What the research does and does not show.

What is well supported

Girls' interest in technology peaks in early adolescence and declines through the teen years. This is the finding the whole programme is built around, and it is consistent across multiple surveys, including Microsoft's research with KRC and work by the National Girls Collaborative Project.

The gap is not ability. Girls match boys on measured performance through this period. What diverges is confidence, belonging, and whether she sees herself as someone who does this.

Teaching coding in a weekly club does teach coding. A randomised controlled trial of year-long weekly code clubs by the National Foundation for Educational Research found a clear positive effect on coding skill, 5.60 against 3.81 on a coding quiz, p < .001. Self-rated ability rose too: 63% of club pupils rated themselves good at making things with code, against 42% of the control group.

What is not supported, including things that work against our own pitch

That same trial found no effect on computational thinking. Both groups improved by roughly the same amount and the difference between them was not statistically significant.

It found no gain in transferable skills such as problem solving or working in a team.

It found no gain in interest in technology careers. Interest in a coding job fell over the year, from 55% at the start to 44% in the club group and 46% in the control group. A year of weekly club did not make those children want a technology career more.

That trial was with Year 5 pupils in England, in school-run clubs, teaching general coding rather than AI, and it is a single study. But it is the closest published test of our own format, and it does not show the outcome this category usually promises.

What the strongest positive evidence actually measures

The best-evidenced result in this field is an independent evaluation of Girls Who Code by the American Institutes for Research. Comparing participants with similar waitlisted students using national enrolment records, it found participants were 13.2 percentage points more likely to major in a computer-science-related field, and 11.5 points for the self-paced version.

Two caveats worth knowing. It measures an intensive summer immersion programme and a self-paced course, not a weekly year-long club. And the evaluators state they could not control for prior grades or motivation.

What we will not claim

• We will not tell you this programme is proven to keep your daughter in technology. Nobody has shown that about a programme like ours.
• We will not quote outcome percentages from our own families until families have finished a year. When we do, we will state how we measured and how many answered.
• We will not present a testimonial as evidence of effect.

What we do claim, and how you check it

Every class ends with something she made. You can see it: sessions are recorded and her work lands in her portfolio. Four times a year she presents it to you at a showcase. If that is not happening you will know within a week, and you can cancel from your portal without talking to anyone.

That is a promise about what we do rather than about who she becomes. It is the only kind worth making before we have a single graduating cohort.

Sources

• Randomised Controlled Trial and Process Evaluation of Code Clubs, National Foundation for Educational Research (PDF). Trial registration ISRCTN17195519.
• Girls Who Code Impact Evaluation Study, American Institutes for Research.
• Girls Who Code Program Evaluation: Final Report, American Institutes for Research, March 2024 (ERIC ED647331).

If you think anything on this site overstates the evidence, write to support@aischoolforgirls.com and it gets corrected here.