Mirror, Mirror on the Wall...
The story of Colbert’s executive dashboard, the golden notebooks he created for Louis XIV, is fascinating because it highlights how early nation states addressed their data management issues. This is truly “old-school” data science at its best, before the age of computers, the internet, and smartphones. Everything was calculated by hand on a sheet of paper or vellum or in one’s head.
Initially, Louis XIV – France’s “Sun King” – valued Colbert’s hard work and the information system he had created. In his book, The Reckoning: Financial Accountability and the Rise and Fall of Nations, Jacob Soll presents the complete story of Colbert’s golden notebooks. Initially, Louis derived great satisfaction from his newfound financial literacy. He wrote to his mother, “I have already begun to taste the pleasures to be found in working on finances myself … no one should doubt that I will continue”1.
Soll writes that starting in 1661, Louis received bi-annual financial updates. At those meetings, Colbert and the King reviewed his expenditures, assets, and the nation’s revenue. The discipline paid off. France became legible and quickly emerged as Europe’s most powerful state. And while he was on the go, the king’s golden notebook (his portable executive dashboard) was in his coat pocket within easy reach.
The system worked well until Colbert died in 1683. Shortly thereafter, Louis discontinued the account books. Soll expresses surprise at how short-lived the entire project was. Why stop when the data management system worked so well? The problem was Louis himself. More precisely, it was his spendthrift ways. Versailles was a money pit. The frequent wars and military campaigns cost an arm and a leg. And finally, the trifecta of wine, beautiful women, and song ate into the king’s cash flow like a ravenous termite colony. The result was red ink everywhere.
While Colbert’s information system made France legible to its King, it also made its King legible to himself. And therein lay the problem. The accounting system was a mirror, reflecting back to Louis all the poor decisions he was making. It hurt his eyes and stung his pride. Rather than face himself, Louis XIV decided to disconnect the monarchy from reality. About a hundred years later, the fruits of that decision were plain to see. With famine ravaging the land, Marie Antoinette was reportedly asked about the death of so many peasants. Her response? “Let them eat cake!” They ate the monarchy instead. So, one might argue that the French Revolution, in a very real sense, was a data management failure. That is, those in charge largely ignored the data that could have saved them.
Not much has changed since the 17th century. Today, our AI models also act as mirrors, reflecting our prejudices and values back to us. We’re surprised and often offended, for example, if a model responds as if all doctors are men and nurses are women. Yet it’s not the model’s fault. Large language models (LLMs) are trained on text humans create. As such, all they do is make us legible to ourselves. This relationship is no different from the one Louis XIV had with Colbert’s information system. The question, then and now, is: “How will we respond?”
Jacob Soll, The Reckoning: Financial Accountability and the Rise and Fall of Nations (New York, Basic Books, 2014) chapter 6.