“A survey of 400 algorithms presented in papers at two top AI conferences in the past few years… found that only 6% of the presenters shared the algorithm’s code” Missing data hinder replication of artificial intelligence studies.