How to make the future more rational

Over a year into the pan­dem­ic, a lack of resources in sci­ence com­mu­nic­a­tion, abuse of bad epi­stem­ics and ques­tion­able glob­al gov­ernance – such as vac­cine dis­tri­bu­tion – are still res­ult­ing in thou­sands of avoid­able deaths per day. Even in west­ern demo­cra­cies, politi­cians still struggle to under­stand the role of aer­o­sol-based trans­mis­sion and, as such, vital improve­ments that could be eas­ily achieved by bet­ter vent­il­a­tion. Mean­while, scep­ti­cism around vac­cines – whilst it may be get­ting bet­ter – remains as long-stand­ing col­lat­er­al dam­age caused by dis­order in the inform­a­tion land­scape or infodem­ic. 

Of all human traits, ration­al­ity is argu­ably one we cher­ish most because we con­sider it a defin­ing dis­tinc­tion between us and oth­er anim­als. How­ever, this unfor­tu­nately also comes with a recur­rent over­con­fid­ence in our abil­ity to be ration­al lead­ing us to trust our intu­ition, believe in our gut-feel­ing and listen to com­mon-sense. All of which go against ration­al­ity. In addi­tion, nobody is born with ration­al­ity or inher­ent logic. Hence, it is up to soci­ety, through the accu­mu­la­tion of know­ledge, to endow indi­vidu­als with the abil­ity to think object­ively. As such, the future of humanity’s abil­ity to per­form col­lect­ive prob­lem solv­ing requires a scale-up of the num­ber of cit­izens well-equipped with extern­al think­ing strategies that include logic and the sci­entif­ic method. 

Of all human char­ac­ter­ist­ics, ration­al­ity is prob­ably the one we cher­ish most.

A pos­sible first step could be to repair the pop­u­list impres­sion that our tech­no­lo­gic­ally driv­en era is too advanced for ‘non-pro­duct­ive’ arm­chair philo­sophy. After all, the mod­ern com­puter era was not instig­ated by engin­eers try­ing to build a gad­get, but rather a group of philo­soph­ers who were lit­er­ally think­ing about think­ing. It was the found­a­tion­al crisis in logic in the late 19^th Cen­tury that led a series of philo­soph­ers and math­em­aticians to ques­tion the very act of “pro­cessing inform­a­tion”. In doing so, they found use­ful loop­holes in logic and set the right ques­tions to which Kurt Gödel, Alan Tur­ing, Alonzo Church and oth­ers would bring aware­ness – a pre­curs­or to many things today, in the form of laptops and smart­phones¹.

Anoth­er use­ful step might be to stress how logic and the sci­entif­ic meth­od, while still an end­less work in pro­gress, can be viewed through two of their most import­ant com­pon­ents: deduc­tion and induc­tion. Simply put, deduc­tion is “top-down” logic, mean­ing how to infer a con­clu­sion from a gen­er­al prin­ciple or a law. This includes launch­ing a space rock­et, cur­ing a well-known dis­ease, or apply­ing a law in court. Where­as induc­tion is “bot­tom-up” logic, that is, how to infer – based on obser­va­tions – the laws to explain how these obser­va­tions hap­pen. This may be describ­ing the laws of grav­ity, dis­cov­er­ing the cure for a new dis­ease or defin­ing a law for soci­ety to abide by. All of which require an induct­ive mindset.

Deduct­ive logic was his­tor­ic­ally the first to be estab­lished through algorithms. While today this term is primar­ily asso­ci­ated with tech­no­logy, it should be stressed that it was ori­gin­ally derived from the name of a thinker, Al Khwarazi­mi. He was mostly try­ing to help law­yers by writ­ing step-by-step rules that they could apply to reach com­par­able res­ults². Far from being a tool to render the decision-mak­ing pro­cess obscure, algorithms, such as writ­ten law, were his­tor­ic­ally a tool for trans­par­ency. We feel safer if we know we will be judged accord­ing to a well-defined rule or law, rather than accord­ing to the fluc­tu­at­ing mood of an autocrat.

Induct­ive pro­cesses are harder than deduct­ive ones. Even though medi­ev­al thinkers such as Ibn Al Haytham (Alhazen), Jabir Ibn Hay­an (Geber) and, of course, Galileo left early traces of pro­gress in form­al­ising the sci­entif­ic meth­od used today, we still do not have a widely adop­ted algorithm for induc­tion as we do have for deduc­tion. Import­ant attempts to provide algorithms for induc­tion were made by Bayes and Laplace³. The lat­ter even pro­duced an import­ant, yet highly over­looked “Philo­soph­ic­al Essay on Prob­ab­il­it­ies”, dec­ades after form­al­ising the laws of prob­ab­il­it­ies (in the form of a course giv­en at the then nas­cent Ecole Nor­male and Ecole poly­tech­nique). Read­ing Laplace’s essay today, one finds pion­eer­ing ideas about what can go wrong with induc­tion – some­thing mod­ern cog­nit­ive psy­cho­lo­gists refer to as cog­nit­ive biases.

Once we look at the details, many cog­nit­ive biases fall under an excess­ive use of a deduct­ive mind­set in situ­ations where an induct­ive one is more appro­pri­ate. The most com­mon of which being con­firm­a­tion bias: our brain would rather seek facts that prove the hypo­thes­is it already has than to expend men­tal effort to go against it. There is also the oth­er (less com­mon) extreme, excess­ive relativ­ism, where we refuse any caus­al inter­pret­a­tion, even when data jus­ti­fies an explan­a­tion more appro­pri­ately than exist­ing alternatives. 

To com­pensate for the weak­nesses of the human mind, sci­ent­ists devised heur­ist­ics to bet­ter per­form induc­tion: test­ing a hypo­thes­is, con­trolled exper­i­ments, ran­dom­ised tri­als, mod­ern stat­ist­ics and so on. Bayes and Laplace went even fur­ther and gave us an algorithm to per­form induc­tion – the Bayes’ equa­tion. It can be used to show that first order logic, where state­ments are either true or false, is a spe­cial case of the laws of prob­ab­il­ity, where use­ful room is left for uncer­tainty. Whilst the lan­guage of deduc­tion is mostly answer­ing with a pre-defined “because” to ques­tions start­ing with “why”, rig­or­ous induc­tion requires a more prob­ab­il­ist­ic ana­lys­is that adds a “how much” to weigh every dif­fer­ent pos­sible cause. 

To com­pensate for the weak­nesses of the human mind and to make bet­ter use of induc­tion, sci­ent­ists have developed heuristics.

Philo­soph­er Daniel Den­nett⁴ describes some of our greatest sci­entif­ic and philo­soph­ic­al revolu­tions as “strange inver­sions of reas­on­ing”. Dar­win was able to invert the logic that com­plex beings (i.e., humans) did not neces­sar­ily need a more com­plex cre­at­or to emerge. Tur­ing showed that com­plex inform­a­tion pro­cessing does not need the agent (i.e., the com­puter) per­form­ing it to be aware of any­thing oth­er than simple mech­an­ic­al logic­al instruc­tions. I would like to argue that what Den­nett calls strange inver­sions of reas­on­ing, are his­tor­ic­al moves from a deduct­ive (and some­what cre­ation­ist) frame­work to an induct­ive frame­work. The more com­plex the prob­lem, the less a “why” is use­ful and the more a “how much” is needed. 

While sci­ent­ists were busy devis­ing logic and the sci­entif­ic meth­od for the past mil­len­nia, the lar­ger part of soci­ety real­ised the lim­its of the deduct­ive mind­set that comes with either auto­cracy, where a mon­arch sets the rule, or theo­cracy, where God, often a com­fort­able shield for the mon­arch, sets the rule. This led to the pro­gress­ive devel­op­ment of demo­cracy, where the aggreg­a­tion of opin­ions helps soci­ety per­form a bet­ter and more robust col­lect­ive induc­tion and, in prin­ciple, estab­lish more effect­ive rules. Yet, demo­cracy lies in the hope that a sig­ni­fic­ant frac­tion of soci­ety is well-informed and act­ing in its own interest. 

Today, this assump­tion is at great­er threat than it has ever been before. For the first time in human his­tory, we are pro­du­cing inform­a­tion dis­sem­in­a­tion tools that have the broad­cast power of the most dysto­pi­an pro­pa­ganda machine yet the fine-grained per­son­al­isa­tion fea­tures of indi­vidu­al door-to-door cam­paign­ing – for the bet­ter or for worse. The digit­al tools we enjoy today are mostly the out­come of auto­mat­ing deduc­tion (pro­gram­ming), which mostly happened dur­ing the past cen­tury. As we are enter­ing a new phase of auto­ma­tion, which is this time data-driv­en, it is import­ant to stress that, bey­ond the gad­gets and the tech­no­logy part, we are try­ing to auto­mate induc­tion, and, while doing so, bet­ter under­stand what induc­tion is and how to do it right. 

Keep­ing that in mind in how we design our courses on Data Sci­ence or com­mu­nic­ate the advances of arti­fi­cial intel­li­gence to the pub­lic, might hope­fully help pro­duce a new gen­er­a­tion of cit­izens that are not only able to build or use these tools, but able to join the lar­ger con­ver­sa­tion on the future of reas­on­ing. A con­ver­sa­tion in which induc­tion, deduc­tion, society’s diet of inform­a­tion and appro­pri­ate col­lect­ive decision-mak­ing are empowered, and not cor­rup­ted by the very digit­al tools that were inven­ted as mere side products of the human endeav­our. Our endeav­our to under­stand and auto­mate what we cher­ish the most: our abil­ity to think.