Noetic Debt: How Much Discovery Should We Trade for Understanding?

Scientific discovery has historically required human understanding. Understanding existing science was how scientists produced the next result. Autonomous AI threatens to sever that connection, turning human understanding from an input to discovery into an optional output, something science can produce or decline to produce at a cost. I call the accumulation of established results that no one grasps ‘noetic debt’ and use examples from mathematics and empirical science to illustrate. Managing this debt in the age of autonomous science is an allocation problem because the machine capacity spent making results intelligible is capacity not spent finding new ones, except where explaining also improves the search. The cost of spending too little can compound: results left unexplained can make the discoveries built on them harder to understand. Once discovery no longer requires understanding, we must choose to cultivate it. The question is therefore not only how much science we can produce, but how much of it we should seek to understand. That question cannot be settled by scientific results alone. It requires deciding what value human understanding has. Behind it sits a prior question: can it still be scientific progress if we understand less science?