Human experts in many scientific fields routinely work with data that are heterogeneous, noisy, and/or uncertain, as well as with heuristics that are unproven and with possible conclusions that are contradictory. We present a deployed software system for cosmogenic isotope dating, a domain that is fraught with these difficult issues. This system, which is called ACE (Age Calculation Engine), takes as inputs the nuclide densities in a set of rock samples from a landform. It answers the scientific question, “What geological processes could have produced this distribution of nuclide concentrations, and over what time scales?” ACE employs an encoded knowledge base of the possible processes that may have acted on that landform in the past, complete with the mathematics of how those processes can affect samples, and it uses a workflow system to encode the computations associated with this scientific analysis. Flexibility and extensibility were critical issues in ACE’s design to allow its scientist-users to modify and extend it after its developers were no longer involved. The success of this is evident; the system remains in active use to this day, several years after the development cycle ended, without a single request for help from the geoscientists to the computer science side of the team. The ACE project website has received over 17,000 hits since 2008, including 2500 over the last twelve months. The software (~20,000 lines of Python code) has been downloaded nearly 600 times as of April 2013, which is a significant number in a research community of a few hundred PI-level scientists.