Say What? Examining Text and Voice Input Modalities for Prompt-Based Programming in Computing Education

Synthesis: Nearly all prior research on LLMs in computing education has used text input, yet voice-enabled interfaces are becoming common. This exploratory study investigated how introductory programming students interact with Prompt Problems — tasks requiring natural-language prompts to generate correct code — under free choice of text or voice (N = 919). For two of three problems, students who typed were more likely to succeed on the first attempt than those submitting unedited voice prompts; editing transcribed voice prompts before submission erased the gap. Most students tried and preferred text, though some used voice complementarily. Qualitative analysis revealed perceived roles, drawbacks, and advantages of each modality, with implications for Multimodal AI tools and instructional design.

• Modality matters for first-attempt success: Unedited voice prompts underperformed typed ones on two of three problems; editing transcribed voice closed the gap — extending Understanding Student Perceptions, Mistakes, and Debugging Approaches when Solving Natural Language Programming Tasks.
• Student preference for text: Despite voice availability, most students chose and preferred typing, relevant to Student Experience design.
• Complementary use: Some students mixed modalities, suggesting non-preferential, context-dependent strategies tied to Teaching Prompt-Based Programming with LLMs: A 45-Minute Lesson with Guided Practice for End-User Programmers practice.
• Instructional design: Findings inform multimodal CS Education tooling and AI Literacy around prompt construction.
• Scales Prompt Problems research: Large N (919) builds on the Programming Intelligent Tutoring Systems Prompt Problems line of work.

What this means for practice

• Learners. Type your first prompt when the wording has to be exact, and reserve voice for rough drafting: compared with text prompting, unedited voice prompts had lower odds of immediate success on two of the three problems (OR = 0.50, 95% CI [0.27, 0.89]; OR = 0.43, 95% CI [0.23, 0.83]).
• Learners. Read and fix the transcript before submitting if you do dictate, because students who edited their voice prompts did not differ significantly from typers (OR = 0.57 and OR = 0.72, both n.s.) — editing erased the gap.
• Instructors. Offer voice as an option rather than a default for prompt-based tasks: 813 students used text only (88.5%), 44 used voice only (4.8%), and 62 tried both (6.7%), and those who did use voice were not deterred by it.
• Instructors. Keep a text fallback for non-native English speakers, who cited transcription errors as a reason for preferring typing.
• Designers. Ship an edit-before-send step and surface transcription quality in Multimodal AI prompt tools, since the unedited-versus-edited voice gap is the clearest actionable difference in the study.

Limitations

• Students self-selected into modality and few chose voice (44 voice-only users of 919), limiting generalizability, and causality cannot be established — whether modality influences outcomes or student characteristics drive modality choice.
• Only 62 students attempted both prompting approaches, so the within-problem persistence analysis never attained sufficient sample sizes.
• Some students, including non-native English speakers, reported transcription issues, so evaluations may reflect technology performance rather than the modality itself, and transcription delay was not measured.
• The three problems were relatively simple and short, and modality preference rests on two self-report reflection questions rather than a validated technology-acceptance instrument.

Citation

Riegel, K., Hua, Y. C., Denny, P., Pădurean, V.-A., & Leinonen, J. (2026). Say What? Examining Text and Voice Input Modalities for Prompt-Based Programming in Computing Education.