Knowledge, noise, and curve-fitting: A methodological argument for justified true belief?

Could considerations of simplicity ultimately lead us to prefer the theory that knowledge is justified true belief, over rival theories that accommodate the standard view that Gettier cases are justified true belief but not K? I consider one source of simplicity considerations: when one engages in a modeling project relying on noisy data, then the more noise that one expects, the more one should prefer to select simpler curves. And the general noisiness of our epistemic verdicts about cases puts such considerations into play.