AI Made by Youth: A Conversational AI Curriculum for Middle School Summer Camps

Yukyeong Song, Gloria Ashiya Katuka, Joanne Barrett, Xiaoyi Tian, Amit Kumar, Tom McKlin, Mehmet Celepkolu, Kristy Elizabeth Boyer, Maya IsraelAAAI 2023 · DOI

Abstract

As artificial intelligence (AI) becomes integrated into the fabric of daily life, the need for K-12 AI literacy has never been more urgent. This paper details the development and evaluation of a specialized curriculum and a custom-built development platform, AMBY (AI Made By You), designed to empower middle schoolers to move from passive consumers to active creators of AI. Through building conversational agents, students demonstrated significant gains in confidence, collaborative sharing, and intent to persist in AI-related fields.

Introduction

From voice assistants like Alexa to complex recommendation algorithms, AI is ubiquitous. However, teaching the underlying mechanics of these systems to young learners presents significant pedagogical challenges. Traditional computer science curricula often feel abstract, while professional-grade development tools—such as Google’s Dialogflow—frequently overwhelm students with steep learning curves and dense, adult-oriented interfaces.

To bridge this gap, researchers at the University of Florida developed a curriculum centered on conversational AI. By leveraging students' existing familiarity with chatbots, the program introduces complex concepts like machine learning, data training, and natural interaction through a relatable and intuitive "hook."

The AMBY Platform and Curriculum Design

A core contribution of this work is the AMBY (AI Made By You) platform. Unlike professional tools, AMBY provides a visual, scaffolded environment specifically for middle schoolers. It features a "Playground" where students can map out conversational flows and a testing panel for real-time interaction with their agents.

The AMBY interface showing the "Jerry Berry" chatbot project, which includes main intents, follow-up intents, and a real-time testing panel.

The curriculum is structured around four primary components:

  1. General AI Lessons: Introducing perception, data representation, and algorithmic bias.
  2. Unplugged Activities: Physical logic exercises, such as a "Yoga activity" where students simulate if-then logic through body poses.
  3. Conversational AI Lessons: Direct instruction on chatbot architecture and intent mapping.
  4. Project Activities: Hands-on development using a Design Thinking framework.

Following the "Use → Modify → Create" pedagogical model, students begin by interacting with existing bots, then progress to tweaking code, and finally build original projects from scratch. This progression ensures that students build technical competence alongside conceptual understanding.

Methodology and Evaluation

The researchers evaluated the curriculum over two years of summer camps (2021 and 2022). The 2022 study involved 32 middle school students (rising 7th and 8th graders). Evaluation metrics included pre- and post-camp attitudinal surveys, video assessments where students explained "What is AI?", and an analysis of the 58 unique conversational projects created by the participants.

One notable project, Jerry Berry, exemplified the students' ability to apply Design Thinking to social goals. This bot was designed to educate users about influential Black historical figures, demonstrating how students could use AI as a tool for social empowerment.

Results and Key Insights

Quantitative analysis revealed statistically significant improvements in student attitudes toward AI. The curriculum successfully bolstered students' self-efficacy and their desire to engage with the technology further.

Statistical analysis of student attitudinal constructs, showing significant pre-to-post improvements in ability beliefs, sharing, and persistence.

Key findings include:

  • Ability Beliefs: Students felt significantly more capable of solving AI problems, with mean scores rising from 2.91 to 3.30 ($p=0.006$).
  • Sharing and Persistence: There were marked increases in students' willingness to discuss their work with others ($p=0.049$) and their intent to pursue future AI learning opportunities ($p=0.019$).
  • Conceptual Shift: Qualitative video data showed a transition in how students defined AI. Initially, students defined AI through examples (e.g., "it's a robot"). Post-camp, they defined it through its functional characteristics, specifically identifying that AI systems "learn" from human-provided input data.

Discussion and Future Work

The transition from Google Dialogflow to the custom AMBY platform was a turning point in the study. The researchers found that age-appropriate tooling is just as critical as the lesson plan itself; reducing interface friction allowed students to focus on high-level conceptual design rather than technical troubleshooting.

While the study did not see a significant shift in "Identity" (students seeing themselves in AI careers), the authors suggest this may be due to the informal, exploratory nature of the summer camp. Future research will focus on long-term knowledge retention and expanding the curriculum to more diverse educational settings to foster a broader, more inclusive generation of AI creators.

Summary updated

BibTeX
@inproceedings{song2023ai,
  author = {Yukyeong Song and Gloria Ashiya Katuka and Joanne Barrett and Xiaoyi Tian and Amit Kumar and Tom McKlin and Mehmet Celepkolu and Kristy Elizabeth Boyer and Maya Israel},
  title = {{AI Made by Youth: A Conversational AI Curriculum for Middle School Summer Camps}},
  booktitle = {AAAI Conference on Artificial Intelligence},
  year = {2023},
  doi = {10.1609/aaai.v37i13.26882},
}
Figure