On a nearby screen, the supercomputer tells you that Box A contains either a million dollars or nothing at all. It allows you to choose between only two options: choose Box A or take both boxes. 

Normally, the decision would be simple, as taking both boxes appears to be the optimal option. However, things aren’t that simple. Prior to entering the room, the supercomputer has already predicted your choice. If it thinks you will only take Box A, it will contain a million dollars. Meanwhile, if it thinks you will take both boxes, it contains nothing. 

The supercomputer has already done this experiment thousands of times and is unerringly accurate in predicting this choice. 

This thought experiment is called Newcomb’s Paradox. In a survey done by The Guardian in 2016 on over 31,000 people, 46.5% chose to take both boxes, while 53.5% chose only Box A. But why? 

At first glance, the choice seems obvious. Choose Box A and you have the chance to walk away with a million dollars. However, the supercomputer has already made its decision and cannot change the past. Thus, there is either money in Box A or there is not. If Box A contains a million dollars, taking both boxes would earn you an extra thousand. This makes taking both boxes seem to be the most profitable choice. 

Two different principles of decision theory clash with each other here: expected utility and strategic dominance. The expected utility principle weighs the outcome’s value (how much money you get) and its probability to determine the best move. Meanwhile, strategic dominance dictates the best choice from all your options (most profitable), regardless of another party’s actions (the supercomputer’s prediction). 

Someone who chooses Box A reasons that the predictor is almost never wrong, so their choice is a reliable sign of what was predicted. By choosing only Box A, they make it overwhelmingly likely that Box A was filled with a million dollars, since that is exactly the choice the predictor would have foreseen. Thus, it becomes a decision between a guaranteed million dollars against the possibility of only getting a thousand dollars. 

Whereas from the other perspective, the boxes are filled already and are immutable. Regardless of whether there truly is a million dollars within Box A, its contents have already been predetermined before entering the room and no decisions within the room can change it. Thus, there is no rational reason not to take both boxes and gain a thousand extra dollars. 

Newcomb’s Paradox continues to fascinate many, including philosophers and mathematicians. Neither choice feels completely right or wrong. Choosing only Box A prioritises outcome, prediction, and statistical success. Choosing both boxes prioritises strict logical reasoning. 

Newcomb’s Paradox is increasingly relevant in a world driven by artificial intelligence and data modelling systems, predicting our choices and behaviour. By challenging the concepts of prediction and human reasoning, this paradox makes us wonder what it truly means to make a rational decision. If we end up in a world where conforming to the whims of artificial intelligence systems will reward us, which choice will we make? Box A, or both boxes?