Cognitive training for war: how to prepare for uncertainty

Over the course of the ten-year peri­od between 2015 and 2025, cog­nit­ive pre­par­a­tion for war­fare has been the sub­ject of intense research and exper­i­ment­a­tion, both with­in defence insti­tu­tions and in the aca­dem­ic and indus­tri­al worlds. French and allied doc­trines have con­verged towards the same ori­ent­a­tion: train­ing must become exper­i­ment­al, instru­mented and refutable.

In 2015, France’s Dir­ect­or­ate Gen­er­al of Arma­ments stated that “the chal­lenge is no longer to repro­duce, but to exper­i­ment”. In 2021, the Concept d’emploi des forces ter­restres (Land Forces Employ­ment Concept) placed “con­front­ing change” at the heart of cog­nit­ive superi­or­ity. Finally, Le Com­bat aéroter­restre en 2040 (Air-Land Com­bat in 2040), a text from the Com­mandement pour le Com­bat Futur (Future Com­bat Com­mand), emphas­ises the need for pre­par­a­tion based on con­straint, speed, adapt­a­tion and beha­vi­our­al robust­ness. These three texts, sup­por­ted by ana­lyses from the RAND Cor­por­a­tion (2020–2022) and NATO MSG-190 (2021), estab­lish a con­sensus: train­ing must increase adapt­ab­il­ity rather than con­form­ity to a model.

At the same time, the civil and para-aca­dem­ic fields have developed a set of prac­tices grouped under the term spec­u­lat­ive scen­ario plan­ning. Inspired by crit­ic­al design and pro­spect­ive innov­a­tion meth­ods, these approaches use fic­tion as a tool for reflec­tion. Based on nar­rat­ives set in plaus­ible futures, they make it pos­sible to ques­tion the social, polit­ic­al or tech­no­lo­gic­al effects of war by remov­ing overly restrict­ive bar­ri­ers of plaus­ib­il­ity. Their stated object­ive is to stim­u­late the ima­gin­a­tion, broaden per­cep­tions of what is pos­sible and engage the pub­lic in think­ing about the future.

In spec­u­lat­ive scen­ario build­ing, sur­prise is a nar­rat­ive device, not an exper­i­ment­al variable

Cul­tur­ally, the res­ult is indis­put­able: spec­u­lat­ive storytelling has brought war back into the demo­crat­ic debate. Cog­nit­ively, how­ever, stud­ies con­duc­ted in both the mil­it­ary and civil­ian spheres show that it has not pro­duced any meas­ur­able pro­gress in terms of pre­pared­ness. Nar­rat­ive devices cre­ate rep­res­ent­a­tions; they do not gen­er­ate mech­an­isms. Their effect­ive­ness lies in under­stand­ing, not in reac­tion. In some cases, we can even identi­fy a per­form­at­ive intent. For example, the CIA’s “The World in 20.. as seen by the CIA” reports, pub­lished at each US pres­id­en­tial elec­tion, are as much an exer­cise in scen­ario-based foresight as they are a polit­ic­al state­ment, present­ing the world with a par­tic­u­lar stra­tegic vis­ion that shapes the minds of both the United States’ allies and competitors.

Reports by the RAND Cor­por­a­tion (Build­ing Bet­ter Games, 2020) and NATO MSG-190 (2021) identi­fy three recur­ring lim­it­a­tions in these formats:

• lin­ear sequen­cing, which is incom­pat­ible with the con­tinu­ous dynam­ics of inform­a­tion warfare,
• delayed feed­back, which sep­ar­ates learn­ing from action rather than integ­rat­ing it,
• the absence of beha­vi­our­al met­rics, which pre­vents the nar­rat­ive from being linked to observ­able performance.

In spec­u­lat­ive scen­ario build­ing, sur­prise is a nar­rat­ive device, not an exper­i­ment­al vari­able. The coher­ence of the nar­rat­ive takes pre­ced­ence over the meas­ure­ment of the decision. This is par­tic­u­larly evid­ent in the IPCC’s cli­mate scen­ari­os, which present nar­rat­ives of pos­sible futures. They are described as nar­rat­ives trans­lated into sets of socio-eco­nom­ic assump­tions. They make it pos­sible to identi­fy the most sig­ni­fic­ant risks for an organ­isa­tion or ter­rit­ory by choos­ing a prob­able future. They do not allow the impact of each vari­able to be measured.

Insti­tu­tions that use these approaches pur­sue a legit­im­ate goal of rais­ing aware­ness. The aim is to foster a com­mon cul­ture of com­plex­ity and encour­age stra­tegic cre­ativ­ity. How­ever, none of the founders¹ ever claimed, or even sug­ges­ted, that these tools could pre­pare an indi­vidu­al for decision-mak­ing in a dynam­ic envir­on­ment or for the cog­nit­ive con­straints spe­cif­ic to oper­a­tion­al situ­ations. Their object­ive was to raise aware­ness, broaden the ima­gin­a­tion, pro­duce rep­res­ent­a­tions, open up dis­cus­sion and pro­voke intel­lec­tu­al debate. The cur­rent con­fu­sion there­fore stems less from a lack of tools than from an inap­pro­pri­ate real­loc­a­tion of their purposes.

But ped­agogy, how­ever par­ti­cip­at­ory it may be, can­not replace exper­i­ment­a­tion. The dif­fer­ence lies in the nature of the evid­ence: scen­ario plan­ning demon­strates through mean­ing; exper­i­ment­a­tion demon­strates through data. Scen­ario plan­ning raises aware­ness of the need; exper­i­ment­a­tion puts pro­ced­ures in place.

Mil­it­ary-civil­ian stud­ies con­duc­ted over the past ten years – in cog­nit­ive engin­eer­ing, decision-mak­ing psy­cho­logy and sys­tem­ic mod­el­ling – have shown that effect­ive train­ing requires three prop­er­ties: time con­straints (pro­longed expos­ure to time pres­sure), endo­gen­ous feed­back (imme­di­ate effect of action on the envir­on­ment) and meas­ur­ab­il­ity (pro­duc­tion of object­ive data). This approach is already being con­sidered in the world of cyber­de­fence: for example, major USCYBERCOM exer­cises pit teams against each oth­er, with each action modi­fy­ing the over­all sys­tem, gen­er­at­ing imme­di­ate feed­back and, above all, enabling per­form­ance to be meas­ured. The res­ults are con­vin­cing: after sev­er­al years of train­ing, some teams have achieved response time gains of over 30%. None of these cri­ter­ia are fully met by spec­u­lat­ive scen­ario plan­ning. Its con­tri­bu­tions remain qual­it­at­ive. Spec­u­lat­ive scen­ario plan­ning is above all an exer­cise in com­mu­nic­a­tion and mediation.

The prob­lem is also one of pace. Inform­a­tion war­fare moves at the speed of the net­work; nar­rat­ive devices oper­ate at the pace of dis­cus­sion. The gap between the speed of the attack and the slow­ness of the debate cre­ates a meth­od­o­lo­gic­al vul­ner­ab­il­ity. As long as train­ing does not align with the real time of the con­flict, it will lag behind by one cycle of adapt­a­tion, or worse: it will cre­ate the false impres­sion that the threat is being taken into account (this is the cri­ti­cism often lev­elled at armies for pre­par­ing for yesterday’s war instead of anti­cip­at­ing tomorrow’s). The urgency is there­fore cog­nit­ive before it is polit­ic­al: learn­ing faster than the threat is changing.

The ten years of mil­it­ary-civil­ian research stud­ied here lead to a clear dia­gnos­is. Spec­u­lat­ive scen­ario plan­ning is a meth­od of accul­tur­a­tion, not a train­ing tool. It pro­motes aware­ness of the issues among the inex­per­i­enced pub­lic but can­not intrins­ic­ally devel­op either resi­li­ence or respons­ive­ness. It remains use­ful for build­ing shared ima­gin­ar­ies, but it is exper­i­ment­al sim­u­la­tion – open, instru­mented, falsifi­able – that pro­duces last­ing learning.

It is not a ques­tion of oppos­ing two worlds, but of cla­ri­fy­ing their respect­ive regimes of truth. Scen­ario plan­ning explains; sim­u­la­tion veri­fies. The former relies on the coher­ence of dis­course, the lat­ter on the resi­li­ence of real­ity. It is between these two approaches that the future of cog­nit­ive pre­par­a­tion is being played out today. At a time when civil and mil­it­ary author­it­ies are recog­nising the need for a cer­tain frugal­ity, we must nar­row our focus: long-term foresight will always remain essen­tial to the defence of the coun­try, but it can­not be achieved at the expense of imme­di­ate les­sons and adapt­a­tions to emer­ging threats.

For script­ing to be dir­ectly pro­duct­ive, it must be accom­pan­ied by the abil­ity to change the para­met­ers of the story’s envir­on­ment. Not to tell one par­tic­u­lar story, but to allow hun­dreds to be ima­gined. It is not so much the con­sist­ency of the story that mat­ters as the weight that the para­met­ers carry in the eval­u­ation of the tools or con­cepts involved. This is the approach that led the aero­naut­ics industry to devel­op the concept of digit­al twins, in which pre­cise mod­el­ling of air­craft, tur­bines and pro­duc­tion lines allows vari­ables to be changed and the imme­di­ate con­sequences to be observed. It is thus pos­sible to replay the same over­all action an infin­ite num­ber of times, vary­ing the para­met­ers and identi­fy­ing the consequences.

From then on, the scen­ario is, so to speak, incid­ent­al. The key thing is that the sys­tem mod­el­ling is rel­ev­ant. With a robust mod­el, we can ima­gine any scen­ario we want. All we need to do is modi­fy the sets of rules of action for the vari­ous act­ors to assess the con­sequences of each of the variables.

Scen­ario plan­ning allows us to take an intel­lec­tu­al step back, which pre­vents us from imme­di­ately dis­miss­ing the very pos­sib­il­ity of the object or cir­cum­stances under study. Exper­i­ment­a­tion should allow us to iter­ate the para­met­ers until we pro­duce imme­di­ately use­ful les­sons, rather than just food for thought.

Ten years of mil­it­ary-civil­ian research have led to a clear dia­gnos­is: scen­ario plan­ning explains, exper­i­ment­a­tion veri­fies. The former relies on the coher­ence of dis­course, the lat­ter on the res­ist­ance of real­ity. The future of cog­nit­ive pre­par­a­tion lies between these two approaches.