Decision Theory
The Box Problem That Broke Decision Theory
A thought experiment invented in the 1960s by a physicist at RAND Corporation has never been solved — and the way you answer it reveals something unsettling about how you think you make choices.
The Idea
Imagine two boxes in front of you. Box A is transparent and contains a small amount of cash. Box B is opaque. A supremely accurate predictor — call it an oracle, an AI, a god, it doesn't matter — has already studied you in depth and made a prediction about what you'll do. If it predicted you'd take only Box B, it placed a large fortune inside it. If it predicted you'd take both boxes, it left Box B empty. The predictor is almost never wrong. What do you do? Two perfectly reasonable positions exist, and they flatly contradict each other. The 'evidential' camp says: take only Box B. The evidence is overwhelming that one-boxers walk away rich. Your choice and the box's contents are correlated, even if the prediction is already locked in. Choose the action that is associated with the better outcome. The 'causal' camp says: take both boxes. The prediction is already made — the money is either in Box B or it isn't, and your decision right now cannot change that. Taking both boxes dominates: you get whatever is in Box B plus the guaranteed cash. Why leave anything on the table? Here's what makes this extraordinary: both arguments are logically coherent. Newcomb's problem, named after physicist William Newcomb and popularised by philosopher Robert Nozick, isn't a trick or a paradox with a hidden solution. It is a genuine rupture in decision theory — a place where rationality seems to argue with itself.
In the World
When Robert Nozick published his analysis of Newcomb's problem in 1969, he admitted he found it deeply troubling — and he'd been thinking about it for years. He wrote, famously: 'To almost everyone, it is perfectly clear and obvious what should be done. The difficulty is that these people disagree.' Philosopher David Lewis, one of the most rigorous minds of the twentieth century, planted himself firmly in the two-box camp. His argument: causality is what matters in decisions, not mere correlation. The predictor's forecast, however accurate, cannot retroactively reach into Box B and remove the money after you choose. You act on the world as it is, not on what your action signals about you. But the philosopher and cognitive scientist David Chalmers, among others, has pointed out that one-boxing works spectacularly in practice — simulated versions of the problem, where a computer plays the role of predictor, reliably reward one-boxers. If you're the kind of person who always takes both boxes, the predictor knew that before you sat down. This is where the problem bleeds beyond philosophy into something personal. Your decision in Newcomb's problem doesn't just reflect a theory of causality — it reflects a theory of yourself. Are you the kind of agent whose choices can be predicted? And if so, does that make your 'free' choice less free, or just more consistent than you'd like to admit?
Why It Matters
Newcomb's problem isn't waiting for you in a laboratory. But its underlying tension — between acting on what you can causally control versus acting on what your choice signals about who you are — shows up constantly. Think about integrity. When no one is watching and there is no practical consequence, why behave well? A pure causalist might say there's no point — the outcome isn't affected. But most people sense that the kind of person you are shapes the kind of choices you make, which shapes the kind of life you get. Your decisions aren't isolated events; they are data points in a pattern that you, and others, are always reading. Mindfulness traditions have long insisted on something similar: that the quality of attention you bring to a choice matters beyond its immediate outcome. You are not just choosing an action; you are rehearsing a way of being. Newcomb's problem makes this philosophically precise. It asks whether you are an agent who acts on isolated causal chains, or one who recognises that you are, in some sense, already the kind of person who will make a particular choice — and that this matters.
A Question to Ponder
When you make a decision today, are you asking 'what action gets me the best outcome from here?' or 'what does the choice I'm about to make reveal about the kind of person I actually am?'
Get a new one of these every morning.
Start learning with Thinkable