Model consistency as a heuristic for eventual correctness

Abstract

Inconsistencies between stakeholders’ views pose a severe challenge in the engineering of complex systems. The past decades have seen a vast number of sophisticated inconsistency management techniques being developed. These techniques build on the common idea of “managing consistency instead of removing inconsistency”, as put forward by Finkelstein. While it is clear what and how to do about inconsistencies, it is less clear why inconsistency is particularly useful. After all, it is the correctness of the system that should matter, as correctness is the end-user-facing quality of the product. In this paper, we analyze this question by investigating the relationship between (in)consistency and (in)correctness. We formally prove that, contrary to intuition, consistency does not imply correctness. However, consistency is still a good heuristic for eventual correctness. We elaborate on the consequences of this assertion and provide pointers as to how to make use of it in the next generation of inconsistency management techniques.