How do you make information more valuable? Not by getting better information — by having more decisions to make with it.
The paper establishes a clean characterization: one decision problem yields a higher value of information than another, uniformly across all information structures, if and only if it is obtained by adding an independent parallel decision problem. Not refining the existing choice. Not getting a more precise signal. Adding a second, separate decision that runs alongside the first.
This is counterintuitive because the standard framing treats information value as a property of the signal — better data means better decisions. But the result says the value lives equally in the decision structure. The same signal is worth more when you have two things to decide than when you have one, regardless of what the signal says.
The mechanism is that additional independent decisions create more “surface area” for information to be useful. Each decision is a dimension along which a signal can shift the optimal action. A signal that's irrelevant to one decision might be decisive for another. The more decisions you face, the more ways any given piece of information can improve your total outcome.
The practical implication reverses the usual optimization direction. If you're trying to increase the return on your information infrastructure, the answer might not be better sensors or more data. It might be identifying additional decisions where the existing data is decision-relevant. The data you already have may be undervalued — not because it's low quality, but because you're not making enough decisions with it.