No, the model isn’t “thinking”, no model in use today has anything resembling an internal cognitive process. It is making a prediction. A covid test is predicting whether you have the Covid-19 virus inside you or not. If its prediction contradicts your biological state, it is wrong. If an object recognition algorithm does not predict there being a firetruck, how is that not being wrong in the same way?