Abstract
This paper describes an experience in teaching Machine Learning (ML) and Natural Language Processing (NLP) to a group of high school students over an intense one-month period. In this work, we provide an outline of an AI course curriculum we designed for high school students and then evaluate its effectiveness by analyzing student's feedback and student outcomes. After closely observing students, evaluating their responses to our surveys, and analyzing their contribution to the course project, we identified some possible impediments in teaching AI to high school students and propose some measures to avoid them. These measures include employing a combination of objectivist and constructivist pedagogies, reviewing/introducing basic programming concepts at the beginning of the course, and addressing gender discrepancies throughout the course.
Introduction
As computer science education expands within K-12 curricula, it often remains limited to introductory programming. There is a critical need to demystify "black box" technologies like Machine Learning (ML) and Natural Language Processing (NLP) for younger learners. However, introducing these advanced topics to high school students presents unique challenges. Students often possess varying levels of mathematical maturity and coding experience. Furthermore, many enter with "sci-fi" perceptions of AI—viewing it through the lens of sentient robots—rather than understanding it as a technical discipline rooted in data patterns. Finally, instructors must balance "objectivist" lecture-based learning with "constructivist" project-based exploration to avoid overwhelming students who lack foundational skills.
Methodology
The authors developed and implemented a four-week intensive curriculum at the UC Santa Cruz COSMOS summer program. The pedagogical approach was a hybrid of two distinct phases:
Phase 1: Foundations (Weeks 1-2)
This phase utilized an objectivist approach, delivering structured content via Google Colab and Python. Students were introduced to basic programming constructs—including loops, functions, and data structures—alongside fundamental ML concepts such as linear and logistic regression, neural networks, and NLP preprocessing.
Phase 2: Application (Weeks 3-4)
The curriculum transitioned to a constructivist model where students worked in teams of 3–4 on real-world projects. Instructors shifted roles from lecturers to facilitators. To ensure technical success, the team provided "cleaned" versions of popular datasets, such as the Ames Housing Price Prediction for regression and Amazon Fine Food Reviews for NLP sentiment analysis.
Key Insights and Analysis
The study revealed several critical shifts in student understanding and highlighted persistent social dynamics in technical education.
Redefining Artificial Intelligence
Initially, students defined AI through behaviorist lenses (human imitation). By the conclusion of the course, definitions shifted toward an engineering-centric view: "processing large amounts of data to learn patterns and trends," indicating a more realistic understanding of the field.
The Gendered Confidence Gap
A significant disparity in self-reported coding confidence was observed at the start of the program. While only 35% of male students reported being "not confident," a staggering 66% of female students expressed the same sentiment.
This gap influenced group dynamics during the project phase. In mixed-gender groups, male students frequently gravitated toward "technical" roles (model implementation and results analysis), while female students often handled "non-technical" aspects like introductions and preprocessing. Interestingly, in groups where female students were the majority, they were more likely to take on technical leadership roles and reported higher overall satisfaction.
Results and Discussion
The program was largely successful, with 71% of students reporting they fully met their learning goals. All teams successfully completed a technical project, culminating in a 20-minute presentation and a conference-style poster. Exposure to real-world applications through field trips (e.g., Microsoft) and guest lectures on topics like "Fairness in ML" proved more effective for teaching ethics than abstract discussions.
However, limitations remain. The small, self-selected cohort may not represent the broader population. Furthermore, despite intensive training, many students reached a "technical ceiling," requiring significant scaffolding to implement models from scratch without assistance.
Conclusion
Teaching AI to high schoolers requires a careful blend of foundational instruction and hands-on application. Addressing the gender confidence gap through intentional group formation and early programming support is essential for fostering an inclusive and effective learning environment. Future work should focus on longer curriculum durations to solidify the programming skills necessary for true technical independence.
Summary updated