Predict the output. A wrong first answer gets feedback on the misconception behind it and a hint, then one more try.