Historian Edward Tenner argues for a new approach to building algorithms that allows greater information diversity and leaves room for just plain luck