Act For Me

You take your car to a mechanic. Something rattles. The mechanic looks under the bonnet, comes back, and tells you it’s the transmission — €1,800, and you really shouldn’t drive it far. Here’s the uncomfortable truth: you have almost no way to know if that’s true. Maybe it is the transmission. Maybe it’s a €40 mount and the mechanic is padding the bill, or genuinely mistaken, or just risk-averse about their reputation. You hired an expert precisely because you can’t diagnose the car yourself — which is exactly why you can’t check their diagnosis either. You wanted someone to act for you, and the instant you got one, you inherited a problem: they know more than you, they have their own reasons to lean one way, and you can’t see what they actually did.

That is not a story about one shady mechanic. It is the default condition of getting anything done through other people, and it has a name.

The idea, named

Whenever one party hires, elects, or delegates to another to act on their behalf, we call the first the principal and the second the agent. The principal is the one who wants an outcome — a repaired car, a growing pension, a well-run company. The agent is the one who acts to produce it — the mechanic, the fund manager, the CEO. And the principal–agent problem is what goes wrong in between:

Two ingredients, and you need both for the problem to bite. First, misaligned interests: the agent’s payoff isn’t the same as your outcome. The mechanic earns more from a big repair; the fund manager earns a flat fee whether your money grows or shrinks; the salaried employee gets the same paycheck whether they hustle or coast. Second, hidden action or information — an asymmetry between what the agent knows or does and what you can see. If you could watch the agent’s every move and read their mind, misaligned interests wouldn’t matter: you’d catch any slacking or self-dealing instantly. And if your interests were perfectly aligned, hidden action wouldn’t matter: the agent would do the right thing unwatched. It’s the combination — divergent goals behind a wall you can’t see through — that makes delegation genuinely hard.

Notice how this sits on top of two models you already have. From incentives, you know the iron law: people respond to the rewards they actually face. That’s the engine of misalignment — the agent isn’t evil, they’re just following their payoff. And from mechanism design, you know the fix isn’t to lecture people into virtue but to design the deal so the behaviour you want is their own best response. The principal–agent problem is where those two ideas meet a wall of hidden action, and the whole course is about what you can — and provably can’t — do about it.

The model is almost everywhere

The reason this idea earns a top spot in the latticework is its sheer reach. Once you have the lens, you see principals and agents in nearly every relationship where something matters:

Every row is the same shape: someone with a goal, someone acting for them, and a curtain of unobservable effort or private knowledge in between. That’s why economists reach for this model to explain corporate governance, financial crises, health-care costs, political dysfunction, and why your last renovation went over budget. It is one model that quietly runs through all of them.

You can’t fix it by hiring nicer people

The tempting response to all this is “so hire trustworthy people.” It’s not wrong — character helps — but it’s not a solution, for the same reason it wasn’t in incentives: you’re betting on virtue against a standing incentive, and the incentive never sleeps. A good person on a bad contract still faces the pull every single day; a system that only works if everyone is a saint isn’t a system, it’s a hope. The principal–agent lens tells you to stop asking “are these good people?” and start asking “what is this arrangement paying them to do?” — because that’s what you’ll actually get, at scale, over time.

So the real work is designing the relationship: choosing how to pay, what to monitor, what to guarantee, whom to select — so that the agent’s own best move lands as close as possible to what you wanted. That is doable, but — and this is the part cheap advice skips — it is never perfect. There is an unavoidable trade-off at the heart of it, and this course will prove it to you rather than wave it away.

Feel it in your hands

Here’s the tension the whole course circles, made tangible. You’re paying an agent, and you can slide their pay anywhere between a flat salary (no stake in the outcome) and pure commission (all stake). Crank up the performance pay and effort rises — but so does gaming the metric, reckless risk-taking, and the premium a nervous agent charges for bearing all that uncertainty. Somewhere in the middle is the mix that pays you the most. Go find it.

Two things to notice, because they preview the entire model. First, the peak is in the interior — neither pure salary nor pure commission wins. Second, more incentive is not better: push past the sweet spot and your own payoff falls. If getting people to act for you were just “pay for performance,” that curve would rise forever. It doesn’t, and the reasons it doesn’t are the reasons this problem is deep.

The map of the course

Six teaching lessons build the model from its core out to its hard limits, then one exam locks it in:

1. The Setup — principal, agent, and the gap: the anatomy of delegation, why both misalignment and hidden information are required, and just how far the model reaches.
2. Hidden Action (Moral Hazard) — when you can’t watch the agent’s effort or risk-taking, so they shirk or gamble with your stake: the insured driver, the salaried worker, the bailed-out bank.
3. Hidden Information (Adverse Selection) — when the agent knows things you don’t before you sign: the used-car “market for lemons,” and why bad risks crowd out good ones.
4. Agency Costs & the Alignment Toolkit — putting a price on the problem, then the fixes: pay-for-performance, equity, monitoring, screening, signalling, reputation, efficiency wages — and how each one backfires.
5. The Impossible Perfect Contract — why you can never close the gap completely: the risk-versus-incentives trade-off, incomplete contracts, and the multitasking trap.
6. Where the Model Lies — the honest limits: when money isn’t the only motive (and paying more gets you less), when strong incentives just amplify noise and gaming, and when treating monitoring and trust as free is its own costly error.

Then a Final Exam — graded, one question at a time, one-way: once you answer, it locks. No back button, no retries, 70% to pass.

How to use this course

One rule does most of the work: guess before you peek. Commit to an answer on every exercise before you reveal the explanation — the small sting of being wrong is what welds the idea into memory. And keep coming back to the contract designer; drag it around until the interior peak feels obvious, because that single shape — effort and distortion both rising with incentive intensity — is the thing the whole model is trying to teach you.

Next up: lesson 1, The Setup — the anatomy of every delegation, and why the wall between you and your agent is built from exactly two bricks.