Hofstader [PDF] is paraphrasing Lewis Namier, who wrote in Personalities and Powers (1955):

A dilettante is one who takes himself more seriously than his work; and doctrinaires enamoured of their theories or ingenious ideas are dilettanti in public affairs. On the contrary, the historical approach is intellectually humble; the aim is to comprehend situations, to study trends, to discover how things work: and the crowning attainment of historical study is a historical sense - an intuitive understanding of how things do not happen (how they did happen is a matter of specific knowledge.)

Namier is essentially arguing against the a theory-driven or result-driven approach to history. Instead of looking for root causes for what happened, or for data that fits the model being applied, the historian should look for the hard data - to see the facts, and develop their understanding of what happened from those facts, rather than going in with some preconception of what must have happened. Hofstader in paraphrasing this particular bit is talking about how the paranoid fail this essential criteria: those who go looking for conspiracies tend to find them, not because they exist, but because in a large enough data set there are bound to be parallels and coincidences, and if you seize on those to exclusion of all else, you are fitting facts to match your theories, not forming theories that match the facts.