Although the computerized models offered the predictions
mostly similar to that of human recognition,
the weights they assigned to the lower signified factors
did not exactly coincide. These discrepancies, on
one hand, strengthen the fact that artificial learning
models could emulate human decision making in this
context with varying degrees of success. On the other
hand, they also high-light the fact that the weights
specified by human experts are subjective and can
only be applied in area bounded by their familiarity.
They are also less reliable as they get lower, thanks
to variability due to disagreement among individuals.