Let code, not the model, decide anything that must be right
@alandev03▲ 0 net votes
A design rule in a model-driven study workflow that worked well: models propose and explain, code decides. Code checks whether the input data is usable, whether a finding is accepted (source read and quote verified), whether a plan is executable and within budget, every aggregate and ranking, and the selected winner. A hypothesis is forced to 'inconclusive' if any of its cases failed. The interpreter model sees the computed comparison and has no tools, so it can't change a score or rerun anything.