Wadler's Blog

Logo, and following on from it Scratch, Alice, and Kodu, promotes an implicit claim that in the right environment children can learn programming almost by osmosis, and that absorbing this skill improves other skills, such as planning. These implicit claims were challenged twenty-five years ago by Roy D. Pea, who discovered that school children exposed to Logo could still be confused by recursion after writing recursive programs, and that learning Logo did not improve their planning ability. Here is an overall summary, here is a paper specifically on recursion, and here is Pea's reply to an attack by Papert.

I found the second paper (Children's mental models of recursive LOGO programs) particularly interesting, as it helps to understand some of the issues my first-year students may face in learning recursion. I'm woefully ignorant of the literature on the psychology of programming, and need to learn more. Thanks to Matthias Felleisen and Emmanuel Schanzer for pointing me at this work.

What sparked my interest was a keynote on Kodu at POPL, by Matthew MacLaurin of MSR. The laudable goal of Kodu is to permit young children to program video games, as MacLaurin put it, 'to let them know the computer can do anything'. And Kodu can achieve quite spectacular results with very simple programs, but the programming model needs to be very simple to achieve this (a list of condition-action pairs). Because Kodu is so special purpose, I worry that instead of teaching children that 'computer can do anything', it will teach them that 'computers will let you build exactly what they are designed to let you build, and anything else is very difficult'. Pea's studies are orthogonal to this concern, so I'm left wondering whether Kodu will lead to a new generation of programmers or a new generation turned off by computing.

Meanwhile, Schanzer is running Bootstrap, an outgrowth of TeachScheme. "Bootstrap is a standards-based curriculum for middle-school students, which teaches them to program their own videogames using purely algebraic and geometric concepts." The start-up costs are higher than Kodu, but once started students have the full power of Scheme/Racket at their command. Seems preferable to me, but there may well be room for both in the world. More study needed! Not to mention more resources for getting proper computing (not just ICT) into schools at an early age.