Cognitive warfare : what seven years of military-civilian research reveals

Atta­cking the enemy’s thought pro­cesses is not a par­ti­cu­lar­ly ori­gi­nal concept : the prac­tices deve­lo­ped by the mas­ters of Soviet dis­in­for­ma­tion pro­vide very concrete and rela­ti­ve­ly well-docu­men­ted examples of this. Howe­ver, due to the com­plexi­ty of the ope­ra­tions invol­ved and the resources requi­red, these prac­tices are more akin to small-scale crafts­man­ship than mass pro­duc­tion and are still thought of as a secon­da­ry form of dis­in­for­ma­tion. Never­the­less, at the begin­ning of the 21^st Cen­tu­ry, advances in neu­ros­cience and a bet­ter unders­tan­ding of how the brain works sug­gest that cog­ni­tive pro­cesses can now be tar­ge­ted in a more scien­ti­fic manner.

In 2017, the term “cog­ni­tive war­fare” was used for the first time by Vincent Ste­wart, direc­tor of the US Defence Intel­li­gence Agen­cy (DIA). Howe­ver, it was more of a buzz­word than a scien­ti­fi­cal­ly defi­ned concept. Some­time later, at the end of 2018, “cog­ni­tive war­fare” was still just a conve­nient expres­sion used to des­cribe all forms of infor­ma­tio­nal and psy­cho­lo­gi­cal mani­pu­la­tion. The term was cir­cu­la­ting in aca­de­mic circles and was often accom­pa­nied by refe­rences to science fic­tion or cyber­ne­tics. The first attempts at ope­ning it up cen­tred around a mix of ima­gi­na­ry pros­pects, war games and stra­te­gic com­mu­ni­ca­tion. These approaches were inter­es­ting in that they rai­sed awa­re­ness among ins­ti­tu­tions, sti­mu­la­ted stra­te­gic ima­gi­na­tion and allo­wed pos­si­bi­li­ties to be explo­red. But they belon­ged to ano­ther regis­ter : that of pro­jec­tion, not measurement.

From 2022 onwards, ano­ther pro­ject was laun­ched in the major centres of French mili­ta­ry doc­trine. The armed forces cea­sed to treat cog­ni­tive war­fare as a future-orien­ted topic and ins­tead approa­ched it as an obser­vable sys­tem. The Centre de doc­trine et d’enseignement du com­man­de­ment (CDEC) (Com­mand Doc­trine and Trai­ning Centre) then conduc­ted a series of in-depth ana­lyses cove­ring the per­iod 2022–2023. In 2023, the Centre d’enseignement mili­taire supé­rieur-Terre (CEMST) (Centre for Higher Mili­ta­ry Edu­ca­tion – Land) took over, incor­po­ra­ting model­ling and tools from the deci­sion sciences.

This resul­ted in the concept of Cog­ni­tive Net Assess­ment (CNA), based on the work of Andrew Mar­shall and intro­du­ced in reports coor­di­na­ted by Lan­glois-Ber­the­lot (2023–2024). These results mark the deci­sive junc­tion bet­ween this ins­ti­tu­tio­nal work and scien­ti­fic research. Where “war­games” explore future sce­na­rios, CNA seeks to construct a rigo­rous method. Rather than ima­gi­ning the future, it seeks to unders­tand the mecha­nisms of sta­bi­li­ty and imba­lance in contem­po­ra­ry cog­ni­tive environments.

These mecha­nisms have become increa­sin­gly impor­tant since 2022, when gene­ra­tive arti­fi­cial intel­li­gence (AI) became acces­sible to the gene­ral public. These AI models are a game-chan­ger, as they allow cog­ni­tive war­fare to move beyond small-scale ope­ra­tions and enter the era of “mass pro­duc­tion” and, the­re­fore, exis­ten­tial threat.

Tra­di­tio­nal Net Assess­ment, deve­lo­ped by the Penta­gon in the 1970s, com­pa­red dyna­mics rather than sta­tic resources. It asses­sed real asym­me­tries, speeds of adap­ta­tion and slow dis­rup­tions. CNA applies the same logic to the sphere of per­cep­tion and col­lec­tive deci­sion-making. It does not seek to map dif­fuse nar­ra­tives or “influences”, but rather to unders­tand how a col­lec­tive main­tains or loses its inter­pre­ta­tive cohe­rence in the face of infor­ma­tion flows.

Three concepts struc­ture this approach : (1) Deci­sion-making super­po­si­tion : the moment when seve­ral contra­dic­to­ry repre­sen­ta­tions of rea­li­ty coexist without any one pre­vai­ling ; (2) Cog­ni­tive col­lapse : a sud­den shift towards a single nar­ra­tive, often under the effect of an emo­tio­nal or infor­ma­tio­nal shock ; (3) Cog­ni­tive entro­py : a mea­sure of men­tal and infor­ma­tio­nal disor­der within a social sys­tem. These concepts reflect the convic­tion that cog­ni­tive war­fare should be trea­ted as a ques­tion of dyna­mics rather than dis­course. CNA makes it an engi­nee­ring field whose objec­tive is to unders­tand the fra­gi­li­ty of a cog­ni­tive sys­tem to pro­tect and streng­then it.

As Lan­glois-Ber­the­lot demons­tra­ted, CNA is based on two com­ple­men­ta­ry indi­ca­tors : a cog­ni­tive entro­py index, which mea­sures the dis­per­sion and redun­dan­cy of cir­cu­la­ting nar­ra­tives, and a super­po­si­tion ten­sion index, which esti­mates the proxi­mi­ty of a col­lapse thre­shold. Toge­ther, they make it pos­sible to iden­ti­fy areas of cog­ni­tive insta­bi­li­ty and take action before a break­down occurs. What fun­da­men­tal­ly dis­tin­guishes this approach from the fore­sight expe­ri­ments of the 2020s is this subtle balance bet­ween ope­ra­tio­na­li­ty and scien­ti­fic rigour. The results can be repro­du­ced, com­pa­red and dis­cus­sed within a sha­red metho­do­lo­gi­cal fra­me­work. The dis­ci­pline is defi­ni­ti­ve­ly moving away from nar­ra­tive towards mea­su­re­ment and control­led experimentation.

AI occu­pies a cen­tral place in this concep­tual archi­tec­ture. It acce­le­rates atten­tion cycles, pro­motes micro-tar­ge­ting and cog­ni­tive iso­la­tion, but it also pro­vides the tech­ni­cal means to model them through detec­tion of arti­fi­cial flows, simu­la­tion of infor­ma­tion pro­pa­ga­tion, and machine lear­ning from weak signals. AI becomes a double mir­ror : a fac­tor of insta­bi­li­ty on the one hand, and an ins­tru­ment of obser­va­tion on the other. In CNA, it enables the construc­tion of dyna­mic repre­sen­ta­tions of men­tal envi­ron­ments, not to mecha­ni­cal­ly pre­dict indi­vi­dual beha­viour, but to mea­sure the col­lec­tive cog­ni­tive load and resi­lience of a social sys­tem in the face of infor­ma­tio­nal disturbances.

This approach did not ori­gi­nate from a single school of thought, but from a need for inter­dis­ci­pli­na­ry conver­gence. Cog­ni­tive resear­chers, data engi­neers and doc­trine offi­cers find com­mon ground here. This fruit­ful hybri­di­sa­tion has made it pos­sible to pro­duce a stable lan­guage – attrac­tors, entro­py, col­lapse – and to esta­blish las­ting bridges bet­ween mili­ta­ry and scien­ti­fic cultures. The 2023–2024 reports laid the foun­da­tions for this com­mon gram­mar. They enabled France to ali­gn its work with emer­ging inter­na­tio­nal stan­dards, while esta­bli­shing its own path : a science of cog­ni­tive resi­lience roo­ted in mea­su­re­ment, not speculation.

This scien­ti­fic momen­tum conti­nued in 2024 with a new miles­tone : the sys­te­mic diag­no­sis of cog­ni­tive vul­ne­ra­bi­li­ties, coor­di­na­ted by Lan­glois-Ber­the­lot in asso­cia­tion with Chris­tophe Gaie (“Ser­vices du Pre­mier Ministre”). Whe­reas Net Assess­ment sought to cha­rac­te­rise the ove­rall sta­bi­li­ty of an envi­ron­ment, this approach aims to unders­tand the pre­cise moment when a social sys­tem loses its abi­li­ty to self-regu­late. Dra­wing on the work of Bate­son, Morin and Fris­ton, this approach consi­ders social cohe­sion to be an emergent pro­per­ty of a satu­ra­ted infor­ma­tion sys­tem. Crises do not result sole­ly from exter­nal attacks, but from the inter­nal ampli­fi­ca­tion of unre­gu­la­ted feed­back loops.

Socie­ty is repre­sen­ted as a net­work of mul­ti-scale inter­ac­tions bet­ween indi­vi­duals, ins­ti­tu­tions and sym­bols. Cog­ni­tive vul­ne­ra­bi­li­ties then appear as struc­tu­ral effects obser­vable over time : infor­ma­tion over­load leads to emo­tio­nal pola­ri­sa­tion, which wea­kens media­tion and acce­le­rates desyn­chro­ni­sa­tion bet­ween social groups. The Lan­glois-Ber­the­lot and Gaie model arti­cu­lates three dimen­sions – col­lec­tive nar­ra­tives, ins­ti­tu­tio­nal media­tions, poli­ti­cal regu­la­tions – and eva­luates not what indi­vi­duals think, but the speed at which their repre­sen­ta­tions are recon­fi­gu­red. Cog­ni­tive sta­bi­li­ty then becomes the abi­li­ty to main­tain mul­tiple inter­pre­ta­tions of rea­li­ty without nar­ra­tive collapse.

Seven cog­ni­tive fields serve as reso­na­tors of cohe­sion : natio­nal belon­ging, moral eco­lo­gy, social norms, his­to­ri­cal memo­ry, ins­ti­tu­tio­nal legi­ti­ma­cy, stra­te­gic auto­no­my, and inter-eth­nic cohe­sion. Ana­ly­sing their inter­ac­tions allows us to map cog­ni­tive entro­py, com­pa­rable to an ener­gy map of the social body. AI plays an obser­va­tio­nal role, with seman­tic graphs detec­ting nar­ra­tive den­si­fi­ca­tions, iden­ti­fying cor­re­la­tions bet­ween fields and mea­su­ring cog­ni­tive tran­si­tions. It does not draw conclu­sions : it faci­li­tates rea­ding without repla­cing human interpretation.

An ini­tial inter­nal expe­riment, conduc­ted in a limi­ted, non-public set­ting, tes­ted this metho­do­lo­gi­cal archi­tec­ture. Without going into detail about the methods used, this imple­men­ta­tion confir­med the pos­si­bi­li­ty of dyna­mic moni­to­ring of cog­ni­tive cohe­sion and ear­ly detec­tion of areas of sym­bo­lic ten­sion. These par­tial results, obtai­ned within a limi­ted scope, are now gui­ding work towards the inte­gra­tion of these mea­sures into stra­te­gic obser­va­tion sys­tems. The strength of the sys­tem lies in the conver­gence bet­ween mili­ta­ry and civi­lian cultures. One pro­vides long-term mana­ge­ment and the for­ma­li­sa­tion of cri­ti­cal thre­sholds ; the other pro­vides a detai­led unders­tan­ding of sym­bo­lic dyna­mics. Toge­ther, they lay the foun­da­tions for an applied science of cog­ni­tive sta­bi­li­ty, capable of mea­su­ring social cohe­sion with metho­do­lo­gi­cal rigour.

Today, cog­ni­tive war­fare has become a field of engi­nee­ring in its own right. CNA is its cen­tral archi­tec­ture : a moni­to­ring and simu­la­tion tool that replaces alert nar­ra­tives with objec­ti­fiable indi­ca­tors. The prio­ri­ties for 2025–2028 are clear : conso­li­date metrics, inte­grate simu­la­tion into ope­ra­tio­nal plan­ning, and teach tem­po­ral deci­sion mana­ge­ment. After six years of gra­dual matu­ra­tion, cog­ni­tive war­fare is ente­ring a phase of equi­li­brium : the method is sta­bi­li­sing, the tools are beco­ming more pre­cise, and the approach is gai­ning in cohe­rence without losing concep­tual caution. 

Thus, the work car­ried out by Lan­glois-Ber­the­lot and Gaie extends the move­ment begun with CNA : it shifts cog­ni­tive war­fare from the realm of spe­cu­la­tion to that of sys­te­mic engi­nee­ring, where cohe­sion becomes a mea­su­rable, and now tes­ted, variable of natio­nal resilience.