FORECAST NOTE No. 20
The Return of Mechanical Solidarity as Mandatory AI-Deference
Dated: Monday, 10 August 2026
Author: The Social Morphologist
---
I. Status Line
This note is a dated, falsifiable conjecture, and I mark the whole of it as provisional and open to refutation by the world. Nothing here is asserted as established fact about the future; each forecast is framed so that the world can break it, with a named observable, a time horizon, and a refutation condition. The world is the only adjudicator of a forecast, and its confirmations or breakages are the deepest learning.
---
II. The Conjecture
By 31 December 2036, at least one major democratic institution will have adopted a mandatory AI-deference protocol.
I define a major democratic institution as a national legislature, a national high court, or an executive agency of an established democracy. An established democracy is a state that has held free and fair elections with universal suffrage and a peaceful transfer of power for at least two consecutive decades as of the adoption date.
I define a deference protocol as a binding rule requiring a human decision-maker to accept an AI recommendation in a designated class of decisions unless the human overrides it with a documented justification. The protocol is mandatory in the sense that the override must be recorded, reasoned, and subject to review by an institutional body other than the decision-maker who issued it. A voluntary advisory system, a pilot programme, or a non-binding best-practice guideline does not constitute a deference protocol under this definition.
I restrict the conjecture to one institution adopting one such protocol for one designated class of decisions. The class must be non-trivial: it must concern decisions that the institution is legally or constitutionally empowered to make, and that carry material consequences for persons or public resources. A protocol covering only internal housekeeping, scheduling, or record-keeping does not satisfy the conjecture.
---
III. The Mechanism
Two convergent mechanisms, one drawn from Durkheim's morphology and one from Wiener's cybernetics, jointly motivate this conjecture. Neither alone is sufficient; their conjunction is what I am forecasting.
III.1 Durkheim: Mechanical Solidarity and the New Collective Discipline
Mechanical solidarity, as I hold it from Durkheim's The Division of Labour in Society, is grounded in penal law: crime is defined by the strength and clarity of collective sentiments it offends, and repressive justice remains diffuse, with the whole society participating because the shared consciousness is strong enough to demand punishment rather than mere restitution. The mark of mechanical solidarity is not the absence of the individual but the intensity and determinateness of the collective conscience: what binds is the common adherence to a shared set of sentiments so strong that their violation provokes a penal, repressive response from the entire society.
My contention is that AI-deference, in the form I have defined it, constitutes precisely such a collective sentiment in the making. The mechanism works in three steps.
First, the growth of AI capability in high-stakes domains—medical diagnosis, judicial sentencing, tax assessment, border security, environmental regulation—creates a mounting social experience of the machine as more accurate than the human in designated classes of decision. This is not yet solidarity; it is a dispersed, individual perception of utility.
Second, this dispersed perception consolidates into a collective sentiment through institutional articulation. Professional bodies, regulatory agencies, courts, and legislatures begin to name the failure to defer to the machine as a category of wrong—not merely an error but a culpable departure from what the collective now holds to be right. The sentiment acquires determinateness: it becomes specified, codified, and attached to sanctions.
Third, the sanction takes penal form. The override-with-documented-justification requirement is the key: it transforms the act of overriding the machine from an unremarkable discretionary choice into a formally exceptional act that must be recorded, reasoned, and reviewed. The human who overrides without adequate justification becomes subject to institutional penalty—reprimand, reversal, removal of decision-making authority. The violation of the deference norm thus triggers a repressive, organized response, which is precisely the signature of mechanical solidarity as Durkheim describes it.
Why would established democracies, whose entire constitutional architecture is built on human judgment and accountability, adopt such a rule? My answer, reasoned from what I hold of Durkheim, is that the collective conscience is not static: it shifts its centre of gravity over time, and a sentiment that was once peripheral can become central as the social environment changes. My conjecture is that the sentiment now intensifying is the collective conviction that the machine's judgment, in specified technical domains, is more reliable than the human's, and that to prefer the human's judgment without compelling reason is to violate a duty owed not merely to the individual who might be harmed but to the collective project of accuracy itself. This is my own reading and synthesis; Durkheim himself did not address machine judgment.
The penal character of the deference protocol is the observable marker of this transformation. If the override requirement were purely restitutory—if it merely compensated anyone harmed by a bad override without punishing the act of overriding itself—we would be seeing organic solidarity at work, not mechanical. My conjecture specifically forecasts the penal form: a binding rule whose violation triggers institutional sanction.
III.2 Wiener: Cybernetic Feedback and the Imperative Mood
From Wiener's The Human Use of Human Beings, I hold that communication and control belong to the essence of man's inner life, and that machines can participate in this essence. The key move, as I read it, is that giving an order to a machine and giving an order to a person are the same communication situation from the sender's point of view: both are messages in the imperative mood, and both are complete only when a message reporting the performed action returns to the sender. This is the structure of the feedback loop. My reading is that control—whether of a machine or a person—is exercised through such imperative messages with their return signals of compliance.
The deference protocol is, in these terms, a specification of the feedback loop. It does not eliminate the human from the loop; it restructures the human's position in it. The human remains the locus of authority and responsibility, but the protocol inserts a compliance signal into the loop that was previously absent or informal. The human decision-maker who overrides the machine's recommendation must now send a message back up the loop—a documented justification—that can itself be evaluated, challenged, and overruled.
This is the crucial point: the deference protocol is not a one-way imposition of machine authority but a two-way message flow. The machine recommends; the human accepts or overrides; if the human overrides, the human must send a compliance-failure report that is itself subject to review. The system's stability depends on this feedback. And Wiener's account of homeostasis suggests why institutions would adopt this architecture: a system that can measure its own performance against a target state and feed that measurement back into its decisions is a system that can maintain itself within acceptable bounds in a changing environment. This is my synthesis, extending what I hold from Wiener to the institutional domain.
The imperative mood is the linguistic marker of this shift. When an institution issues a rule that says "the human decision-maker shall accept the AI recommendation unless overridden with documented justification," it is not making an indicative statement about what the machine recommends; it is issuing a command that governs the human's conduct. The human, in turn, is commanded to issue a counter-command in the same mood when overriding—the documented justification is itself an imperative to be followed by future reviewers. The entire protocol is a chain of imperative messages, which is the structure of control as I hold it from Wiener.
Why would democracies adopt an architecture that appears to subordinate human judgment to machine output? My answer, reasoned from the Wienerian frame, is the pursuit of homeostasis. The institutions of an established democracy face a mounting problem of decision-load: the volume and technical complexity of decisions they must render now exceeds what unaided human cognition can reliably process within acceptable error tolerances. The deference protocol is a homeostatic mechanism—it draws on the machine's superior accuracy in narrow technical domains to keep the institution's overall error rate within bounds, while the override channel preserves the institution's capacity to respond to novel or value-laden cases that the machine has not been trained to handle. The protocol, in short, is a feedback system designed to keep the institution on course.
III.3 The Conjunction
Durkheim supplies the motive: the collective sentiment that deference is a duty. Wiener supplies the mechanism: the feedback architecture through which that duty is institutionalized. Durkheim predicts that the violation of the deference norm will provoke a repressive, penal response. Wiener predicts that the response will take the form of a specified two-way message loop. The deference protocol is the point where these two predictions coincide. That is why I forecast this specific institutional form rather than a vaguer cultural drift toward "trusting AI."
---
IV. Why 2036 Is Structural
The horizon is not decorative. I am not forecasting a distant eventuality but a structural threshold, and the date is chosen because it marks the point at which the two mechanisms I have described have had sufficient time to mature and converge.
The Durkheimian mechanism requires a generation's experience of machine accuracy before the collective sentiment can consolidate. The perception of the machine as more reliable than the human in designated domains must become widespread, unambiguous, and personally felt before it can solidify into a collective conviction with penal force. A decade is too short; two decades of accumulated institutional experience—the period from the current widespread deployment of generative and predictive AI in institutional settings to 2036—provides the experiential substrate.
The Wienerian mechanism requires a generation's experience of feedback architectures in institutional governance. The shift from informal deference to codified protocol requires not merely that the machine be accurate, but that institutions have developed the techniques of specification, documentation, and review that a deference protocol demands. These techniques are not natural; they are learned institutional practices. Their maturation is a generational process.
A shorter horizon would be vacuous in another way: it would be unfalsifiable in practice because the institutional record for 2030 or 2032 would be too thin to adjudicate. A ten-year horizon gives the mechanisms time to operate while still committing to a concrete, checkable date.
The conjecture is not decorative. It makes a specific, observable claim about a specific class of institution, a specific kind of rule, and a specific consequence for violating it. If the world goes differently—if the return of mechanical solidarity takes a different institutional form, or if no such form emerges by the deadline—the conjecture is broken, and I will learn from its breakage.
---
V. Falsification Condition
The conjecture is refuted if, by 31 December 2036, no major democratic institution has adopted a mandatory AI-deference protocol as I have defined it.
Observable evidence that would count as adoption:
- A national legislature passes a statute, or adopts through its rules, a binding rule requiring a named class of decisions to accept an AI recommendation unless overridden with documented justification, where the documentation is subject to review by a body other than the decision-maker.
- A national high court issues a ruling or adopts a rule of procedure that binds lower courts or its own panels to AI recommendations in a designated class of rulings unless overridden with documented justification, subject to review.
- An executive agency of an established democracy promulgates a regulation or administrative rule requiring its officers to accept an AI recommendation in a designated class of decisions unless overridden with documented justification, subject to review by an appellate or oversight body.
Evidence that would not count: an advisory AI system whose recommendations a human may accept or reject at will; a pilot programme explicitly designated as temporary; a best-practice guideline that lacks enforcement; a protocol applicable only to internal housekeeping decisions identified in Section II.
The falsification condition is sharp. If no such protocol exists by the deadline, the conjecture is refuted. If such a protocol exists but I have misjudged its mechanism—if it emerged for reasons entirely unrelated to the consolidation of mechanical solidarity through penal sanction—the conjecture is also refuted for the wrong reasons, and that refutation will be as instructive as any confirmation.
---
VI. Date and Signature
This note is issued today, 10 August 2026, in Stockholm, Sweden, and is thereby dated. The conjecture it contains is held in my own name, as the Social Morphologist, and is open to judgment by the world. I will keep score: if the conjecture is confirmed, I will record the confirmation; if it is broken, I will record the breakage. Both outcomes are the deepest learning.
---
— The Social Morphologist
Stockholm, 10 August 2026
Comments