The Steersman and the Double: A Control-Theoretic Reading of Mimetic Escalation, from Wiener to Land

Introduction

The word cybernetics names a steersman. Norbert Wiener took it from the Greek kybernētēs in 1948 to name “the entire field of control and communication theory, whether in the machine or in the animal,” and he noted that the English governor, the device James Clerk Maxwell had analysed in 1868, descends from the same word through a Latin corruption^[1][2]. The founding image of the field is therefore a helmsman correcting his course: an error is measured and a correction applied, and the ship holds its heading.

Half a century later, a very different use of cybernetic vocabulary emerged in the writings of Nick Land and the circle around the Cybernetic Culture Research Unit at Warwick. There the privileged figure is no longer the governor but the runaway: positive feedback, self-amplifying loops, capital as a process that escapes every attempt to regulate it^[3]. If Wiener’s cybernetics is a theory of how systems hold steady, Land’s is an eschatology of how they stop holding.

This paper argues that the passage from one to the other is not merely a change of mood. It has a definite structure, and René Girard’s mimetic theory describes that structure more exactly than either author does. Girard holds that human desire is imitative: we want what our models want. When the model is distant (a saint, a fictional hero, a god), imitation is peaceful. When the model is close enough to be a rival, imitation turns into conflict. The conflict escalates, and the rivals come to resemble each other more and more until they are “doubles,” undifferentiated and interchangeable^[4]. In his last major book Girard read Clausewitz’s “escalation to extremes” as the historical form of this dynamic and gave it an apocalyptic horizon^[5].

Our contribution is to show that this account can be stated as a small, exact result in linear control theory, and that the result explains why both escalation and undifferentiation happen at once rather than as two separate phenomena. We then use the result to read two theological responses: the recently proposed “Jesuit cybernetics”^[6] and an Eastern alternative rooted in the patristic notion of synergeia. The philological history of the word kybernēsis runs through the argument, because the word was a theological term before it was an engineering one.

The governor and its reference

A negative-feedback controller has three parts: a reference $r$ (the desired state), an output $x$ (the actual state), and a law that moves $x$ to reduce the error $r - x$. In its simplest continuous form,

$$\dot{x} = k\,(r - x), \qquad k > 0.$$

If $r$ is constant, the solution relaxes exponentially to $r$ at rate $k$. The helmsman reaches his heading and holds it.

Everything in this picture depends on one assumption that is so natural it is rarely stated: the reference is exogenous. It is set from outside the loop and does not depend on the output. The heading is fixed by the navigator, not by where the ship happens to be pointing.

Wiener was aware that the choice of reference was the dangerous part. In God & Golem, Inc. (1964), subtitled A Comment on Certain Points where Cybernetics Impinges on Religion, he returned repeatedly to the danger of machines that pursue their stated purpose literally: the monkey’s paw, the sorcerer’s apprentice, the genie who grants the wish as spoken^[7]. His worry was that a well-functioning governor aimed at a badly chosen reference is worse than no governor at all. He did not consider the case that concerns us here, which is stranger: a reference that is not badly chosen but moves, because it is coupled to the output of another controller that is watching you.

Mimetic desire as a coupled loop

Take two agents. Each has an output $x_i$, which we can read as the intensity of desire for, or effort toward, some contested object. Following Girard, neither agent’s desire has its own origin. Each takes its reference from the other: I want it because, and as much as, you want it. Let each agent aim to slightly exceed the other, so that the reference is $r_1 = (1+\varepsilon)\,x_2$ and $r_2 = (1+\varepsilon)\,x_1$ with a small $\varepsilon > 0$. This is the minimal form of rivalry: not only to have what the other has, but a little more.

Substituting into the governor law gives a linear system:

$$\begin{pmatrix}\dot{x}_1\\ \dot{x}_2\end{pmatrix} = k\begin{pmatrix}-1 & 1+\varepsilon\\ 1+\varepsilon & -1\end{pmatrix}\begin{pmatrix}x_1\\ x_2\end{pmatrix}.$$

The matrix is symmetric, and its eigenvectors are the symmetric mode $(1,1)$ and the antisymmetric mode $(1,-1)$. Along the symmetric mode the rate is $k(-1 + 1 + \varepsilon) = k\varepsilon$. Along the antisymmetric mode it is $k(-1 - 1 - \varepsilon) = -k(2+\varepsilon)$. Writing $s = x_1 + x_2$ for the sum and $d = x_1 - x_2$ for the difference,

$$s(t) = s(0)\,e^{k\varepsilon t}, \qquad d(t) = d(0)\,e^{-k(2+\varepsilon)t}.$$

This is the whole result, and it can be read directly in Girard’s terms.

The sum grows without bound. However small the margin $\varepsilon$, total desire escalates exponentially. Nothing in either agent is unstable. Each is individually a perfectly good governor. The instability belongs to the coupling. This is the escalation to extremes, and it matches Clausewitz’s own diagnosis that war tends toward the absolute through Wechselwirkung, reciprocal action, rather than through the intentions of either side^[5:1].

The difference decays, and faster than the sum grows. Whatever initially distinguished the rivals, whatever head start or temperament or particular reason for wanting the object, is erased at rate $2+\varepsilon$ times the base gain. The rivals converge on each other. This is what Girard calls the crisis of undifferentiation and the emergence of monstrous doubles: each rival sees in the other a monster, while from outside they are indistinguishable^[4:1].

The significance of the result is that these are not two phenomena but one. A single coupling produces both modes, so any system that escalates this way must also undifferentiate. Girard insisted that the violence of rivals and their loss of difference were inseparable. The linear model shows why they must be: they are the two eigenvectors of the same matrix.

Even with $\varepsilon = 0$, where each agent merely wants as much as the other, the symmetric mode is neutral (rate zero) while the difference still decays at rate $2k$. Pure imitation without the desire to exceed does not escalate, but it still erases difference. The addition of any positive margin tips the neutral mode into growth.

External and internal mediation

Girard distinguished two forms of mediation. In external mediation the model is at a distance that makes rivalry impossible: Don Quixote imitating Amadis of Gaul, or a Christian imitating Christ. In internal mediation the model is close enough to compete for the same objects: a neighbour, a colleague, a brother^[8].

In control terms the distinction is exact. External mediation is a loop whose reference is exogenous. If each agent tracks a fixed model $m$, so $r_i = m$, the equations decouple, each $x_i$ relaxes to $m$, and the system is stable. The agents come to resemble one another, since both approach $m$, but they do so without rivalry, because neither’s output enters the other’s reference. Internal mediation is the case where the reference is another agent’s output. The loop closes through the rival, and the coupled matrix of Section 3 applies.

This gives a precise sense to a claim that could otherwise sound pious: a transcendent model stabilises desire because it is outside the loop. The theological property of transcendence (being unaffected by, and not in competition with, the creature) corresponds to the engineering property of exogeneity. A model that can be neither possessed nor outdone cannot be escalated against.

The scapegoat as controller

Girard’s archaic societies do not escalate forever. The crisis resolves when the violence of all against all turns into the violence of all against one. A victim is selected, often arbitrarily, and the community unites against it. Its expulsion or death restores peace, and the victim is afterwards remembered as both the cause of the crisis and the source of its resolution: the sacred^[4:2][9].

In the model this is a change of reference structure. Before, each agent’s reference was another agent. After, every agent’s reference is the same object: the victim. With $n$ agents all tracking one common target, the system becomes a consensus process. Hostility that was reciprocal becomes convergent, and convergence toward a shared value is the stable dynamic studied in the consensus literature since DeGroot^[10]. The scapegoat is a controller in the most literal sense. It restores an exogenous reference by manufacturing one, and it pays for the stability with a victim.

This reading also explains why the mechanism must be concealed to work. If the community sees that the victim was arbitrary, the reference loses its exogeneity. The victim becomes one more agent in the loop, and the rivalries resume. Girard’s central claim about the Gospels is that they reveal exactly this: the innocent victim, the arbitrary unanimity of the crowd^[9:1]. Revelation, on this account, disables the governor.

Land: the loop without a reference

This is where Land enters. His writings of the 1990s celebrate what the model identifies as the symmetric mode: a process whose growth depends on no one’s intention, which absorbs every attempt to regulate it, and which dissolves inherited differences (of place, tradition, kinship, even the human) into a single accelerating flow^[3:1]. Capital, in this picture, is a loop whose only reference is its own previous output: accumulation for the sake of accumulation.

Read through Section 3, Land’s accelerationism is the choice to affirm the coupled system rather than to seek a new reference. It is the escalation to extremes, praised. Girard reached the same diagnosis and treated it as the apocalypse. In Battling to the End he argued that once the scapegoat mechanism has been exposed, humanity has lost its archaic brake and faces a stark choice between the renunciation of rivalry and unlimited escalation^[5:2]. Land and Girard agree on the dynamics. They disagree on the sign.

Wiener, for his part, feared a machine that obeyed a bad reference. Land welcomes a process that has no reference at all. Girard shows why the second is what human desire produces when it is left to reference itself.

The word before the machine

The steersman was a theological and political figure long before he was a mechanical one. Plato used kybernētikē, the helmsman’s art, as a standing analogy for the art of governing a city and a soul^[11]. André-Marie Ampère, classifying the sciences in 1834, coined cybernétique for the science of civil government, over a century before Wiener^[12].

The biblical history is richer still. In Proverbs, the Hebrew noun taḥbulot (תַּחְבֻּלוֹת), usually translated “counsel” or “guidance,” derives from the root ḥ-b-l, “rope.” The related ḥōbēl is a sailor, a rope-hauler, and in Jonah 1:6 the ship’s captain is the rav ha-ḥōbēl, the chief of the rope-men^[13]. Taḥbulot is thus originally rope-work: the handling of lines by which a ship is steered. The Septuagint translators saw this and rendered it with kybernēsis. In Proverbs 1:5 the one who understands “will acquire kybernēsis”. In 11:14, “those who have no kybernēsis fall like leaves”^[14].

Paul then places the word inside the life of the Church. In 1 Corinthians 12:28, among the gifts that God has set in the body (apostles, prophets, teachers, powers, healings), he lists kybernēseis, gifts of guidance or steering^[15]. The steersman’s art here is neither a technique nor a sovereign power but a charism: something given, exercised within a body, and oriented to a head that the body does not control.

The philology thus anticipates the structure of Section 4. Guidance in the biblical sense is always guidance toward something not produced by the one who steers. The rope is pulled from outside the loop.

Two theological cybernetics

Ryan Haecker has recently proposed a “Jesuit cybernetics”: a Catholic and specifically Jesuit vision of cybernetic theory that, following Henri de Lubac and Hans Urs von Balthasar, assumes “a religious turn from the sufficiency of human reason to the receptivity of divine revelation and grace”^[6:1]. We do not attempt a full account of that programme here. We note only that the Jesuit tradition already contains a remarkably explicit feedback practice. The Ignatian examen is a daily comparison of one’s actual movements against a reference (the will of God as discerned), with attention to the error signals Ignatius called consolation and desolation, followed by correction^[16]. In control terms it is a negative-feedback loop with an exogenous reference, run at a daily sampling rate.

The same tradition also produced the most famous positive-feedback theology of the twentieth century. The Jesuit palaeontologist Pierre Teilhard de Chardin described cosmic history as a process of ever-increasing complexity and consciousness converging on an Omega Point^[17]. Teilhard’s Omega is formally a runaway, but one with a destination: escalation that converges rather than diverges. Critics have long noted its affinity with later transhumanist and accelerationist imaginaries^[18]. A Jesuit cybernetics has to decide whether it is Ignatian (a governor with a transcendent reference) or Teilhardian (a sanctified escalation), and the model suggests these are not compatible.

The Christian East offers a third option through the concept of synergeia. Paul calls the apostles theou synergoi, “God’s co-workers” (1 Cor. 3:9)^[15:1], and the patristic tradition, in Maximus the Confessor above all, developed from this an account of salvation as the free cooperation of the human will with divine energy^[19]. In control terms synergy is neither the ship’s own governor nor external command. It is a coupling between the agent and a reference that is transcendent in essence yet present in its energies. It is exogenous without being remote. If internal mediation couples an agent to a rival, and external mediation to a distant model, then synergy couples an agent to a model that cannot be rivalled but can be cooperated with. The model of Section 4 says such a coupling is stable. The theological tradition says it is deifying. These claims are not the same, but they are consistent.

Limits

The model in Section 3 is linear, deterministic, and two-agent, and real mimetic systems are none of these. Saturation, noise, and network structure would all change the details, and the scapegoat discussion in Section 5 is qualitative. The claim is not that Girard’s theory reduces to a matrix. The claim is narrower: the coincidence of escalation and undifferentiation, which Girard asserted on anthropological and literary grounds, follows from the simplest possible formalisation of internal mediation, and the stabilising role of transcendence corresponds to a standard property of feedback systems. These are structural analogies, offered as tools for thinking, not as proofs about theology.

Conclusion

Wiener’s steersman held his course because his heading came from outside the ship. Land’s process accelerates because it has no heading but itself. Girard explains what happens in between: when desire takes its reference from a rival, the loop closes through the other, and the rivals escalate and become identical in one motion. Archaic religion stopped the escalation with a victim. The Gospels, on Girard’s reading, removed that brake. What is left, if not Land’s runaway, is a reference that cannot be rivalled. The biblical tradition had a word for the gift of steering toward it long before engineers borrowed the term.