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.