The Roman Who Vouched for Everyone
In 63 BCE, Marcus Tullius Cicero wrote a letter of introduction for a young man named Gaius Trebatius Testa, recommending him to Julius Caesar for a position in Gaul. The letter was effusive in its praise, describing Trebatius as "exceptionally talented" and "worthy of your friendship." There was just one problem: Cicero had met the man exactly once, at a dinner party three weeks earlier.
Photo: Marcus Tullius Cicero, via www.stellabooks.com
This wasn't unusual. It was standard operating procedure.
The Roman system of commendatio — formal letters of recommendation — functioned as the empire's primary networking mechanism. Wealthy Romans routinely wrote glowing endorsements for people they barely knew, and everyone understood the game. The letters weren't really about the candidates. They were about the recommenders signaling their own importance, maintaining their social networks, and banking favors for future use.
Sound familiar?
The Economics of Social Capital
What Cicero understood intuitively, modern behavioral economics has quantified: recommendation systems aren't broken when they prioritize the recommender's interests over accuracy. They're working exactly as human social psychology designed them to work.
Consider the medieval guild system. Master craftsmen were required to vouch for their apprentices before they could practice independently. In theory, this protected consumers from incompetent workers. In practice, it created a closed network where masters recommended apprentices not based on skill, but on family connections, political loyalty, and personal debt.
The Worshipful Company of Goldsmiths in London kept detailed records from the 14th century onward. Analysis of these documents reveals that apprentices with family connections to existing guild members were endorsed at rates 300% higher than equally skilled outsiders. The system wasn't failing to identify talent — it was successfully identifying tribal membership.
The American Innovation
When European colonists arrived in America, they brought the recommendation letter with them, but the frontier environment created a fascinating adaptation. Without established social hierarchies, recommendations became less about confirming existing status and more about creating new networks from scratch.
Benjamin Franklin, perhaps America's greatest networker, perfected what we might now call "strategic vouching." His letters of introduction were masterpieces of calculated ambiguity. He would recommend people he'd never met to colleagues he barely knew, using language precise enough to sound meaningful but vague enough to avoid specific claims.
Photo: Benjamin Franklin, via ilarge.lisimg.com
"Mr. Johnson comes highly recommended by mutual acquaintances for his industrious character and sound judgment," Franklin might write, when he'd heard of Johnson third-hand and knew nothing about his judgment whatsoever.
Franklin understood that the value of a recommendation lay not in its accuracy but in its social function: creating connections, demonstrating influence, and building the recommender's reputation as someone worth knowing.
The Digital Age Changes Nothing
LinkedIn has digitized and scaled the Roman commendatio system without fundamentally altering its psychology. The platform's recommendation feature allows users to endorse skills they've never witnessed in people they've barely worked with. A 2019 study found that 73% of LinkedIn recommendations contained specific claims about abilities the recommender had no direct knowledge of.
Yet the system persists because all parties understand the implicit contract. Candidates get credentials that look impressive to algorithms and HR departments. Recommenders get to display their networks and professional generosity. Employers get a filtering mechanism that, while imperfect, still provides more signal than a resume alone.
The genius of the system isn't that it accurately identifies talent — it's that it identifies people who are embedded in professional networks, which turns out to be a reasonably good proxy for competence in collaborative work environments.
Why Fake Vouching Works
The durability of recommendation systems across cultures and centuries suggests they serve a deeper psychological function than quality control. They create the illusion of reduced risk in high-stakes decisions.
When a Roman senator hired someone based on Cicero's recommendation, he wasn't just getting a worker — he was getting Cicero's implicit guarantee that if things went wrong, the senator could blame Cicero rather than taking full responsibility for his own judgment.
Modern hiring managers exhibit identical behavior. A 2018 study of Fortune 500 companies found that employees hired through internal recommendations were no more likely to succeed than those hired through other channels, but managers reported feeling "more confident" about recommendation-based hires and were less likely to be criticized if those hires failed.
The recommendation system works because it distributes psychological risk, not because it improves outcomes.
The Honest Signal in Dishonest Systems
Perhaps most remarkably, recommendation systems generate genuine information despite being built on mutual deception. The signal isn't in what the letters say — it's in who writes them and for whom.
When Cicero recommended Trebatius to Caesar, he was revealing important truths: that he had access to Caesar, that he was willing to spend social capital on this particular candidate, and that he believed his relationship with Caesar was strong enough to survive a bad recommendation.
Similarly, when a senior executive writes a LinkedIn recommendation for a junior colleague, she's signaling that this person is worth her public association, that she's confident enough in their basic competence to attach her name to their career, and that she believes the relationship has future value.
These signals are often more reliable than the specific claims in the recommendations themselves.
The Eternal Return
Five thousand years of evidence suggests that humans will always create systems for vouching for each other, and those systems will always be gamed, and everyone will always know they're being gamed, and the systems will persist anyway.
This isn't a bug in human social organization — it's a feature. Recommendation systems don't exist to perfectly identify talent. They exist to create networks, distribute risk, and allow strangers to do business with each other in a world where perfect information is impossible.
The next time you write a LinkedIn recommendation for someone you barely know, remember: you're participating in one of humanity's oldest social technologies, refined over millennia to do exactly what you're doing. You're not breaking the system. You are the system.
