Machine learning could fuel a reproducibility crisis in science

&quot;The high accuracy of predictions in tests often fools people into thinking the models are picking up on the &#x27;true structure of the problem&#x27; in a human-like way...&quot;<p>A form of confirmation bias?