transfer was associated with vague or inaccurate explanations. The authors suggest a potential causal relationship between transfer and explanation that warrants further quantitative testing.

Implications

The findings support integrating AI literacy into STEM Education curricula rather than relegating it to extracurricular settings, consistent with calls to embed AI learning in practical contexts via Active Learning. For Curriculum Design, the results highlight that subject context is not automatically supportive: AI concepts are best learned when concretized in familiar, hands-on disciplinary knowledge (as in the cell-classification and neuron lessons) rather than only through abstract academic texts. The study's emphasis on assessing students' AI Ed Evaluation through both conceptual and interdisciplinary-reasoning measures offers a model for evaluating integrated AI learning.

The authors also offer practical guidance for educators: activate prior subject knowledge before introducing AI, use age-appropriate materials (converting academic articles into accessible formats, possibly with generative AI assistance), select contexts where students are competent and confident, and design scalable, modular lessons that fit within state-mandated standards and existing buffer time.

Connected Concepts

Citation

Zha, S., Maulucci Bragdon, M., Gong, N., Wang, J., Leavesley, S., Eaton, R., & Bosarge, E. (2025). A case study of integrating AI literacy education in a biology class. International Journal of Artificial Intelligence in Education, 35, 2453–2477.