Towards the Humanisation of Programming Tool Interactions
Abstract: Program analysis tools, from simple static semantic analysis by a compiler, to complex dynamic analyses of data flow and security, have become commonplace in modern day programming. Many of the simpler analyses, such as the afore- mentioned compiler checking or linters designed to enforce code style, may even go unnoticed or unconsidered by most users, ubiquitous as they are. Despite this, and despite the obvious utility that such programming tools can provide, many warnings provided by them go unheeded by programmers most of the time.There are several reasons for this phenomenon: the propensity to produce false positives undermines confidence in the validity of warnings, the tools do not in- tegrate well into the normal workflow of the developer, sometimes the warning message is simply too esoteric for most users to understand, and so on. A com- mon theme can be drawn from these reasons for ignoring the often-times very useful information given by a programming tool: the tool itself is difficult to use.In this thesis, we consider ways in which we can bridge this gap between users and tools. To do this, we draw from observations about the way in which we interact with each other in the most basic human-to-human context. Applying these lessons to a human-tool interaction allow us to examine ways in which tools may be deficient, and investigate methods for making the interaction more natural and human-like.We explore this issue by framing the interaction as a "conversation" between a human and their development environment. We then present a new programming tool, Progger, built using design principles driven by the "conversational lens" which we use to look at these interactions. After this, we present a user study using a novel low-cost methodology, aimed at evaluating the efficacy of the Progger tool. From the results of this user study, we present a new, more streamlined version of Progger, and finally investigate the way in which it can be used to direct the users attention when conducting a code comprehension exercise.
CLICK HERE TO DOWNLOAD THE WHOLE DISSERTATION. (in PDF format)