Fail Forward

In the AI Story

The framing comes from design thinking but its developmental grounding is older, running through Bandura's mastery experience research and productive failure work in educational psychology. The common finding across these traditions: learners develop most robustly when given genuine opportunities to fail, followed by structured reflection that converts the failure into understanding. The failure is not the opposite of learning; it is the mechanism. Remove the failure and you remove the learning.

Applied to AI, the practice has a specific grammar. A parent who discovers her child submitted AI-assisted work does not confiscate the laptop. She asks questions. "Show me what you asked it. Show me what it gave you. Which parts did you keep? Why those parts? What would you change if you did it again?" The questions are not punitive and they are not disguised lectures. They are genuine inquiries into the child's process, and they produce the metacognitive reflection that builds the capacity to evaluate AI output critically over time.

The practice's most difficult requirement is parental restraint. Parents trained by worst-first thinking respond to a child's AI mistake with the instinct that more supervision is needed. Fail forward asks parents to respond instead with the instinct that more conversation is needed — and to trust that the conversation, repeated across dozens of encounters, will produce the judgment that surveillance never could. The trust is hard because its evidence accumulates slowly; the child does not become a more critical AI user overnight. She becomes one across months of scaffolded reflection, and the parent must tolerate the intermediate period during which the capacity is being built but not yet demonstrated.

The institutional analog is the teacher's shift from evaluating outputs to evaluating process — grading questions rather than essays, as the Brooklyn teacher Skenazy cites did. This is fail-forward at the classroom level: students are not punished for producing imperfect AI-assisted work; they are required to reflect on the imperfections and learn from them. The reflection is the education. The essay is merely the occasion for it.

Origin

Skenazy adapted the fail-forward concept from design thinking literature and integrated it with the self-efficacy research that has grounded her framework since the Free-Range Kids founding. The AI application emerged in her writing and speaking during 2024–2026 as schools began implementing prohibition-based AI policies.

Key Ideas

Mistakes as curriculum. The specific mistakes a child makes with AI are not failures to be eliminated but data to be examined, and the examination is where the learning lives.

Non-punitive structure. Fail-forward conversations require parents and teachers to distinguish genuine curiosity from disguised evaluation; the child's honest engagement depends on the distinction.

Metacognition as goal. The practice aims to develop the child's capacity to think about her own thinking — the meta-level awareness that is the actual skill the AI age demands.

Restraint as discipline. The hardest part of fail forward is not designing the reflection but resisting the instinct to prevent the mistake in the first place.