5 principles for taking action in the face of uncertainty

While there are known prin­ci­ples for man­ag­ing sit­u­a­tions of rel­a­tive uncer­tain­ty, it is less clear how to deal with sit­u­a­tions of great uncer­tain­ty, i.e. where the unknown reigns supreme. In these rare cas­es, there is no longer any ref­er­ence to exist­ing knowl­edge (these are known as ‘unknown unknown’ sit­u­a­tions, as opposed to ‘known unknown’ sit­u­a­tions where there is an exist­ing ref­er­ence). In such cas­es, man­age­ment requires rules to those already used in the past.

The first wave of the COVID-19 cri­sis was an exem­plary case of this type of sit­u­a­tion¹. Dur­ing this peri­od, it was impos­si­ble to refer to a past event. Dubbed the ‘flu bug’ for a short time, the high mor­tal­i­ty rates observed quick­ly con­tra­dict­ed this asser­tion. What’s more, hos­pi­tal stays in inten­sive care were much longer than those usu­al­ly observed for oth­er infec­tious res­pi­ra­to­ry virus­es, and loss of smell was a pre­vi­ous­ly unknown symp­tom. There were innu­mer­able ‘sur­pris­es’, mak­ing any rea­son­ing by ref­er­ence to exist­ing knowl­edge tricky.

In response, organ­i­sa­tions grap­pled their way through the process and made, often rushed, deci­sions. Some con­sid­ered that these actions were based on intu­ition and flair. How­ev­er, on clos­er inspec­tion, they reveal a cer­tain form of man­age­ment. This is shown by a study based on inter­views with more than 120 play­ers in the French hos­pi­tal sys­tem, pub­lished recent­ly in the Euro­pean Man­age­ment Review².

The man­age­ment in ques­tion is based on a ratio­nal­i­ty that is nei­ther Carte­sian nor close to the con­cept of High-Reli­a­bil­i­ty-Organ­i­sa­tion (HRO), which is often used in uncer­tain sit­u­a­tions. Rather, it is a prag­mat­ic ratio­nal­i­ty, which con­sists of con­duct­ing a col­lec­tive inves­ti­ga­tion to test hypothe­ses in the field. Five prin­ci­ples emerge, reflect­ing the abil­i­ty of French hos­pi­tal play­ers to demon­strate real­ism in the face of the unknown.

This first prin­ci­ple encour­ages us to design actions with­out wait­ing for per­fect infor­ma­tion. The actions envis­aged at this stage rep­re­sent hypothe­ses. They are derived from the ini­tial results of the sur­vey, which selects them on the basis of the fol­low­ing meth­ods: (i) learn­ing from unusu­al events that can be observed in the field (anom­alies such as the abnor­mal­ly long dura­tion of stays in inten­sive care for patients with COVID-19, which is sur­pris­ing giv­en our knowl­edge of the effects of infec­tious virus­es of the res­pi­ra­to­ry sys­tem); (ii) con­sol­i­dat­ing the reli­a­bil­i­ty of the infor­ma­tion gath­ered by tri­an­gu­lat­ing dif­fer­ent sources or by assess­ing the pro­file of the issuer of the alert (what some peo­ple call ‘epis­temic vig­i­lance’); (iii) and when the hypothe­ses are con­tra­dic­to­ry, to organ­ise debates between the var­i­ous stake­hold­ers in order to reach a col­lec­tive deci­sion (as in the case of the recep­tion of the Chi­nese curves on the inci­dence of the pan­dem­ic, which led to sev­er­al med­ical spe­cial­ists and epi­demi­ol­o­gists being brought togeth­er to com­pare their points of view). In this activ­i­ty, any mod­el­ling effort is also use­ful, but it is not suf­fi­cient alone, because it can nei­ther pro­vide reli­able con­text (the rela­tion­ship between nation­al pro­jec­tions and a local sit­u­a­tion), nor pro­vide a suf­fi­cient­ly con­sis­tent pre­dic­tion, due to a vari­ety of new cri­te­ria that dis­rupt the ‘mod­el’.

This prin­ci­ple requires the hypoth­e­sis to be test­ed in the field, which is the only way to judge its rel­e­vance. Such recourse to the field takes place despite the sur­round­ing igno­rance, and con­di­tions for action that are rarely opti­mal. The objec­tives are guid­ed by a search for evi­dence and learn­ing, while the imple­men­ta­tion con­sists of cir­cum­scrib­ing the test in var­i­ous small stages, each of which is fol­lowed by a rapid return. This incre­men­tal approach opti­mis­es the assess­ment of the rel­e­vance of the action, wast­ing as lit­tle time as pos­si­ble and avoid­ing obvi­ous errors. One exam­ple at the start of the COVID-19 cri­sis was the pri­or­i­ty actions tak­en in the health hos­pi­tal sec­tor (lim­it­ing patient vis­its to hos­pi­tal, set­ting up quar­an­tine mea­sures, staff pro­tec­tion pro­ce­dures, increas­ing the num­ber of inten­sive care beds, to name but the main ones).These mea­sures quick­ly proved their worth. At the same time, they revealed the weak­ness of actions tak­en in EPHAD estab­lish­ments, where many elder­ly and vul­ner­a­ble patients were exposed to the risk of the virus. This obser­va­tion led to the rapid exten­sion and rein­force­ment of actions in this sec­tor, as rec­om­mend­ed in the fol­low­ing principle.

This prin­ci­ple ensures that organ­i­sa­tions trans­late the results of field tests appro­pri­ate­ly, adapt­ing their actions if nec­es­sary. Depend­ing on the feed­back and the col­lec­tive delib­er­a­tion that fol­lows, the play­ers main­tain the ini­tial hypoth­e­sis or mod­i­fy or even refor­mu­late it. The mul­ti­dis­ci­pli­nary nature of the col­lec­tive delib­er­a­tions and the open-mind­ed­ness of the par­tic­i­pants are essen­tial cri­te­ria in this revi­sion. The more the delib­er­a­tion involves a vari­ety of expert view­points, the more like­ly it is that the update will be rel­e­vant.  Sim­i­lar­ly, the more the atti­tudes expressed accept the con­clu­sions of the delib­er­a­tion, the more like­ly it is that the cho­sen hypoth­e­sis will be adopted.

This third prin­ci­ple con­cludes an over­all approach. Com­pris­ing of three stages: (i) def­i­n­i­tion of a hypoth­e­sis; (ii) field test­ing; (iii) updat­ing of the hypoth­e­sis on the basis of the results, it is char­ac­ter­is­tic of an abduc­tion method. We start with a hypoth­e­sis, check its rel­e­vance by glean­ing observ­able facts, and then deduce whether it should be main­tained or replaced by anoth­er. This approach is even more like­ly to be effec­tive if the inves­ti­ga­tion is car­ried out col­lec­tive­ly in order to cap­ture as many clues as pos­si­ble. It is also large­ly depen­dent on the atti­tudes of the mem­bers involved, as the fol­low­ing prin­ci­ple makes clear. 

Accord­ing to this prin­ci­ple, play­ers are encour­aged to express their doubts and to show humil­i­ty (fal­li­bil­ism), because the knowl­edge they have acquired is extreme­ly frag­ile, sub­ject to the appear­ance of new observ­able facts. Sim­i­lar­ly, they are urged to avoid sim­pli­fi­ca­tion by dichotomies between yes and no (dual­ist posi­tions), as these gen­er­al­ly reduce the abil­i­ty to select and inter­pret clues. With­out these two atti­tudes, there is a high risk of mak­ing erro­neous deci­sions and giv­ing them too much weight.

An exam­ple of the impor­tance of this prin­ci­ple was the debate on hydrox­y­chloro­quine as a treat­ment for the virus. The fact that the hypoth­e­sis of such a treat­ment was put for­ward was not in itself shock­ing. It was even jus­ti­fied in the light of what was known about the sub­ject. On the oth­er hand, the fail­ure to ques­tion it when the tri­als car­ried out had not pro­duced con­vinc­ing results is evi­dence of an over­ly assertive, dual­is­tic posi­tion in favour of the ‘yes’ option. 

In this quest for appro­pri­ate atti­tudes, the envi­ron­ment out­side the play­ers involved in the sur­vey plays an impor­tant role, like­ly to trig­ger harm­ful pressures.

Unknown sit­u­a­tions must be man­aged by those who have the most pre­cise knowl­edge of the event. In par­tic­u­lar, if those on the ground have the exper­tise that is being built up (which is often the case), they must be giv­en pri­or­i­ty in the actions to be tak­en. One con­se­quence of this is that they need to be pro­tect­ed from exter­nal pres­sures out­side the scope of the inves­ti­ga­tion. The lat­ter can in fact ham­per the effort under­tak­en. The COVID-19 cri­sis high­light­ed two such pressures:

1. Insti­tu­tion­al pres­sure, which may have called into ques­tion local actions in the name of giv­ing pri­or­i­ty to deci­sions from high­er hier­ar­chi­cal lev­els, even though these are irrel­e­vant in terms of their respec­tive expertise.
2. Pres­sure from the media, which may have steered debates towards con­fronta­tions devoid of nuance, because of the forms of expres­sion imposed. Once again, the debate on hydrox­y­chloro­quine is a good exam­ple: the exchanges on TV often turned into car­i­ca­tured oppo­si­tions between the for and against, lock­ing the play­ers into posi­tions of defend­ing a point of view, far removed from the atti­tudes of fal­li­bil­i­ty and anti-dual­ism that are nec­es­sary, but also dam­ag­ing. As a result, peo­ple work­ing in the field have found them­selves faced with patients who want to under­go treat­ment at all costs, even though there is no evi­dence to sug­gest otherwise. 

Unknown sit­u­a­tions must be man­aged by those who have the most pre­cise knowl­edge of the event. 

The pre­ven­tive mea­sures against these exter­nal pres­sures are to be found at the lev­el of col­lec­tive delib­er­a­tions organ­ised by the play­ers hold­ing the exper­tise. They need to be cau­tious about mak­ing over­ly sim­plis­tic state­ments about the envi­ron­ment. They must also pro­tect them­selves from pres­sure by unit­ing col­lec­tive­ly. How­ev­er, this organ­ised pro­tec­tion must not cut off the exper­tise from infor­ma­tion pro­duced else­where. The prin­ci­ple of col­lec­tive inves­ti­ga­tion means that clues can be gleaned. They must there­fore strike a care­ful bal­ance between pro­tec­tion and selec­tion of sur­round­ing information. 

The five prin­ci­ples are inspired by the the­o­ret­i­cal approach­es of prag­ma­tism, a North Amer­i­can school of thought from the ear­ly 20^th Cen­tu­ry. Tak­en togeth­er, they form a spe­cif­ic form of man­age­ment at hos­pi­tal lev­el, known as High-Prag­mat­ic-Organ­i­sa­tion (HPO), in ref­er­ence to this school of thought. An HPO organ­is­es col­lec­tive sur­veys, engages in abduc­tion process­es, relies on play­ers whose atti­tudes cul­ti­vate doubt and humil­i­ty, and pro­tects itself from exter­nal pres­sure³⁴. By act­ing in this way, French hos­pi­tal stake­hold­ers have shown that hos­pi­tals can oper­ate accord­ing to their own prin­ci­ples to face up to the unknown.

This new con­cep­tu­al term also alludes to anoth­er con­cept, that of High-Reli­a­bil­i­ty-Organ­i­sa­tion (HRO), which is often used to express the man­age­ment required in sit­u­a­tions of uncer­tain­ty⁵. An HRO is an organ­i­sa­tion capa­ble of deal­ing with cri­sis sit­u­a­tions where uncer­tain­ty reigns, by apply­ing dif­fer­ent prin­ci­ples (a hos­pi­tal, but also a nuclear pow­er plant or a sys­tem for organ­is­ing air flights, can thus be assim­i­lat­ed to HROs in the event of a crisis).

With­out going into detail, the pro­posed prin­ci­ples dif­fer from those just described on one essen­tial point: the absence of a ref­er­ence, and there­fore of the pos­si­bil­i­ty of express­ing reli­a­bil­i­ty. As a reminder, reli­a­bil­i­ty aims to min­imise devi­a­tions from a nor­mal state that sets per­for­mance stan­dards⁶. It refers to antic­i­pa­tion by trig­ger­ing pre­ven­tive actions⁷. With HROs, nuclear pow­er plant con­trol oper­a­tors can, for exam­ple, shut down reac­tors if they think that oper­a­tions have entered these zones. Sim­i­lar­ly, aero­plane pilots can refuse to fly if they think the equip­ment or weath­er con­di­tions are dangerous.

How­ev­er, when the play­ers are faced with the unknown, this rea­son­ing is inop­er­a­tive, because the very pur­pose of action is to define what these stan­dards are, and by deduc­tion the prin­ci­ples of pre­ven­tion. For exam­ple, when a remote mon­i­tor­ing sys­tem was set up for patients with mild forms of COVID-19, the ini­tial feed­back showed just how rel­e­vant the sys­tem was. Pre­vi­ous­ly, man­agers had admit­ted these patients with­out real­is­ing that they were occu­py­ing beds unnec­es­sar­i­ly and increas­ing the risk of the virus spread­ing. The pre­ven­tive nature of the action could only be judged fol­low­ing the empir­i­cal test.

Final­ly, it can be said that although the prin­ci­ples of ORH man­age­ment can be found, the absence of a ref­er­ence to nor­mal­i­ty lim­its the appli­ca­tion of the con­cept in the event of an unknown sit­u­a­tion. In these sit­u­a­tions, the ref­er­ence is not deter­mined in advance but is con­struct­ed a pos­te­ri­ori. For this rea­son, the con­cept of HPO, and its five prin­ci­ples, seems more appro­pri­ate to deal with them.