Interview with Lorraine Daston: “The rules book began with an everyday observation of the dazzling variety and ubiquity of rules. Every culture has rules, but they’re all different.
I eventually settled on three major meanings of rules: rules as laws, rules as algorithms, and finally, rules as models. The latter meaning was predominant in the Western tradition until the end of the 18th century, and I set out to trace what happened to rules as models, but also the rise of algorithmic rules. It’s hard to imagine now, but the word algorithm didn’t even have an entry in the most comprehensive mathematical encyclopedias of the late 19th century.
To get at these changes over time, I cast my nets very wide. I looked at cookbooks, I looked at the rules of warfare. I looked at rules of games. I looked at rules of monastic orders and traffic regulations, sumptuary regulations, spelling rules, and of course algorithms for how to calculate. And if there’s one take-home message from the book, it is a distinction between thick and thin rules.
Thick rules are rules that come upholstered with all manner of qualifications, examples, caveats, and exceptions. They are rules that are braced to confront a world in which recalcitrant particulars refuse to conform to universals—as opposed to thin rules, of which algorithms are perhaps the best prototype: thin rules are formulated without attention to circumstances. Thin rules brook no quarter, they offer no sense of a variable world. Many bureaucratic rules, especially bureaucratic rules in their Kafkaesque exaggeration, also fit this description.
The arc of the book is not to describe how thick rules became thin rules (because we still have thick and thin rules around us all the time), but rather to determine the point at which thick rules become necessary—when you must anticipate high variability and therefore must tweak your rule to fit circumstances—as opposed to the stable, predictable settings in which we turn to thin rules.
In some historically exceptional cases, thin rules can actually get a job done because the context can be standardized and stabilized…(More)”.