A door is open or it is closed. A person is employed or they are not. An application has been approved or it has been rejected. These are the categories institutions prefer, and the preference is not irrational. Binary outcomes are tractable.
They can be recorded, communicated, appealed, audited, and assigned. The person who made the decision can be identified. The basis for the decision can be stated. The person affected by the decision knows what the decision was. The system moves forward because the question has been answered and the answer is unambiguous.
Reality does not consistently produce binary situations. A door can be ajar. A person can be working under conditions that do not fit either employment or its absence. An application can address some of the relevant criteria fully, others partially, and others not at all. The binary categories are imposed on situations that do not naturally produce them, and the imposition works well enough that the gap between the category and the situation is not usually examined. The decision is made. The record shows a decision. The system continues.
Fuzzy logic is the formal apparatus for handling the ajar door—the mathematics of gradual truth, of situations that are not fully one thing or fully another, of conditions that are met to a degree. It was developed for precisely the situations that binary logic handles badly, and it works in engineering contexts where the inputs are continuous and the outputs need to track them. The thermostat that does not simply switch heat on or off but modulates it in response to the actual temperature differential. The braking system that does not simply apply full braking or none but adjusts the force to the conditions. The control system that matches its response to the gradient of the input rather than to a threshold that divides the input into two states.
These applications are successful and uncontroversial. The thermostat’s fuzzy logic does not disturb anyone. The question worth examining is why the same logic has not migrated into the institutional domains where it would be at least as applicable—and why the institutions that govern most of human life continue to produce binary outputs for inputs that are demonstrably not binary.
The first part of the answer is about accountability, which is the most important structural feature of any system that makes decisions affecting people. Accountability requires that a decision can be traced back to a decision-maker, that the basis for the decision can be stated, and that the decision can be evaluated against a standard. Binary decisions satisfy these requirements in ways that fuzzy decisions do not.
If a candidate is hired, someone decided to hire them. The decision can be reviewed. The criteria can be stated. The candidate who was not hired can be told why, in terms that refer to the criteria rather than to a score that represents a degree of satisfaction of the criteria. The score is the problem. A candidate who receives a score of 0.6 on a qualification assessment has been assessed as 0.6 qualified. What does that mean for the person who has to make the next decision? Hire or not hire? The score does not answer the question. It describes the degree of fit between the candidate and the criteria. The decision requires a threshold—a point at which 0.6 becomes yes or no. The threshold is the binary that the fuzzy score defers but does not eliminate.
The binary is not avoided by fuzzy logic. It is moved. The decision still has to be made. The question is where in the process the conversion from degree to category occurs, and who is responsible for the conversion, and on what basis. In the direct binary system, the conversion is visible and attributed. In the fuzzy system that eventually produces a binary output, the conversion is somewhere in the process, and the accountability for it is distributed across the algorithm, the threshold-setter, the reviewer, and the decision-maker in ways that are harder to trace and harder to assign.
The second part of the answer is about what institutions require from their own representations of their activity. An institution that presents its decisions as the output of a clear process, applied consistently, producing determinate outcomes, is presenting itself as a machine—reliable, objective, not subject to the variation that human judgment introduces. The presentation is a legitimising strategy. It claims that the institution’s outputs are the product of the rules rather than the people applying them, which makes the outputs less contestable and the institution less vulnerable to the charge that its decisions are arbitrary or biased.
Fuzzy logic would complicate this presentation. A system that acknowledges that its outputs are degrees of truth rather than determinate facts is acknowledging that the process is interpretive—that the conversion from degree to category involves judgment, and that the judgment is not fully determined by the rules. This is accurate. It describes what actually happens in most institutional decision-making, where the rules are applied to situations that do not perfectly fit the categories the rules define, and the fit is assessed through a process that involves human judgment at multiple points. The binary output conceals this. The fuzzy representation would expose it.
The exposure is the problem. An institution whose decision-making is visibly interpretive is an institution that is harder to defend against contestation, because the interpretation is a point of entry for challenge. The person who received the binary rejection can be told the criteria were not met. The person who received the 0.6 score and a rejection can ask why 0.6 was insufficient, why the threshold is where it is, whether the threshold is applied consistently, whether the interpretation of the criteria that produced 0.6 rather than 0.7 was correct. The questions are legitimate. The fuzzy representation makes them possible. The binary representation prevents them by presenting the outcome as determined rather than judged.
There is a further complexity, which is that fuzzy logic requires a consistent ontology of degrees. If a candidate is 0.6 qualified, the 0.6 has to mean something specific—a place on a defined scale, produced by a defined measurement process, comparable to other 0.6s produced by the same process. This requires the criteria to be specified precisely enough that degrees of satisfaction can be assessed consistently across cases. In engineering control systems, the inputs are physical measurements and the membership functions are defined mathematically. The 0.6 is precise because the input is precise and the function that maps it to the degree is precise.
In institutional decision-making, the inputs are rarely physical measurements and the criteria are rarely specified with sufficient precision to support consistent degree assessment. The criterion of relevant experience, applied to a hiring decision, is not a physical measurement. Two assessors evaluating the same candidate against the same criterion will produce different assessments, not because they are applying different thresholds but because they are interpreting the criterion differently. The fuzzy score 0.6 produced by one assessor and the 0.7 produced by another are not measurements of the same thing. They are the outputs of two different interpretations of a criterion that is insufficiently specified to support consistent measurement.
The binary system has the same problem, but it conceals it. Both assessors decide yes or no. The outcomes may differ for the same reason—different interpretations of an under-specified criterion—but the binary output does not expose the divergence the way the degree would. The 0.6 and the 0.7 are visibly different. The yes and yes are not, even if they were produced by different reasoning.
What the resistance to fuzzy logic reveals about institutions is not primarily a preference for clarity over accuracy, though that preference is real. It reveals the depth of the investment in the appearance of determinacy—in the presentation of institutional decision-making as rule-governed, objective, and not contingent on interpretation. This presentation is a form of authority. The institution that decides on the basis of rules is exercising authority derived from the rules, which is more defensible than authority derived from judgment, because judgment is personal and contestable and rules are not.
The appearance of determinacy requires the binary output. The binary output requires the threshold. The threshold requires the conversion from the continuous reality to the discrete category. The conversion is the act of judgment that the binary system is supposed to eliminate. It does not eliminate it. It hides it—in the specification of the threshold, in the definition of the criteria, in the measurement process that produces the scores the threshold is applied to. The judgment is present throughout. The binary output makes it invisible by presenting the outcome as determined.
Fuzzy logic would make the judgment visible at every stage. The degree of satisfaction is visibly a degree. The conversion is visibly a choice. The threshold is visibly a decision. The institution is visibly interpreting rather than merely applying. This is a more accurate picture of what the institution is doing. It is a less comfortable picture for an institution whose authority depends on presenting what it does as rule-governed and determinate.
The thermostat does not need to be authoritative. It needs to work. The institution needs to be authoritative. Whether it works is a secondary consideration.
The binary is not chosen because it is more accurate.
It is chosen because it is more defensible.
The ajar door does not appear in the record.
The record shows open, or closed.
