On Comprehension of Genetic Programming Solutions: A Controlled Experiment on Semantic Inference

Applied to the problem of automatic program generation, Genetic Programming often produces code bloat, or unexpected solutions that are, according common belief, difficult comprehend. To study comprehensibility produced by Programming, attribute grammars obtained Programming-based semantic inference were compared manually written ones. According established procedure, research was carried out as a controlled classroom experiment involved two groups students from universities, and consisted background questionnaire, tests feedback questionnaire after each test. The tasks included in required identification various properties attributes grammars, correct grammar list choices, correcting rule an grammar. It automatically generated field inference, this are indeed significantly harder comprehend than This finding holds, regardless whether comprehension correctness, i.e., how many correctly comprehended, efficiency is considered, quickly comprehended.