could it be the models haven't changed, but because of their probabilistic nature, they trigger all sorts of human biases that make us draw incorrect conclusion

could it be the models haven't changed, but because of their probabilistic nature, they trigger all sorts of human biases that make us draw incorrect conclusions about their behavior?