Data Comics for Education: Evaluating Effectiveness, Benefits, and the Ethics of AI-Assisted Creation

Synthesis: Data comics combine sequential visual narratives with data Visualization to improve student engagement with Generative AI in educational settings. This paper evaluates the effectiveness of AI-assisted creation of data comics, finding that they significantly enhance student engagement and comprehension compared to traditional visualization formats. The study also examines ethical dimensions including authorship attribution of AI-produced content, accuracy of generated visuals, and the risk of misleading representations. These findings have direct implications for K-12 education, where Active Learning approaches benefit from engaging visual materials.

Key Findings

• In a within-subjects study with 60 university students, participants consistently performed better with GenAI-assisted data comics than with conventional visualizations across information retrieval and comprehension tasks, with the largest advantage in insight comprehension tasks.
• The comprehension benefit held independent of prior visualization literacy, suggesting data comics can support learners regardless of their starting skill with charts and graphs.
• Students rated data comics as more engaging and easier to understand than conventional visualizations, and several highlighted the value of the narrative style for information recall — although the authors note this construct was not directly measured.
• Participants raised ethical concerns about GenAI-driven misinformation and ownership, pointing to authorship attribution of AI-produced content and the accuracy of generated visuals as open questions.
• Perceived limitations were widespread: two-thirds of participants (N = 36) flagged downsides, most commonly information overload (N = 18), with "too busy" layouts slowing the location of key insights.

Study Design & Method

The study used a within-subjects experimental design in which 60 university students completed information retrieval and comprehension tasks using both conventional visualizations and data comics created with assistance from generative AI tools. Task performance was compared across the two formats, and students also provided qualitative feedback on engagement, comprehension, and perceived limitations. The authors frame the work against the challenge of preparing students from diverse disciplines to interpret and use data for reasoning and critical thinking in their future professional practice.

What this means for practice

• Instructors. Use AI-assisted data comics when the goal is comprehension of insights rather than lookup: correct rates were higher with comics (median 0.750) than with conventional visualizations (median 0.333), and the benefit held independent of prior visualization literacy.
• Instructors. Earmark comics for multiple-insight questions, where the advantage over conventional visualizations was largest (median 1.000 versus 0.333).
• Learners. Check AI-generated panels against the underlying data before citing them — 45 of the 60 participants raised misinformation as a concern, and hallucinated or altered figures travel silently once embedded in a narrative.
• Designers. Keep panels deliberately sparse: of the 36 participants who named a downside, 18 pointed to information overload, with "too busy" layouts slowing the location of key insights.
• Instructors. Have students document the generative tool, prompt, and source data behind each comic, since 37 participants flagged ambiguous ownership of AI-generated and reprocessed images, and require that kind of transparency from any Generative AI pipeline used for instructional visuals.

Limitations

• The 60 participants were predominantly from similar academic backgrounds (STEM), which the authors state limits generalizability to groups with different professional contexts.
• Only four pairs of visualizations were compared, so the breadth of data-comic and visualization types assessed is narrow.
• Evaluation covered only the first two levels of Bloom's taxonomy — retrieving data points and comprehending insights — leaving more complex cognitive tasks such as manipulating data stories untested.
• The comics were not embedded in instructional activities or connected to learning outcomes, and the authors note that creation still requires substantial human involvement.

Citation

Zirui Shan, Vanessa Echeverria, Yuheng Li, Yi-Shan Tsai, Roberto Martinez-Maldonado (2026). Data Comics for Education: Evaluating Effectiveness, Benefits, and the Ethics of AI-Assisted Creation.