A model built to teach
A one minute film about why DuckyHelper teaches the way it does. A student asks for the answer, a real study shows why answers are not learning, and seven principles from real papers each meet the part of the app that uses them. Narrated, with captions.
What it shows
- The real pill, close up: a learner asks, can you just tell me the answer?
- A real page of a PNAS study of AI in high school math (Bastani et al. 2025), highlighted where it finds that plain GPT-4 can harm learning, and where a tutor with safeguards largely avoids the harm.
- A wall of real first pages of papers on how people learn, counting up to 100+.
- 1,000,000+ tutoring examples, then the model card: one model, built to teach.
- Seven principles, each on its own real paper page (Bloom 1984, Bastani et al. 2025, Mayer and Moreno 2003, Sweller and Cooper 1985, Roediger and Karpicke 2006, Chi et al. 1994, Cepeda et al. 2006), each met by the real DuckyHelper pill: a reply that guides instead of answering, the pen drawing a pendulum, worked steps, a quiz, and a quiz again three days later.
Chapters
One film, seven papers
Each principle in the film stands on one real paper, numbered as the film numbers them. Pick one to see its beat, then read the page that explains the research and what DuckyHelper does because of it.
Teach one on one
Bloom (1984) found students tutored one on one scored about two standard deviations above a regular class. The film shows the paper's real figure. The 2 sigma problem
Guide, don't hand over answers
The same PNAS page, highlighted where a tutor with safeguards largely avoided the harm, against Ducky's real reply: let's get there together. Why answers are not learning
Draw it as you explain
Mayer and Moreno (2003): people learn more from words with pictures. Ducky's real pen underlines the line, then draws a pendulum. Pictures and words
Show a worked example
Sweller and Cooper (1985): for beginners, studying a solved problem teaches more than solving alone. Then the real Steps card. Worked examples
Quiz, don't re-read
Roediger and Karpicke (2006): a week later, the tested group recalled 56% and the re-reading group 42%. The paper's own Figure 1, then a real quiz. Retrieval practice
Have them explain it back
Chi and colleagues (1994): students prompted to explain each line to themselves understood more. The learner's own words fill the screen. Self-explanation
Space it out
Cepeda and colleagues (2006), 317 experiments: spaced review lasts longer than cramming. The pill says it will quiz you again in three days. Spaced practice
Trained on 1,000,000+ tutoring examples
One real exchange pulls back to a million, then the model card: one model, built to teach. It is trained on 1,000,000+ tutoring examples and inspired by 100+ learning-science papers. How the model is built to teach
Transcript
(On screen, a learner asks: Can you just tell me the answer?)
A study in PNAS found answers can hurt.
So we asked: how do people actually learn?
More than a hundred papers on learning.
Trained on over a million tutoring examples.
One model. Built to teach.
Teach one on one.
Guide. Don't hand over answers.
Draw it as you explain.
Show a worked example.
Quiz, don't re-read.
Have them explain it back.
Space it out.
DuckyHelper. A tutor, not an answer machine.
Sources
- Bastani, H., Bastani, O., Sungu, A., Ge, H., Kabakcı, Ö., & Mariman, R. (2025). Generative AI without guardrails can harm learning: Evidence from high school mathematics. Proceedings of the National Academy of Sciences, 122(26), e2422633122. (accessed 2026-10-09)
- Bloom, B. S. (1984). The 2 Sigma Problem: The Search for Methods of Group Instruction as Effective as One-to-One Tutoring. Educational Researcher, 13(6), 4-16. (accessed 2026-10-09)
- Mayer, R. E., & Moreno, R. (2003). Nine Ways to Reduce Cognitive Load in Multimedia Learning. Educational Psychologist, 38(1), 43-52. (accessed 2026-10-09)
- Sweller, J., & Cooper, G. A. (1985). The Use of Worked Examples as a Substitute for Problem Solving in Learning Algebra. Cognition and Instruction, 2(1), 59-89. (accessed 2026-10-09)
- Roediger, H. L., & Karpicke, J. D. (2006). Test-Enhanced Learning: Taking Memory Tests Improves Long-Term Retention. Psychological Science, 17(3), 249-255. (accessed 2026-10-09)
- Chi, M. T. H., De Leeuw, N., Chiu, M.-H., & Lavancher, C. (1994). Eliciting Self-Explanations Improves Understanding. Cognitive Science, 18(3), 439-477. (accessed 2026-10-09)
- Cepeda, N. J., Pashler, H., Vul, E., Wixted, J. T., & Rohrer, D. (2006). Distributed practice in verbal recall tasks: A review and quantitative synthesis. Psychological Bulletin, 132(3), 354-380. (accessed 2026-10-09)
Ready to learn
DuckyHelper is ready on your Mac. Learn with Eddy on its blackboard, or with Ducky right on your screen.
