Crowd-Based Programming for Reactive Systems

Abstract

End-user applications aimed at the public in general (mobile and web applications, games, etc.) are usually developed with feedback from only a tiny fraction of the millions of intended users, and are thus built under significant uncertainty. The developer cannot really tell a priori which features the users will like, which they will dislike, and which ones will help create the desired outcome, such as high usage or increased revenue. In these cases, providing adaptive capabilities can be the key factor in the application's success. Existing self-adaptive techniques can provide some of the needed capabilities, but they too must be planned, and leave the developers, and much of the development process, "out of the loop". We propose a development environment that allows the wisdom of the crowd to influence the very structure and flow of the program being created, by voting upon behavioral choices as they are observed in early versions of the working program. The approach still allows the developers to retain known desired behaviors, and to enforce constraints on crowd-driven changes. The developers can also react to ongoing crowd-programmed feedback throughout the entire lifetime of the application.