How scientists fool themselves – and how they can stop