The mind is drawn to what's specific.
When a prediction needs to be made, the instinct is to look closely at the situation in front of you — the details, the context, the particular shape of this case. That's where reasoning tends to begin, and for most people, it's also where it ends.
The problem with that approach isn't that the details are wrong. It's that they're doing too much work. A prediction built entirely from the specifics of one situation has no stable foundation underneath it. It feels confident — because the details feel real and concrete — but it isn't grounded in anything beyond the story being told. There's no reference to what generally happens in situations like this one. No sense of how similar cases have played out across time. Just this instance, treated as if it holds the complete answer.
That's how forecasts go quietly off track. Not through obvious errors, but through a kind of narrowing — where the specific crowds out the general, and the prediction ends up shaped more by the texture of one case than by the actual distribution of outcomes that history supports.
When the encouraging details dominate, the estimate drifts too high. When the discouraging ones do, it drifts too low. In either direction, the result is the same: a forecast anchored to a single story instead of the wider range of what tends to happen.
The core issue isn't that specific reasoning occurs. It's that it often occurs alone — without anything more stable underneath it to keep the prediction from drifting wherever the details happen to lead.
The reversal is simple, but it changes how the whole process works.
Instead of starting with the specific situation and building a prediction from there, you start somewhere else entirely — with the general pattern. What do outcomes typically look like in situations of this kind? Before any unique details enter the picture, what does the broader statistical record suggest?
That becomes the first reference point. Not the final answer, but the starting place. A rough baseline drawn from how things tend to go across a wide range of similar situations, before any story is attached.
Only after that anchor is in place do the specifics come in. And when they do, their role becomes more limited — they're there to adjust the baseline, not replace it. To shift the estimate up or down based on what's genuinely distinct about this particular case, while keeping the broader pattern in view.
This sequence doesn't remove judgment from the process. It gives judgment something stable to work from. The specific and the general both contribute — just in the right order. General first. Specific second.
The mechanism that makes this work is the base rate — the frequency at which something tends to occur across a large number of similar cases.
It isn't drawn from the situation at hand. It's derived from the broader category that situation belongs to. And because it comes from many instances rather than one, it carries a built-in stability that case-specific reasoning alone doesn't have.
When a forecast begins with the base rate, it starts from something grounded. That initial reference point sets a reasonable position — calibrated to what actually tends to happen — before any of the story enters. Then, as the details of the specific case become relevant, they layer in carefully, adjusting the estimate only as far as the evidence actually warrants.
The result is a prediction that reflects both the general pattern and the particular situation — one that isn't pulled too far in either direction by the details alone. Balanced not by accident, but because it was built that way, from the ground up.