Ten reservoir models fit the same production history. Which one is true, and does the question make sense?
Its coordinates: 6.8 Philosophy of Science · 6.8.1 Scientific Explanation · 6.8.2 Scientific Realism +1 more
In my work we build computer models of oil reservoirs deep underground, which nobody will ever see. We calibrate them against what we can measure: decades of production rates, pressures in the wells, the water that starts to come up with the oil. We call this history matching.
The awkward fact every reservoir engineer learns early: many very different models match the history equally well. One has a sealing fault in the north; another has no fault but a band of low-permeability rock. Both reproduce thirty years of data. They predict quite different futures.
In practice we keep an ensemble of models and report a range. But I keep wondering what we are doing philosophically.
- Is this what philosophers call the underdetermination of theory by evidence? I've read that Duhem said a hypothesis is never tested alone, only together with auxiliary assumptions, and that Quine made it much stronger.
- If the evidence can't decide, is there still a fact about whether the fault is there? Of course there is, I want to say; the rock is what it is. But then what are my ten models: ten guesses at one truth, or ten tools, none of which is "true"?
- And does it matter, if the ensemble is what we actually use to decide where to drill?