The asymmetry of attention

There is a quiet law that governs how organizations see themselves, and it is simpler than the org charts and the dashboards make it look. An institution attends to what it measures and records, and it ignores what it does not. This is not a moral failing of any particular institution; it is closer to a fact of perception. What is written down can be returned to, questioned, counted, compared, improved. What is never written down has no handle by which attention can grasp it. It is not that the institution decides to ignore the unrecorded. It is that the unrecorded never enters the field where deciding happens at all.

For most of institutional history this asymmetry was tolerable because the consequential decisions tended to leave traces by their nature. A decision of weight was made by people, in a place, after some deliberation, and the deliberation left residue — a memo, a meeting, a signature, a file. The trace was a byproduct of the process, not a deliberate act. You did not have to choose to record a serious decision; making it seriously generated the record almost automatically. The asymmetry of attention was real, but it bit hardest on the trivial, and we could live with the trivial being invisible.

What automation broke

Automation severs the link between consequence and trace. An automated decision can be enormously consequential — it can deny, approve, flag, rank, or route a person's life in some material way — and leave no residue whatsoever, because nothing about making it required a residue to be made. The deliberation, such as it was, happened inside a process that produces an output and discards everything else as a matter of efficiency. The decision that would once have generated a memo by the sheer fact of being taken seriously now generates a row in a table that is overwritten on the next pass. Consequence and trace have come apart, and the asymmetry of attention, which used to spare only the trivial, now swallows the grave.

The result is a class of decision that is consequential and invisible at the same time, and the combination is more dangerous than either property alone. A consequential decision that leaves a record can go wrong, but the institution can find out — it can review, question, and correct. A trivial decision that leaves no record can go wrong harmlessly. But a consequential decision that leaves no record can go wrong, fall below the threshold of institutional attention, and stay wrong indefinitely, because there is nothing for attention to land on. No one is ignoring it. It is simply not there to be noticed. The institution cannot attend to a decision it has no evidence ever happened.

What is recorded can be reviewed, questioned, improved. What is not recorded cannot be ignored, because it was never there to ignore. The unrecorded decision is invisible, and invisible is worse than unaccountable.

Recording is choosing what can be seen

This reframes what it means to decide what to record. We tend to treat the recording decision as a downstream, administrative matter — once the real decision is made, the question of what to log is housekeeping. But the asymmetry of attention inverts that order of importance. The choice of what to record is a choice about what the institution will ever be able to see. It sets the boundary of the institution's own perceptual field, the line between the decisions it can later examine and the decisions that have already passed beyond its reach. To choose not to record a class of decision is to choose that the institution will be blind to that class forever, regardless of how consequential those decisions turn out to be.

And the blindness is not symmetric in its costs, which is the final turn. An unrecorded decision is not merely unaccountable in the sense that you cannot assign blame for it — that would be bad enough. It is unaccountable in the deeper sense that you cannot even know there is anything to assign blame for. Accountability presupposes visibility; you cannot hold to account what you cannot see. So the unrecorded decision sits in a category below the unaccountable: it is the decision that has been removed from the institution's capacity to know about itself. A wrong you can see and fail to fix is a scandal. A wrong you cannot see at all is a structural condition, and it persists not because anyone permits it but because nothing in the system is capable of registering that it is there.

The remedy is not exhortation to pay more attention, because attention cannot be pointed at what leaves no mark. The remedy is to restore the trace — to make the consequential decision generate a record by construction, the way the deliberated decision once did, so that the thing exists in a form attention can reach. A Decision Receipt is, among other things, an answer to the asymmetry of attention: it drags the automated decision back above the threshold of institutional perception, where it can be reviewed, questioned, and improved like any decision that left a paper trail in the older world. Without it, the institution does not have a smaller field of view. It has a field of view with holes in exactly the places where the most consequential things are happening, and no way to know the holes are there.

— Dispatches · Summit Cognitive