The Opening Shot
The people who understand cognitive warfare best may be institutionally unable to do anything on what they understand so well. Not because they are stupid. Not because they are cowardly. Because they have been trained to treat the problem as if it were not physical.
Information is physical. Thoughts are physical. Every synaptic firing is an electrochemical event. Every attention allocation is a measurable behavioral output. [Empirical, confidence: high — see attention-economy and platform-design literature cited below] The adversary does not debate whether cognition is a battlespace. The adversary engineers it.
We — who had a hand in building the internet, who optimized the platforms, who wrote the legal frameworks that govern them — are fighting while sitting on our own hands. Not because an enemy forced us into this position. Because we voluntarily did it ourselves, with categories that no longer match the world we inhabit.
We are not losing everywhere. We are losing specific, identifiable battles — in specific modalities, against a specific architecture — and pretending the loss is total is as much a category error as pretending the problem is unphysical. [Inferential, confidence: moderate — this essay argues for where and why, not that the outcome is settled]
Two Axes, Not One
Before going further, a conflation has to be cleared, because getting it wrong here poisons everything downstream.
“Intelligentized warfare” (智能化战争, zhìnénghuà zhànzhēng) is not China’s term for cognitive warfare. It is the successor concept to “informatized warfare” (信息化) — the digitization of the battlespace, networks, data links, C4ISR. Intelligentization adds AI as a decision-accelerant: autonomous targeting, predictive logistics, algorithmic command-and-control. [Empirical] This axis runs human decision → network-enabled decision → AI-accelerated decision. It is about how fast a side can close its OODA loop. It is operational and material.
“Cognitive warfare” (认知战, rènzhī zhàn) and “cognitive domain operations” (认知域作战, rènzhī yù zuòzhàn) sit on a different axis entirely. [Empirical, source: PLA Daily, Qiushi, AMS military academic journals] This is not about faster decisions. It is about whose decision architecture is operating at all. The cognitive domain is described in PLA writing as the space of consciousness, knowledge, and belief — not the network, but the human-adjacent endpoints of the network. “Mind superiority” (制脑权) and “command of cognition” (制认知权) name the goal: dominance over what an adversary can perceive as threat, imagine as response, coordinate as will.
Intelligentized warfare asks: how fast can we decide?
Cognitive warfare asks: can the adversary decide at all, on his own terms?
The first is friction reduction in your own system. The second is friction injection into the enemy’s system — not to slow his OODA loop, but to make the loop irrelevant because he is no longer orienting on the same reality. Collapsing these two into one Chinese “master concept” is the kind of error that reads as insight to a general audience and as amateurism to a RAND desk officer. They are parallel investments, not synonyms.
What the Adversary Actually Built
The PLA does not speak of the cognitive domain as metaphor. The terms above — 认知战, 认知域作战, 制脑权 — circulate in open military journals. They are theorized, debated, revised across years of publication. [Empirical] The transparency of Chinese doctrine is not the problem, and pretending otherwise flatters us at no cost to the adversary.
The problem is transparency of authorization. China publishes the theory. It does not publish who is targeted, when, with what resources, under what internal sign-off. The decision layer — the actual deployment against an actual population — has no equivalent to congressional hearings, FOIA litigation, or public court challenges to platform law. [Inferential, confidence: moderate — based on comparative absence of equivalent accountability mechanisms in PRC/Russian systems relative to the US]
This is not a claim that one side thinks in public and the other in secret. Both sides think out loud, extensively, in published military and academic literature — Lefebvre’s reflexive control theory (рефлексивное управление) and its Soviet and Russian successors are as openly published as the PLA’s cognitive-domain writing. Reflexive control did not stay a 1960s theoretical exercise — it was carried forward through the Soviet General Staff and into post-Soviet Russian military-theoretical writing by figures like Leonenko and Makhnin, who worked out its application to command-and-control and information-psychological confrontation (информационно-психологическое противоборство) doctrine. This is worth stating plainly because Western commentary often collapses it into the “Gerasimov Doctrine” — a term Gerasimov himself never used and that Mark Galeotti, the analyst who coined it, has since disavowed; the real doctrinal lineage is older, more specific, and runs through Lefebvre’s mathematics, not through a single 2013 speech. The asymmetry is narrower and sharper than “opacity versus transparency”: only one side is structurally required to expose who decided and under what authority once the theory becomes an operation.
The Doctrine We Have, The Capability We Do Not
Current US joint doctrine — JP 3-04, the 2023 DoD Information Operations Strategy, JADC2, “all-domain operations” — explicitly frames domains as simultaneous, not sequential. [Empirical] “All-domain” is a term of art chosen precisely to reject a layered stack. On paper, the US does not see layers where China sees a field. It sees the field.
What it does not have is the executable authority to operate on it.
The legal and organizational permission to act against the domestic cognitive environment — or coherently against an adversary’s civilian information space in peacetime — does not exist in the form the strategy documents assume. [Inferential] We can write “all-domain, simultaneous, cognitive-as-precondition.” We cannot execute it, because the theory-to-practice bridge China has built through Party-state fusion has no analogue here, by design and by choice.
This is not conceptual blindness. It is a doctrine/capability gap, and it is more dangerous than ignorance, because it produces the simulation of readiness. The analyst can point to JP 3-04 and say “we know.” The operator cannot point to a legal framework, a budget line, or a command authority that lets them act on that knowledge against the actual battlespace.
The Court, The Code, and The Fracture
In July 2024, the Supreme Court decided Moody v. NetChoice. The ruling held that platforms’ algorithmic curation — selecting, ranking, promoting content — is protected editorial speech under the First Amendment, explicitly analogized to a newspaper’s editorial judgment. [Empirical, source: SCOTUS opinion, 2024]
The Court left two doors open. First, it noted the ruling does not necessarily reach algorithms that respond solely to user behavior rather than enforcing an editorial viewpoint — “non-expressive” curation may receive different treatment. Second, it named antitrust and consumer-protection law as untouched by the decision: “nothing said here puts regulation… off-limits as to a whole array of subjects.” [Empirical, direct citation of the Kagan opinion]
Here is the dilemma this creates. The most dangerous algorithmic curation may not be the one with an opinion. It may be the one with no opinion at all — that simply learns what keeps a user scrolling and serves it without ideological filter. The Chinese approach to a Western population is not chiefly to tell it what to think; it is to make it incapable of sustaining a thought long enough to act on it. And the pure engagement-optimization engine — the “non-expressive” recommender — may be the only kind of algorithm the First Amendment does not protect.
But there is a technical fracture underneath the legal one. In a deployed recommender system, there is no clean separation between the part enforcing community guidelines and the part predicting engagement. It is typically the same model, the same weights, a blended objective function. [Inferential, based on production ML architecture] The Court can draw the line cleanly. The weights cannot.
The First Amendment, as currently interpreted, requires a legal category — the pure engagement engine with no editorial fingerprint — that may not correspond to anything that exists in a deployed system at scale. We have built a doctrine that assumes a separation between editorial judgment and behavioral prediction that production machine learning was never built to implement. This is not a bug in the law. It is a mismatch between legal epistemology and engineering reality — and unlike the intelligentized/cognitive conflation in Section II, this one is ours, not the adversary’s.
The Operational Lane That Exists
The precise, defensible claim is not “treat the algorithm as critical infrastructure.” Moody narrowed that path.
The lane the Court left open is narrower and specific: the FTC treating engagement-optimization design as a consumer-protection and antitrust matter — dark patterns, addictive-design liability, market-dominance-enabled dependency. [Empirical, based on ongoing FTC enforcement on dark patterns and children’s online safety] This is not hypothetical. It is live, and it is checkable against current FTC enforcement activity. It is the theory-to-practice bridge that actually exists, not the one the strategy documents assume.
The US does not lose cognitive battles because it is free. It loses specific battles because its freedom is architected in ways an adversary can exploit without reciprocity — not because the adversary hides its thinking and we expose ours, but because only one side is required to expose who decided, and under what authority, once doctrine becomes operation. The First Amendment is not the weakness. The asymmetry in accountability at the authorization layer is.
The Closing Weight
We spent decades building a domestic information environment optimized for engagement and are now surprised an adversary can exploit it. [Symbolic — not load-bearing]
The attention consumed by that environment is physical. The infrastructure that allocates it is physical. The legal architecture governing it is physical. The doctrine-capability gap is physical, in the sense that it has a budget line, an org chart, and a missing signature. The only thing that was never physical is our insistence that none of this counts as a battlespace.
We are not losing the war. We are losing the specific engagements where that insistence still holds — and every one of them is identifiable, and every one of them is fixable, if we stop mistaking the map for the terrain.
When you wake up on Monday Morning
For the analyst who accepts the frame, the first move is not legislative. It is cartographic.
Map the domestic information environment as a supply chain with single points of failure. Not narratives. Not “misinformation.” Infrastructure: which platforms allocate attention at scale, which recommendation architectures operate without editorial transparency, which engagement-optimization systems create dependency cascades measurable in dwell-time and return-rate metrics. This is CISA’s lane in a way Moody doesn’t reach — assessment is not regulation of curation; it’s the difference between mapping a fiber-optic cable’s failure points and telling the phone company what to carry on it. The legal firewall against domestic operations is real. The first breach is conceptual: someone with authority must name the mapping as defensive, not surveillance.
For the operator with FTC reach, the move is enforcement acceleration. The dark-patterns and children’s-online-safety dockets are live and demand attention. The theory that engagement-optimization design constitutes a consumer-protection harm does not require new statutory authority. It requires a plaintiff with standing, a theory of harm that survives motion to dismiss, and a willingness to treat platform design as product design, not speech. The Court left this lane open. The cases must be built.
For the strategist with budget authority, the move is organizational. The doctrine/capability gap is not a training problem. It is a structural problem: the legal permissions, the command authorities, the budget lines for cognitive defense against civilian-targeting adversary operations do not exist in the form the strategy documents assume. Someone must ask Congress for them, explicitly, with the political cost named. The alternative is to continue funding the simulation of readiness.
For the lawyer with appellate reach, the move is test-case preparation. The “non-expressive” curation category in Moody is undeveloped. It will remain undeveloped until a case forces the Court to confront whether a blended-objective recommender can be unbundled for First Amendment purposes. The case will lose at the district level. It may lose at the circuit. It only needs to reach the Court with a record clean enough to matter. This is a decade timeline. The adversary operates on shorter cycles. The case must be filed anyway.
None of this wins the war. Each of it wins a specific, identifiable engagement. The essay’s closing claim — that every loss is fixable — is not a promise. It is a conditional. If we stop treating cognition as unphysical, if we stop pretending our doctrine equals our capability, if we accept that the legal architecture we cherish also creates exploitable asymmetries — then specific battles become winnable. Not all of them. Not tomorrow. The ones we can name.
The fire is physical. The hand that tends it must be physical too. Monday morning, someone must pick up the phone. The legal firewall is not a shield for the nation; it is currently a vacuum chamber where our capabilities go to die. If the “cognitive warriors” remain asleep, they will continue to mistake the courtroom theater for the actual battlespace. They will celebrate narrow, semantic legal victories while completely losing command of cognition on the ground.
And we will continue trying to swim in the mirror.
