Title: Learning To Read Code First: Evaluating a Novel Pedagogy for Learning Programming with an Interactive Visual Tutoring System
Advisors: Amy Ko (iSchool) and Steve Tanimoto
Abstract: What knowledge does learning programming require? Prior work has focused on theorizing and teaching writing and problem solving skills. We reconceptualize program reading as a separate skill from writing, and propose a novel formal theory of program tracing knowledge based on symbolic execution. Within program reading, we separate program literary comprehension and program behavior comprehension (which includes program tracing), and provide a new explanation for the “computer is like a person” misconception among novices. We use this theory to argue for initially separating the learning of reading and writing, as well as carefully delineating them throughout learning programming. Based on our formal theory of program tracing knowledge, we propose pedagogical principles for teaching program tracing within a program reading curriculum. To assess learning within this new pedagogy, we build a new tutorial system - PLTutor - with novel technical capabilities that enable 1) curricular flexibility 2) assessments of reading knowledge at multiple levels of granularity 3) visualization of program--instruction--machine-state relationships with code token and sub-expression granularity. We evaluate learning gains among self-selected CS1 students using a block randomized controlled lab study comparing PLTutor with Codecademy(a writing tutorial). We cautiously interpret our results due to small sample size and assessment validity concerns. We find some evidence of improved learning gains on the SCS1, with average learning gains 66% higher than Codecademy (3.94 vs. 2.37 out of 27 questions).