Agentic education · Learning science

Actionable companion · Feedback & scaffolding

Learning with agency: Predicting Age and Visual-Motor Integration Using Origami Photographs: Deep Learning Study

AI-assisted editorial implementation guide, separate from the original research. Proposed actions are not claims or recommendations attributed to the source authors. No independent peer review of this guide is asserted.

Read original paperSearch paper agentCC BY 4.0

Evidence and context

Title and abstract screening; stage labels describe reported settings, not verified numerical ages. Design labels require full-text appraisal. Reported stage(s): Early childhood · preschool, Children · stage unspecified. Verify the participant population and task before transferring the findings.

Original authors’ abstract, reproduced under the linked source license:

Abstract Background Origami is a popular activity among preschool children and can be used by therapists as an evaluation tool to assess children’s development in clinical settings. It is easy to implement, appealing to children, and time-efficient, requiring only simple materials—pieces of paper. Furthermore, the products of origami may reflect children’s ages and their visual-motor integration (VMI) development. However, therapists typically evaluate children’s origami creations based primarily on their personal background knowledge and clinical experience, leading to subjective and descriptive feedback. Consequently, the effectiveness of using origami products to determine children’s age and VMI development lacks empirical support. Objective This study had two main aims. First, we sought to apply artificial intelligence (AI) techniques to origami products to predict children’s ages and VMI development, including VMI level (standardized scores) and VMI developmental status (typical, borderline, or delayed). Second, we explored the performance of the AI models using all combinations of photographs taken from different angles. Methods A total of 515 children aged 2-6 years were recruited and divided into training and testing groups at a 4:1 ratio. Children created origami dogs, which were photographed from 8 different angles. The Beery–Buktenica Developmental Test of Visual-Motor Integration, 6th Edition, was used to assess the children’s VMI levels and developmental status. Three AI models—ResNet-50, XGBoost, and a multilayer perceptron—were combined sequentially to predict age z scores and VMI z scores using the training group. The trained models were then tested using the testing group, and the accuracy of the predicted VMI developmental status was also calculated. Results The R2 of the age and the VMI trained models ranged from 0.50 to 0.73 and from 0.50 to 0.66, respectively. The AI models that obtained an R2>0.70 for the age model and an R2>0.60 for the VMI model were selected for model testing. Those models were further examined for the accuracy of the VMI developmental status, the correlations, and the mean absolute error (MAE) of both the age and the VMI models. The accuracy of the VMI developmental status was about 71%-76%. The correlations between the final predicted age z score and the real age z score ranged from 0.84 to 0.85, and the correlations of the final predicted VMI z scores to the real z scores ranged from 0.77 to 0.81. The MAE of the age models ranged from 0.42 to 0.46 and those of the VMI models ranged from 0.43 to 0.48. Conclusion Our findings indicate that AI techniques have a significant potential for predicting children’s development. The insights provided by AI may assist therapists in better interpreting children’s performance in activities.

What kind of evidence is this?

Supporting AI learning science · Design requires full-text appraisal

Stage: Early childhood · preschool; Children · stage unspecified. Screening labels describe reported context and study-design language. They are not a risk-of-bias rating or proof of effectiveness for an age band.

Participant ages · source passage

Methods A total of 515 children aged 2-6 years were recruited and divided into training and testing groups at a 4:1 ratio.

Study methods · source passage

The data included in our study were collected as part of an ongoing AI project (An Artificial Intelligence System for Assessing Gross Motor and Fine Motor, Handwriting, Attention and Emotion in Children with Developmental Delay), and some of the data were used in the preliminary study [13]. Preschool children enrolled in kindergarten were recruited to the study. The inclusion criteria were as follows: aged 2-6 years, ability to complete group administrations, and informed consent given by caregivers. The exclusion criterion was any obvious physical or mental disability.

Limitations and uncertainty · source passage

Several limitations were noted in this study. First, a small sample size of children aged 24 to 29 months (n=10) was used. This limited dataset may have resulted in insufficient data for AI learning, potentially increasing prediction errors for this age group. Second, most children recruited in the study exhibited typical development, and those with developmental conditions had mixed diagnoses. Consequently, the results may predominantly reflect the performance of typically developing children. Moreover, the model may tend to predict no developmental issues, increasing the likelihood of false negative results. Third, the images used for the model prediction were restricted to dog origami products, as only these were used to train the model. This limitation restricts the model’s applicability and generalizability to other origami forms. Fourth, due to constraints in evaluation time and research personnel, only the standardized VMI score was used for model prediction. However, executing origami activities also requires manual dexterity, cognition, and visual perception, which were not included in the assessment. Fifth, since the predictions were primarily based on an AI model and conducted via the internet, there are concerns about children’s privacy and data security. Furthermore, it is recommended that clinicians ensure appropriate use and understanding when interpreting the results. Future studies should consider using standardized assessments of other developmental dimensions as model prediction outcomes to provide a more comprehensive understanding of children’s development through origami activities. Additionally, increasing the sample size of young children and children with development conditions and incorporating different origami products in model training would enhance the model’s utility and robustness.

Design the system before the pilot

Use this study to inform a learning mechanism or implementation question. It is not direct evidence that autonomous agents improve learning. Keep the proposed agent behavior separate from what the source actually tested.

  • Give feedback on the learner’s reasoning, then fade hints. Record whether the learner can complete the next task without that scaffold.
  • Include a delayed assessment and an unfamiliar transfer problem without AI. Immediate supported performance cannot establish durable learning.
  • Use common rubrics, blind scoring where feasible and a fresh unaided reasoning task. Document permissible assistance and avoid treating AI-detector scores as proof of misconduct.

A 90-day district pilot

  1. Days 1–15 · Appraise. Have the curriculum lead and a teacher read the methods and limitations. Record the study design, sample, comparator, outcomes and evidence gaps. Define a local question that the study can help inform.
  2. Days 16–30 · Design. Use a brief adult-mediated story, turn-taking or language activity. An educator controls every interaction; the agent never becomes an unsupervised companion. Preserve hands-on play and human conversation. Name an accountable owner and set the baseline, success threshold and budget before implementation.
  3. Days 31–60 · Test. Start with a small voluntary cohort, preserve an accessible non-AI route and require teacher review. Use synthetic or de-identified examples; approve data handling and tool terms through district processes.
  4. Days 61–90 · Decide. Measure independent naming, explanation or turn-taking in a human-led activity after the tool is removed. Record frustration, overstimulation and the amount of adult help. Compare with baseline, document uncertainty and stop or revise if learning worsens, access is unequal or critical errors persist.

The board decision brief

Report the learning problem, evidence quality, participating population, total cost including review time, measured benefit, errors, access differences and remaining uncertainties. Make a dated recommendation to stop, revise or expand. A positive perception survey alone does not establish learning improvement.

Attribution and reuse

Chien-Yu Huang; Yen-Ting Yu; Kuan-Lin Chen; Jenn-Jier Lien; Gong-Hong Lin; Ching-Lin Hsieh. Predicting Age and Visual-Motor Integration Using Origami Photographs: Deep Learning Study. JMIR Formative Research (2025-01-10). https://doi.org/10.2196/58421

Companion prepared 2026-10-30. Original authorship and DOI are preserved. The abstract is an attributed source extract; the implementation plan is new AI-assisted editorial material, offered under the same CC BY 4.0 license. This guide has not undergone independent peer review.