Neurophysiological Measure of Effort in Program Comprehension

Abstract

The recent proliferation of inexpensive electroencephalography (EEG) devices is fueling a rising interest in associating detectable indicators of brain activity with human performance factors. In this thesis, the focus is on programmer effort in program comprehension tasks. Traditionally, measures of effort are made using self-reported surveys (NASA-TLX), task timing, and task accuracy. This work explores the feasibility of using EEG to produce a more direct and quantitative measure of effort. Effort is measured across a number of tasks with varying difficulty and comparisons are made between traditional and EEG measures of effort. Initially, the program comprehension tasks are ranked in order of complexity as computed by a number of classic software complexity metrics, such as Halstead’s complexity metrics and McCabe’s cyclomatic complexity. Likewise, we compute a ranking of tasks based on observed effort as a basis of comparison between EEG and complexity measures.