The impact and potential of early exposure to machine learning

Proponents of machine learning education on its significance and implementation for a younger generation.

Kids born today will never know a world where the technology they interact with is not intelligent. Toddlers are often playing with iPads before they can walk, and they may soon be learning to speak by using Alexas. This motivates the question: how might machine learning education fit into an increasingly AI-driven world?

Natalie Lao ’16 PhD ’20 is Executive Director of the App Inventor Foundation, a nonprofit built off of the MIT App Inventor Project that empowers students across the globe to create mobile apps for social good. She emphasizes the importance of educating students in machine learning (ML) early on. The number one reason, she stated, is simply that interacting with the world involves interacting with ML, from scrolling social media to getting a loan at a bank. The second reason is that ML knowledge is increasingly relevant as students become voters and involve themselves in policy creation and evaluation. “Students are going to become voters,” she stated, “and they need to be informed enough so that they feel competent voting on [ML] policies and asking the right questions.” For example, how are tech companies regulated? How is the usage of data regulated, such as avenues of redress if someone is training a generative AI model on data without consent? What are the court cases currently going on that people need to be aware of to protect themselves?

Lao argued the importance of educating students about these topics as young as possible. Incorporating ML knowledge into their worldview is easier when they’re younger, she explained. A report released in 2019 by Purdue University found that exposing kids to STEM early in elementary school provides students with the foundation to pursue STEM-related careers [1]. The report also finds that “girls are five percent less likely to recall learning STEM concepts in elementary and middle school”, highlighting a gender disparity that can be found in STEM fields later in life.

ML education has the potential to create profound implications in the world. To support this vision, she proposed a framework for evaluating ML competency in her PhD thesis, establishing three categories: knowledge, skills, and attitudes. Knowledge encompasses concepts including the ethics of artificial intelligence and societal impact, while practical abilities such as debugging, analyzing artifacts, and participating in conversations about AI fall into the skills bin. The third category, attitudes, Lao finds the most challenging to address. She emphasizes the role of self-efficacy theory, or the belief in one’s own ability to achieve goals, particularly among underrepresented students. Lao notes that these students often do not see people who look like them or come from similar backgrounds succeeding in AI and technology, which can lead to a decrease in self-efficacy. To counteract this, she advocates for the importance of mentorship and role modeling in ML education. By exposing students to diverse individuals who are only a few steps ahead in their careers and are successfully navigating the field, students can begin to envision a future for themselves in similar roles.

Inspired by a conversation with Lao and his passion for STEM education, Kent Brought ’26, joined a team advised by Vincent Monardo, Lecturer of Electrical Engineering and Computer Science, to create MacLea, an online educational tool for machine learning geared toward elementary and middle schoolers. In a field crowded with learning opportunities for adults, this project uniquely targets young learners, highlighting an effort to engage the next generation in machine learning early on. "In high school, I made several trips to local elementary schools to help establish robotics teams in an area where STEM is not a dominant field,” Brought reflects. Since then, he has had an interest in different ways of helping kids learn new things. 

MacLea provides guided tutorials on topics including Convolutional Neural Networks, as well as optimizers, epochs, and loss. Similar to Scratch, a block-based visual programming language created by the MIT Media Lab, users assemble blocks rather than writing text themselves to build a machine learning pipeline. Brought’s goal for the project is to prove that it is possible to make complex machine learning concepts accessible at an early age. The platform will be piloted in the 8th grade STEM program at John D. Runkle School this year, with the hope of gathering feedback and extending the tool to target younger students. 

How will ML be taught in classrooms of the future? Monardo thinks it’s an open question. “If we’re adding computer science and machine learning into the curriculum,” he contemplates, “what are we taking out?”

Lao advocates for adding AI to current computer science (CS) or technology classes, rather than a broad curriculum integration across every class, which are two ways people are approaching the issue. However, Lao also recognizes the value of AI's presence in seemingly unrelated subjects, reflecting on the evolution of internet skills in education: once a specialized topic, now a ubiquitous tool in every classroom. She draws parallels to technology she was taught in school “I had a class for typing and how to use Google. Nowadays, in every single class, you’re expected to use Google for research, you’re expected to type papers.” 

By integrating ML education into K-12 education, Lao, Brought, and Monardo hope to ensure that the next generation of students are not only consumers of technology but also its informed creators and regulators. The endeavors of educators and innovators are laying the groundwork for a future where ML is not just understood but embraced, empowering students with the knowledge and skills to thrive in an AI-saturated world.