What Human Judgement Is: A Hypothesis

Abstract

This paper states a hypothesis about what human judgement is, how it works, and what it costs to engage it or to decline it. Human judgement is the faculty for rendering a verdict where the data that would settle the matter cannot be had in time, or at all. It is a relationship to data: where facts exist or can be gathered, theygovern, and the faculty operates only in the residue. Automation exists because data reached scale, so it is confined by construction to where data speaks; the residue is not a frontier machines have not reached yet, it is the place they cannot reach. Subtract what the animals share and what the machines can do, and what survives both cuts is the no-data verdict, arguably the one experience that is human and only human. In the residue, the operation is the largest survivable bet that returns its verdict in time, made actionable through a disposable model; the value usually reveals itself in losing the bet, which locates a boundary no data could supply. The operation carries one requirement, borne cost, and two gates, discomfort on the way in and resolution on the way out. Two learnings run off it: an execution learning that is largely conscious, domain-specific, and poorly transferable, and a learning of the judging itself that is mostly unconscious, gated by resolution, plausibly consolidated in sleep, and transferable. The faculty is, as far as we know, the most reliable builder of durable self-respect: first-hand, self-witnessed evidence that one can handle what comes. Comfort structures relieve the discomfort of the no-data region; some fund the next attempt and some fund the next avoidance. If the hypothesis holds, reframing the relationship to comfort is a path to reducing avoidable human suffering.