Future.tex

In this paper we proposed a framework in which code maintainability data and comprehension effort data are combined to support continuous updates of interest payment estimates, which in turn supports real-time prioritization of debt removal efforts. The primary contribution of our proposed framework is the integration of developer activity data with static code metrics and the concomitant improvements in understanding of developer comprehension effort and in the accuracy with which interest payments can be estimated. An initial investigation of data that we collected from ABB developers demonstrates the feasibility of the framework and provides examples of how the developer activity data work in concert with structural code metrics to reveal new information about developer comprehension effort.

Our next step will be to conduct a large-scale statistical analysis of comprehension and structural metrics to better understand the correlations and levels of technical debt that drive increased comprehension effort.  With this analysis we will determine how to calculate interest payments for technical debt items.  This will include resolving what portion of comprehension effort is for technical debt and the portion that is inherent in comprehending the ideal code structure.  

We plan to further develop the framework and to use it to answer a number of questions about the relationships between developer comprehension effort and technical debt.  For example, we plan to include other comprehension metrics such as the number of edits to a class that will allow evaluation of the Shotgun Surgery smell where multiple classes are modified for a change.  To improve the accuracy of the comprehension data, we plan to detect the central class based on edit actions as well as navigation and search actions during a session.  We also plan to conduct an observational study of developers to validate the estimates  of interest payments during maintenance activities.