Rules
Lorraine Daston
Why I Recommend This
I was reviewing an AI governance proposal when a researcher flagged an edge case the framework couldn't handle. The authors had tried to specify every scenario in advance. The more complete their rules became, the more brittle they were. Then I read Daston.
She distinguishes thick rules (loaded with caveats, examples, room for judgment) from thin rules (mechanical, context-free, aspiring to cover everything). Her insight is historical: for most of Western history, people understood that rules couldn't anticipate every situation. Discretion was the feature. Somewhere along the way, we forgot.
The Book
Daston traces 2,500 years of Western rule-making through three persistent forms: rules as algorithms (procedures to execute), rules as laws (commands to obey), and rules as models (exemplars to imitate). The last form—paradigms requiring interpretation—dominated until the Enlightenment but has largely been forgotten.
Her central argument: the shift from thick to thin rules reflects a transformation in what we consider rational. Cold War thinkers elevated the algorithm to the ideal form of reasoning. But the ancient problem remains unsolved—no universal can anticipate every case it will encounter.
Passages That Stayed With Me
"Even if rule and particular clearly match, they almost never align perfectly. Tailoring and tweaking will be necessary to smooth over the gap between universal and particular."
Entire specialties—equity, casuistry, case law—grew in this gap.
"An algorithm would then be seen not as the perfection of a rule, but as a pathological case of a rule."
The recipe-algorithm comparison inverts the usual hierarchy.
"By driving the exercise of discretion underground, rules-as-algorithms blow up the bridges that connected universals to particulars."
We eliminated judgment from rule-following, then wondered why our systems became rigid.
"Thick rules create a flexible order out of a disorderly world, but thin rules need an order that already exists in order to function at all."
Thin rules don't impose order—they presuppose it.
Read This If...
- • You design systems (organizations, AI, policies) and want to understand why specification always fails at the edges
- • You're interested in how Cold War thinking reshaped what we consider rational
- • You work in a domain where judgment matters but keeps getting squeezed out by demands for "consistency"
Skip This If...
- • You want prescriptive guidance on when to use thick vs. thin rules
- • You prefer tight arguments to historical panoramas