Comparing MDD and CcD in the Bug Localization Context: An Empirical Evaluation in Video Games

The development of video games usually involves two main methods: Code-centric Development (CcD) and Model-Driven Development (MDD). CcD uses code languages that provide more control but it requires more effort in order to deal with implementation details. MDD raises the abstraction level (avoiding implementation details) by means of software models that are transformed into code or interpreted at run-time. Raising the abstraction level favors the participation of non-technical roles such as level designers or artists that are essential for video game development. However, bug localization, which is crucial for identifying faults, is less explored in MDD despite its advantages. This work examines how MDD and CcD impact bug localization by using a commercial video game that has been released on PlayStation 4 and Steam. We compare bug localization in terms of performance, productivity, and user satisfaction. The results showed that bug localization in MDD led to higher satisfaction among subjects. However, the differences in performance or productivity depended on experience and favored CcD for professionals. Our findings suggest that bug localization practices performed suboptimally in models, indicating a knowledge gap in addressing bugs within MDD environments. With the rising popularity of MDD in video games, there is a need to explore alternative forms of bug localization for MDD.