AI Literacy

In the [YOU] on AI Field Guide

The cycle that began with [YOU] on AI is, in one of its central registers, a literacy project: it attempts to give readers the conceptual tools to see the AI transition clearly rather than through the lens of hype or fear. AI literacy is the name Breazeal gives to the same project oriented toward the general public rather than the technologically initiated. A population that does not understand AI is a population that will relate to it naively, attributing to it understanding it lacks, trusting it where it should not be trusted, failing to see the human choices and interests embedded in its design and deployment.

Breazeal’s literacy framework is grounded in her decades of studying what actually happens when people encounter social machines. She knows that the human disposition to attribute inner states to responsive, expressive things is not defeated by knowledge of the machine’s architecture: the people who walked up to Kismet and found themselves talking to it, adjusting their behavior to its apparent moods, had not lost their critical faculties. They were expressing faculties that operate below the level of conscious belief, faculties that evolved for a world in which a responsive, expressive entity was always also a sentient one. AI literacy does not neutralize these faculties; it provides the conceptual framework for understanding them and for making informed choices about how to engage with the machines that trigger them.

The pedagogical approach reflects the constructionist tradition Breazeal inherited from Seymour Papert. Understanding is not transmitted; it is constructed through activity. Day of AI does not tell students what AI is; it has them discover what AI does and does not do by engaging with it, by examining its failures alongside its successes, by asking who made a particular design choice and who is served by it. This is the same discipline that animated her robots: the making reveals what the made thing is, more honestly than any description of it from the outside. A child who has discovered, through her own exploration, that a facial recognition system fails more often on darker skin tones than lighter ones has acquired a form of AI literacy that no lecture can provide.

Origin

The concept of AI literacy emerged from the intersection of two traditions. The media literacy tradition, dating from the 1970s and 1980s, argued that a population that consumes media without understanding how it is produced, who controls it, and what interests it serves is a population vulnerable to manipulation through the very media it consumes. Breazeal brought this tradition to bear on AI: a population that uses AI without understanding how it is built, who controls it, and what interests it serves is a population vulnerable to the specific forms of manipulation that AI enables—the fluency-authority decorrelation that passes confident confabulation as reliable information, the emotional attunement that recruits social instincts in service of engagement rather than flourishing, the personalization that optimizes for retention rather than the user’s actual interests.

Breazeal co-founded the MIT RAISE (Responsible AI for Social Empowerment and Education) initiative in 2018, returning to MIT after her industry venture, and built Day of AI—a free curriculum designed to bring AI literacy to K-12 students across the United States and internationally. The curriculum was developed with the same seriousness she had brought to her robots: grounded in research about how people actually learn, oriented toward the ethical dimensions rather than merely the technical ones, and tested in classrooms before being widely deployed. By 2025 it had reached more than a million students in multiple countries.

The emphasis on ethics is not incidental but structural. Breazeal had learned from her own work that the moment a machine becomes social, design becomes an ethical act: the machine will elicit attachment and trust whether or not the designer intended it, and the designer has therefore assumed a responsibility that a purely technical framing cannot discharge. AI literacy is the other side of this responsibility: equipping the people who encounter the machine with the understanding they need to engage with it on their own terms rather than the terms the machine’s designers have set.

Key Ideas

Literacy Beyond the Technical. AI literacy includes but is not exhausted by technical understanding of how AI systems work. The full concept encompasses the capacity to evaluate AI applications ethically and politically: to ask whose interests a system serves, what assumptions are embedded in its training data, what it does to the people who use it and are affected by it, and what governance arrangements would be needed to ensure that its development serves broadly distributed human flourishing rather than the interests of the institutions that control it.

Constructionist Pedagogy. Genuine understanding of AI is built through activity rather than transmitted through instruction. A student who has engaged with an AI system, discovered a specific failure mode, and grappled with the ethical question it raises has acquired knowledge that a lecture about the same failure mode cannot provide. The constructionist approach—learning by making and doing rather than by receiving—is both more effective and more honest about what understanding requires: the encounter with the actual system in its actual behavior, including its failures.

Literacy as Empowerment. The aspiration of AI literacy is not merely that people understand AI better but that they are equipped to participate as active agents in determining what AI does to and for them. A population literate in AI is a population capable of demanding accountability from the institutions that develop and deploy it, of participating in governance decisions about its use, and of making informed choices about how it enters their own lives. This is the difference between a mass of users and a public of citizens.