Are algo­rithms tak­ing over?

Algo­rithms, spe­cial forms of arti­fi­cial intel­li­gence, are influ­enc­ing our mod­ern world more and more. They analyse a lot of data in a short time and dis­cov­er new, sur­pris­ing infor­ma­tion and con­nec­tions. We humans sim­ply place our trust in the poten­tial of algo­rithms because our minds are not able to under­stand the com­plex­i­ty of infor­ma­tion pro­cess­ing and smart devices often make our lives more pleas­ant by reliev­ing us of the cog­ni­tive work. How is this dis­rup­tive tech­nol­o­gy chang­ing our every­day life, our soci­ety and our democracy?

In the lat­est of her Duet inter­views, Dr Cal­daro­la, author of Big Data and Law, and Sophie Borchert dis­cuss the pos­si­bil­i­ty of a shift from democ­ra­cy towards an algo­c­ra­cy tak­ing place.

Not every­one is famil­iar with the word algo­c­ra­cy. “Cra­cy” stands for rule and “algo” comes from algo­rithm. By “algo­c­ra­cy” do you mean the rule of algo­rithms? Is algo­c­ra­cy, like democ­ra­cy, a new form of gov­ern­ment? Or mere­ly a form of tech­nol­o­gy (per­haps even a dis­rup­tive tech­nol­o­gy) and/or a social stimulus?

Ms Borchert: The lack of famil­iar­i­ty with the term is hard­ly sur­pris­ing for it is quite new, a neol­o­gism in fact. Orig­i­nal­ly it illus­trat­ed the soci­o­log­i­cal dis­tinc­tion between mar­ket and bureau­cra­cy and in that con­text describes a sys­tem in which human behav­iour is not con­trolled by eco­nom­ic mech­a­nisms or the admin­is­tra­tive machin­ery, but by algo­rith­mic process­es that link both with each oth­er under the influ­ence of globalisation.

I look at algo­c­ra­cy from the per­spec­tive of the polit­i­cal sci­ences, in par­tic­u­lar legal sci­ences, where it joins sim­i­lar, in part odd con­cepts like medi­acra­cy (rule of the media) or gynecoc­ra­cy (rule of women) – just as well as democ­ra­cy (rule of the peo­ple). There­fore, I inter­pret algo­c­ra­cy in sim­ple terms as the rule of algo­rithms, as you recog­nised cor­rect­ly, focus­ing on the gov­ern­men­tal aspect of this rule.

Although tech­nol­o­gy makes up a cen­tral part of that mod­el and rule always adheres to social char­ac­ter­is­tics because of the state’s roots in soci­ety, I wouldn’t reduce algo­c­ra­cy to that. It is more than the sum of its parts, name­ly a form of gov­ern­ment or more pre­cise­ly a form of rule. The dif­fer­ence lies in the fact that the for­mer out­lines only the for­mal con­struct of a state, the out­er struc­ture in which rule is embed­ded, where­as the lat­ter also com­pris­es the inner workings.

Thus, when talk­ing about algo­c­ra­cy, I mean a sys­tem of pow­er based on algo­rith­mic, dig­i­tal­ly imple­ment­ed process­es in a pub­lic dimen­sion. That has noth­ing to do with arti­fi­cial forms of some sort of super-intel­li­gence from the world of sci­ence fic­tion that usurp state author­i­ty and take pow­er to rule accord­ing to their own “will“ to oppress mankind. No, I refer to far more mun­dane and latent, but no less pow­er­ful trends which might be able to infil­trate democracy.

In this respect, the aim of my study is the analy­sis of the con­nect­ing fac­tors and effects of those trends from a jurispru­den­tial point of view to answer the ques­tion whether we are real­ly about to drift from the demo­c­ra­t­ic rule of the peo­ple towards an auto­crat­ic rule of algorithms.

My favourite quote:

“Any fool can know. The point is to understand!”

Albert Ein­stein

Did you end up with the provoca­tive title of your dis­ser­ta­tion “From democ­ra­cy to algo­c­ra­cy?” because algo­rithms are “more intel­li­gent” than the mass intel­li­gence of peo­ple or more auto­crat­ic because they can process more facts, infor­ma­tion and knowl­edge and are not judgmental?

Defin­ing intel­li­gence is a bit of a tricky issue. Though inde­fin­able for humans, it is quick­ly attrib­uted to algo­rithms – often too swift­ly, with­out know­ing the tech­ni­cal details.

Algo­rithms per se are not intel­li­gent, they needn‘t have any­thing to do with com­put­ers. They are basi­cal­ly step-by-step-instruc­tions for the con­ver­sion of input into out­put. All steps are clear­ly defined in the form of a con­di­tion­al if-then-scheme, so that each input pro­duces exact­ly one out­put. Sim­ple recipes and alge­bra­ic func­tions belong to such deter­min­is­tic rules. Fur­ther­more, trans­lat­ed into dig­i­tal code an algo­rithm can be used as soft­ware for data pro­cess­ing, often fol­low­ing sta­tis­ti­cal or heuris­tic pro­ce­dures in order to auto­mate rea­son­ing (infer­ence).

Based on this, Arti­fi­cial Intel­li­gence (AI) refers to dig­i­tal sys­tems of algo­rithms hav­ing the abil­i­ty to “learn“. AI is inde­ter­min­is­tic, tech­ni­cal­ly speak­ing. With regard to a spec­i­fied tar­get, it is able to react to unpre­dictable changes in its envi­ron­ment or its input by autonomous­ly adapt­ing its inter­nal code struc­ture. This doesn’t hap­pen on the basis of lin­ear pro­grams con­sist­ing of fixed com­mands, but dynam­ic chains of actions; after one “if“ come sev­er­al “then“ com­mands. The adap­tion, called machine learn­ing, takes place through train­ing of the algo­rith­mic mod­el by means of var­i­ous data sets – valu­able, not to say the most valu­able resources in the age of Big Data.

For this rea­son, AI is supe­ri­or to us humans in terms of scale and speed of infor­ma­tion pro­cess­ing. But this tech­nol­o­gy is nei­ther in a posi­tion to break through the pro­grammed lim­its nor to devel­op cre­ativ­i­ty, emo­tions or even a con­scious mind. It just imi­tates human ratio­nal think­ing and behav­iour as so-called Weak AI. We are far from the devel­op­ment of Strong AI and a tech­no­log­i­cal singularity.

Nev­er­the­less, an enor­mous haz­ard is already inher­ent even in Weak AI. It is true that in the absence of its own will AI is inca­pable of form­ing its own opin­ion. In fact, it is neu­tral. But human judge­ment can become appar­ent in tech­ni­cal bias­es, either in the design of the machine learn­ing mod­el, in the train­ing of data or dur­ing its prac­ti­cal appli­ca­tion after the train­ing. In con­nec­tion with anoth­er weak spot fatal con­se­quences can occur: The con­tin­u­ous adap­tion of imple­ment­ed rules and with that the con­sis­tence of the out­put can­not be retraced by the human user, espe­cial­ly not in case of deep learn­ing occur­ring in Arti­fi­cial Neur­al Net­works. Con­se­quent­ly, bias­es remain unde­tect­ed and pos­si­bly false results are accept­ed with­out ques­tion. When they con­tribute to gov­ern­men­tal deci­sions, like a sen­tence in court or a rejec­tion of an admin­is­tra­tive appli­ca­tion, the demo­c­ra­t­ic char­ac­ter of these deci­sions must be called into question.

Let’s go back to the word “rule” for a moment. Can only sub­jects rule or what about a tech­nol­o­gy? You said you defined algo­c­ra­cy as “a gov­ern­ment-scale sys­tem of rule based on algo­rith­mic, dig­i­tal­ly imple­ment­ed process­es”. Have I under­stood you cor­rect­ly in that algo­rithms are a tech­ni­cal tool of a rul­ing author­i­ty? Then who is the ruler – the one who devel­ops and monop­o­lis­es the algo­rithms (e.g. GAFAM), or the peo­ple who use them? Who or what legit­imis­es rule? Fur­ther­more, your quote from Albert Ein­stein sug­gests that it is cru­cial to “under­stand”? Do we under­stand algo­rithms? Can peo­ple con­scious­ly decide for or against dig­i­tal process­es in the age of fake news, deep fakes, and agnos­tics, which Daniel Goeude­vert crit­i­cised in his Duet Inter­view? Or are we pow­er­less in the face of them because we can­not under­stand them – espe­cial­ly giv­en their com­plex­i­ty and speed – and do we trust them more than our own knowl­edge and instincts? I see a risk of us los­ing the abil­i­ty to crit­i­cal­ly exam­ine some­thing due to algo­rithms and mean­while our sense of judg­ment, deci­sion-mak­ing and prob­lem-solv­ing dimin­ish­es as a con­se­quence. It is ques­tion­able whether an indi­vid­ual with doubts about the “cor­rect­ness” of the algo­rith­mic can result, but with­out a lob­by and fund­ing, would they even be in a posi­tion to crit­i­cal­ly view the algo­rith­mic sys­tem and its effects? This begs the ques­tion: will algo­rithms take on a life of their own and actu­al­ly become the new rul­ing force – espe­cial­ly when they learn to sup­ply them­selves with the nec­es­sary elec­tric­i­ty and doesn’t that make us think of AI weapons that can then no longer be switched off?

Actu­al­ly, you have already answered the first and the last ques­tion by your­self. Algo­c­ra­cy is based on algo­rith­mic process­es, sim­i­lar to epis­toc­ra­cy being based on knowl­edge and plu­toc­ra­cy on mon­ey. Nev­er­the­less, algo­rithms don’t ele­vate them­selves to being rulers, but remain a mere vehi­cle. That might sound rather strange and the­o­ret­i­cal. In this regard, let me explain the intel­lec­tu­al basis cre­at­ed by the soci­ol­o­gist Max Weber. He defined pow­er as the “oppor­tu­ni­ty to impose one’s will in a social rela­tion­ship even if there is some oppo­si­tion, regard­less of what the oppor­tu­ni­ty has been based on“ (own trans­la­tion). It makes up the essence of rule, which, for its part, Weber defines as an “oppor­tu­ni­ty for a cer­tain type of order to be obeyed by an assigned group of people“(own trans­la­tion). Weak AI, the only tech­ni­cal­ly fea­si­ble device at the moment, lacks impor­tant qual­i­ties, such as the desire for pow­er, that are nec­es­sary for it to qual­i­fy as rule. If one day the cre­ation of a Super-AI with an own mind should be suc­cess­ful, the sit­u­a­tion might be a dif­fer­ent one. But then the world would be turned upside down any­way, not only with regard to con­sti­tu­tion­al law.

The next ques­tion is hard­er to answer. Who is the actu­al ruler, if not algo­rithms them­selves, espe­cial­ly when their lack of trans­paren­cy forces us to react to their out­put unin­tel­li­gi­bly? My answer may sound unsat­is­fac­to­ry, but I don’t know (yet). In the mat­ter of rein­ing in algo­rithms, the chance of obe­di­ence depends on under­stand­ing and con­trol­ling them. Only the per­son who can artic­u­late an order knows how to make use of the imple­ment­ed medi­um. And only some­one who can ensure that the medi­um has car­ried out the order com­plete­ly can expect obe­di­ence and the recip­i­ent is able to retrace it. This have to be the peo­ple as the sov­er­eign in a democ­ra­cy accord­ing to Arti­cle 20 (2) of the Grundge­setz (Ger­man Basic Law, GG) and yet giv­en the opac­i­ty, com­plex­i­ty and auton­o­my of intel­li­gent pro­grams, which are used by new gen­er­a­tions of media to per­son­alise the flow of infor­ma­tion and by the gov­ern­ment to increase the effi­cien­cy of its task ful­fil­ment, that sup­po­si­tion will be a chal­lenge. I’m par­tic­u­lar­ly inter­est­ed in whether peo­ple are able to have free will at all and whether that can be con­vert­ed by state rep­re­sen­ta­tives giv­en that algo­rithms can seem­ing­ly influ­ence opin­ion and deci­sion mak­ing. If not, peo­ple are no longer rul­ing but rather the group con­trol­ling algo­rithms are those in pow­er, such as their devel­op­ers, oper­a­tors or data man­agers – as „algo­crats“. Anoth­er option is that the algo­crat­ic ten­den­cies assume small­er pro­por­tions than feared and peo­ple, in spite of demo­c­ra­t­ic deficits, keep their sov­er­eign­ty. I haven’t come to a final con­clu­sion on that so far.

Who or what legit­i­mates algo­crat­ic rule, that was the sub­ject of your third ques­tion. It was Max Weber again who for­mu­lat­ed impor­tant ideas con­cern­ing that issue by iden­ti­fy­ing pow­er as a very volatile, unsta­ble phe­nom­e­non. Thus, true rule requires sta­bil­is­ing the bal­ance of pow­er, rep­re­sent­ed by anoth­er attribute: legit­i­ma­cy. Legit­i­ma­cy means accep­tance, in oth­er words the jus­ti­fi­ca­tion of gov­ern­men­tal domin­ion, which aris­es from the con­vic­tion of those being ruled that it is legit­i­mate. In this con­text Weber set up three ide­alised forms of legit­i­mate rule. Approval of the per­form­ing of pow­er results either from a belief in con­for­mi­ty with a for­mal, nor­ma­tive set of rules and reg­u­la­tions (ratio­nal domin­ion), in the integri­ty of tra­di­tion­al, sacro­sanct hier­ar­chies (tra­di­tion­al domin­ion) or in the aura of the per­son in charge of pow­er (charis­mat­ic domin­ion) and the com­mand accept­ed on that basis. With regard to algo­c­ra­cy, I have near­ly run out of expla­na­tions once again. If peo­ple retain sov­er­eign­ty under algo­rith­mic influ­ence, then legit­i­ma­cy is ensured accord­ing to Arti­cle 20 (2) GG. But if algo­crat­ic vio­la­tions become appar­ent in the demo­c­ra­t­ic sys­tem, there is a need to clar­i­fy whether this new form of domin­ion is still legit­i­mate or whether it over­steps the bound­ary of ille­git­i­ma­cy. Going by Weber’s tril­o­gy, you could take ratio­nal rule into con­sid­er­a­tion, if algo­rithms were able to con­sis­tent­ly fol­low the legal sys­tem and its inher­ent val­ues – which they are in fact not capa­ble of because of a lack of tech­ni­cal log­ic. Alter­na­tive­ly, the legit­i­ma­cy of algo­c­ra­cy might come from charis­mat­ic domin­ion – if you assume that cit­i­zens trust the per­fec­tion of algo­rith­mic infor­ma­tion pro­cess­ing with­out ques­tion (automa­tion bias). In fact, it is not that easy because a pure­ly soci­o­log­i­cal term of legit­i­ma­cy can­not be applied to polit­i­cal and legal sci­ences that eas­i­ly. Actu­al approval isn’t enough. A nor­ma­tive aspect is nec­es­sary if domin­ion is to be legal­ly legit­i­mate; pub­lic recog­ni­tion must be pro­vid­ed by the law. That’s not the case with regard to blind trust in technology.

Your title sug­gests that there is a shift from democ­ra­cy to algo­c­ra­cy. We often hear in the media that democ­ra­cy is being “threat­ened” and indeed research has long shown that democ­ra­cy is in decline. Does algo­rith­mi­sa­tion – as a dis­rup­tive tech­nol­o­gy – endan­ger democ­ra­cy or are there oth­er rea­sons for this decline? Is it meth­ods like the echo cham­bers that shape the opin­ion of a group which form a “dom­i­nant” opin­ion and sift out “insignif­i­cant opinions”?

The fact that there is a con­stant trans­for­ma­tion between democ­ra­cy and algo­c­ra­cy is not new. Both of them are oppo­site ide­al types that don’t real­ly exist in that spe­cif­ic form. In every state, process­es of democ­ra­ti­sa­tion and auto­crati­sa­tion are con­stant­ly tak­ing place.

Cur­rent­ly many coun­tries are being seized by a so-called third wave of auto­crati­sa­tion. As we know democ­ra­cy has already been threat­ened in the past. What’s new about this threat is the vari­ance and simul­tane­ity of those threats. Democ­ra­cy is in the throes of a cri­sis of unprece­dent­ed extent, called poly­cri­sis which is occur­ring with increas­ing fre­quen­cy. At the turn of the mil­len­ni­um, more peo­ple lived in states with dete­ri­o­rat­ing rather than improv­ing demo­c­ra­t­ic con­di­tions. In the year 2020 only 4 % of the world pop­u­la­tion lives in the lat­ter situation.

The rea­sons for this are diverse and often rein­force each oth­er. Democ­ra­cy is endan­gered by glob­al­i­sa­tion and the accom­pa­ny­ing decline of states and par­lia­ments. Bound­aries become blurred, val­ues begin to sway and bond­ing forces that have kept soci­ety togeth­er in its pre­vi­ous shape fade. On top of that we can­not neglect the cli­mate cri­sis that shows demo­c­ra­t­ic states have thus far been inca­pable of act­ing and are seem­ing­ly help­less due to poor attempts at man­ag­ing the sit­u­a­tion and thus the gap in an already frag­ment­ed soci­ety has increased. Algo­rith­mi­sa­tion, the spread of algo­rith­mic appli­ca­tions, adds more fuel to that explo­sive mixture.

In this con­text, I dis­tin­guish between two areas: algo­rith­mi­sa­tion affects the estab­lish­ment and the exer­tion of demo­c­ra­t­ic rule – the estab­lish­ment you allud­ed to in your ques­tion. Democ­ra­cy is the domin­ion of the peo­ple. Peo­ple are the sov­er­eign, from whom all pow­er in a state must come from and to whom it must be attrib­ut­able, accord­ing to Arti­cle 20 (2) GG. That hap­pens by a process of objec­tive for­ma­tion, dur­ing the course of which a col­lec­tive will devel­ops out of the indi­vid­ual opin­ions that even­tu­al­ly become part of the state will. When algo­rithms are imple­ment­ed into that process and lim­it the set­ting of the for­ma­tion of a per­son­al opin­ion in the inter­net (where it hap­pens increas­ing­ly), the rep­u­ta­tion of the emerg­ing demo­c­ra­t­ic sys­tem is tar­nished from the very start. Major­i­ty and minor­i­ty opin­ions are nipped in the bud. The diverse range of opin­ions is both homogenised and polarised.

Democ­ra­cy, how­ev­er, lives from the very plu­ral­i­ty of a con­struc­tive polit­i­cal dis­course. Of course, we can­not just blame any iso­la­tion in imag­i­nary spaces on fil­ter algo­rithms and web track­ing by cook­ies. Humans nat­u­ral­ly tend to focus their atten­tion on infor­ma­tion that match their own view, and to ignore such infor­ma­tion that chal­lenges it (selec­tive expo­sure). Opin­ion-form­ing con­tent is designed to suit per­son­al pref­er­ences as well, but by nat­ur­al self-select­ed per­son­al­i­sa­tion in the echo cham­bers you men­tioned, not by algo­rith­mic pre-select­ed per­son­alised set­tings in fil­ter bub­bles. Tra­di­tion­al media like news­pa­per and tele­vi­sion have to fil­ter their con­tent, too. How­ev­er, the per­son­alised fil­ter­ing in the inter­net exceeds the human-dri­ven selec­tion, for exam­ple, in case of ampli­fi­er-algo­rithms that inten­tion­al­ly present eye-catch­ing news to pro­long the time spent on the website.

Algo­rithms are com­plex and, depend­ing on their char­ac­ter­is­tics, they lack trans­paren­cy and thus trace­abil­i­ty, and they take on – also autonomous­ly – an ever-increas­ing num­ber of human tasks. In the event of their “takeover of pow­er”, are they depen­dent on the legit­i­ma­cy of the peo­ple, mean­ing would they have to be elect­ed, because they can make deci­sions and quick­ly devel­op a basis for deci­sions based on the amount of data? Or are algo­rithms tools and will they remain tech­ni­cal tools that only require legal reg­u­la­tion such as the EU’s AI reg­u­la­tion which spec­i­fies what they are per­mit­ted and under what “con­di­tions”? If so, can the peo­ple devel­op guide­lines for deal­ing with these “intel­li­gent”, pos­si­bly even “more intel­li­gent” algorithms?

As I already men­tioned, when algo­rithms are employed by the state’s exec­u­tive, leg­isla­tive or judi­cial author­i­ty, they affect the exer­tion of demo­c­ra­t­ic rule. The cru­cial point con­cerns demo­c­ra­t­ic legit­i­ma­tion of sov­er­eign deci­sions, jus­ti­fi­ca­tion of the gov­ern­ment act­ing in spe­cif­ic sit­u­a­tions to exert their rule , which is an indis­pens­able con­di­tion for a democracy.

State author­i­ty which stems from the peo­ple must be account­able to them accord­ing to Arti­cle 20 (2) GG – please par­don my rep­e­ti­tion, but the prin­ci­ple of the people’s sov­er­eign­ty can’t be empha­sised often enough. The pre­vail­ing view dif­fer­en­ti­ates between two types of legit­i­ma­tions: 1) Per­son­al legit­i­ma­tion refers to the jus­ti­fi­ca­tion of the per­son decid­ing (Is the deci­sion to be made by this per­son?) while 2) fac­tu­al legit­i­ma­tion con­sid­ers the jus­ti­fi­ca­tion of the deci­sion on its mer­its (Should this deci­sion be made?). Observ­ing a cer­tain lev­el of legit­i­ma­tion is suf­fi­cient so that deficits of the one ele­ment can be bal­anced by the oth­er. A com­plete sub­sti­tu­tion is impossible.

A ques­tion­able issue is already there when we look at the qual­i­fi­ca­tion of the algo­rith­mic out­put as a deci­sion. Even the most intel­li­gent, entire­ly autonomous pro­gram is to be clas­si­fied as belong­ing to Weak AI and there­fore bound to the cod­ed require­ments. At any rate it should be clear to every­one that an algo­rithm can’t be elect­ed by itself because being elect­ed is a per­son­al deci­sion demand­ing an eli­gi­ble subject.

Nev­er­the­less, the out­put of an algo­rithm can have a con­trol­ling effect that moulds its deci­sion­al char­ac­ter and makes it a pow­er­ful instru­ment. It is to be not­ed that ful­ly auto­mat­ed deci­sions are sub­ject to strict lim­i­ta­tions. But semi-auto­mat­ed deci­sions can also have a con­sid­er­able impact on human behav­iour because of their sup­posed unim­peach­able sta­tus (automa­tion bias).

In this case, the account­abil­i­ty of the out­put exerts a pro­found influ­ence on demo­c­ra­t­ic legit­i­ma­tion, so that an algo­rith­mic deci­sion is still con­sid­ered to be part of the people’s will. If this becomes impos­si­ble because of the lack of trans­paren­cy of a “black box algo­rithm “, deficits in legit­i­ma­tion arise. They can affect per­son­al legit­i­ma­tion, when a deci­sion is attrib­uted to an offi­cial or insti­tu­tion in charge in a way that is unclear, or fac­tu­al legit­i­ma­tion, when the algo­rith­mic deci­sion-mak­ing cor­re­sponds not to the law, but to its own legal terms which are in some way hard to follow.

Whether and how far such legit­i­ma­tion deficits can be com­pen­sat­ed for is anoth­er dif­fi­cult ques­tion. Legal con­trol, such as the AI reg­u­la­tion of the Euro­pean Union, is vital for such a (high) risk tech­nol­o­gy in any case. How­ev­er, if AI is not banned com­plete­ly, the effec­tive­ness of the reg­u­la­tors depends on its tech­no­log­i­cal capa­bil­i­ty to x‑ray the black box. If sci­ence fails to do so, all require­ments on ver­i­fi­ca­tion, doc­u­men­ta­tion and soft­ware qual­i­ty assur­ance must remain paper tigers.

Many com­pa­nies try to direct­ly reach end con­sumers to increase sales. Algo­rithms also enable the state to get in “direct” con­tact with cit­i­zens and to take their wish­es into account – the best require­ment for a democ­ra­cy. Are algo­rith­mi­cal­ly gen­er­at­ed fil­ter bub­bles there­fore nec­es­sary for con­sen­sus build­ing, and, if so, why?

You are think­ing about algo­rith­mi­cal­ly dri­ven sup­port of direct demo­c­ra­t­ic ele­ments, I sup­pose. To be hon­est, I doubt that a direct democ­ra­cy real­ly is the “best“ form of democ­ra­cy. It sure­ly has its ben­e­fits as seen in Switzer­land or the ancient Attic democ­ra­cy which are often upheld as shin­ing exam­ples of this form of democ­ra­cy. But the good per­for­mance of those sys­tems is caused by his­toric, ter­ri­to­r­i­al or struc­tur­al fac­tors which aren’t found every­where, and cer­tain­ly not in Germany.

On the oth­er hand, direct demo­c­ra­t­ic ele­ments can help enhance imple­ment­ing the prin­ci­ple of the people’s sov­er­eign­ty qual­i­ta­tive­ly in a rep­re­sen­ta­tive democracy.

A delib­er­a­tive democ­ra­cy, which focus­es on pub­lic delib­er­a­tion in order to inte­grate cit­i­zens in deci­sion-mak­ing, is a good exam­ple for our dis­cus­sion. The more con­vinc­ing argu­ment in terms of its con­tent and not (only) the for­mal major­i­ty should deter­mine a deci­sion because by con­sid­er­ing all argu­ments an agree­ment on the “best“ option can then be reached – at least the­o­ret­i­cal­ly. Think­ing one step beyond, a delib­er­a­tive process could gain a lot from algo­rith­mic appli­ca­tions, for exam­ple, online fora or bots that eval­u­ate and process the range of opinions.

But I wouldn’t claim that fil­ter bub­bles are nec­es­sary for reach­ing a con­sen­sus. First of all, there is a dif­fer­ence between pri­vate com­pa­nies that fol­low eco­nom­ic cal­cu­la­tions and make a pre-selec­tion from the infor­ma­tion pre­sent­ed and non-com­mer­cial insti­tu­tions which put all the infor­ma­tion togeth­er for delib­er­a­tive pur­pos­es. Sec­ond­ly, any con­sen­sus depends on a giv­en quo­rum. Assum­ing una­nim­i­ty, a high­ly delib­er­a­tive pro­ce­dure might be suit­able to achieve and not to block a demo­c­ra­t­ic process.

When a sim­ple or qual­i­fied major­i­ty is suf­fi­cient to make a deci­sion, such as in the Ger­man legal sys­tem, it would be inap­pro­pri­ate to try and con­ceive of a deci­sion process where all par­tic­i­pants must share the result. An estab­lished rep­re­sen­ta­tive democ­ra­cy accepts that there will always be a minor­i­ty with a dif­fer­ent view and pro­vides rights and priv­i­leges for its pro­tec­tion. Thus, a mas­sive exten­sion of direct demo­c­ra­t­ic ele­ments dri­ven by algo­rithms is not real­ly nec­es­sary at this time.

We are cur­rent­ly con­front­ed with many crises; you your­self have used the key­word “poly­cri­sis”. Let’s look at the envi­ron­men­tal cri­sis. Some claim that cap­i­tal­ism will shrink at this time due to a lack of fos­sil fuels so that we can­not deal with the envi­ron­men­tal cri­sis with­out chaos ensu­ing. They even go so far as to assert that a demo­c­ra­t­i­cal­ly planned econ­o­my is required for the dis­tri­b­u­tion and rationing of scarce resources (e.g. drink­ing water, ener­gy). Are algo­rithms suit­able for a “demo­c­ra­t­i­cal­ly” planned econ­o­my because they can over­see and analyse the com­plex sit­u­a­tions more quick­ly and per­haps make “right/fair” deci­sions regard­ing dis­tri­b­u­tion and rationing?

The econ­o­my has to be redesigned in order to suc­cess­ful­ly deal with cli­mate change. Whether that means reject­ing cap­i­tal­ism, I can­not say as I am no economist.

One thing is cer­tain – if we want to stick to cap­i­tal­ism, we must reform it, at least in the places where it hin­ders cli­mate pro­tec­tion and cop­ing with its con­se­quences. Espe­cial­ly in this con­text, (more or less) intel­li­gent, fast, (seem­ing­ly) neu­tral and fault­less algo­rithms seem to an obvi­ous choice. In mod­er­ate forms they are already being used for eco­log­i­cal pur­pos­es, for exam­ple in “Smart Farm­ing“ in agri­cul­ture, when field robots are taught via machine learn­ing where to weed, fer­tilise or sow. On a larg­er scale, parcels and machines are inter­con­nect­ed and exter­nal data sources about weath­er and topol­o­gy are inte­grat­ed to opti­mise sow­ing and har­vest­ing strate­gies. With “Smart Meter­ing“ in ener­gy indus­tries intel­li­gent mea­sure­ment sys­tems con­tribute to coor­di­nat­ing the gen­er­a­tion and con­sump­tion of ener­gy and there­fore to improv­ing ener­gy management.

A far more rad­i­cal step would be a trans­for­ma­tion of the free mar­ket econ­o­my into a planned econ­o­my, where the scarci­ty in resources would force state inter­ven­tion. This involve­ment could start with local bot­tle­necks, and then extend to region­al, nation­al or even glob­al levels.

If this path were cho­sen by pol­i­tics, I would sup­port using algo­rithms to solve dis­tri­b­u­tion issues. These depend on more vari­ables than one per­son or even whole task forces could ever con­sid­er, let alone eval­u­ate cor­rect­ly. Big Data, gigan­tic data stocks lying more or less organ­ised in data ware­hous­es, data lakes or data swamps, and their pro­cess­ing meth­ods might pro­vide the nec­es­sary due diligence.

Oth­ers rely on human inno­v­a­tive pow­er to over­come crises. Will algo­rithms find faster and bet­ter solu­tions than humans and are they there­fore the solu­tion to being cli­mate-neu­tral by 2045 in accor­dance with legal require­ments, despite a pos­si­bly long mar­ket launch due to test­ing, stud­ies etc.?

I think we have to aban­don the idea that we are able to cope with the many crises which we caused by only using our nat­ur­al cog­ni­tive resources. By using algo­rithms to con­nect, com­mu­ni­cate and trans­act glob­al­ly, mov­ing huge amounts of data while doing so – in 2020 the esti­mat­ed glob­al data vol­ume came to 50 zettabytes – in frac­tions of a sec­ond, we have opened the prover­bial Pandora’s box and relin­quished our hold on technology.

Dig­i­tal dimen­sions know nei­ther space nor time, at least not as we per­ceive it, and have devel­oped process­es we can’t inter­vene with, even though we per­suade our­selves into think­ing so. We live in a VUCA world, the char­ac­ter­is­tics of which, such as volatil­i­ty, uncer­tain­ty, com­plex­i­ty and ambi­gu­i­ty, are reflect­ed in every major cri­sis of our time. Instru­men­tal­is­ing algo­rithms to nav­i­gate this world is an ardu­ous task, but I don’t see any alter­na­tive. Even if we don’t reach cli­mate neu­tral­i­ty in 2045 by using algo­rithms because reforms and inno­va­tion in pol­i­tics and sci­ence are being blocked, with­out them we are sure­ly nev­er going to achieve that goal.

Mr Hans-Jür­gen Jakobs said that the new “ism” was monop­oly. Prof Dr Schupp advo­cates an uncon­di­tion­al basic income, while oth­ers advo­cate the shrink­ing of cap­i­tal­ism, and you are inves­ti­gat­ing whether algo­c­ra­cy will soon gain strength. What will our soci­ety look like in the future? Is an algo­c­ra­cy the desired sys­tem change that will guar­an­tee suc­cess in over­com­ing these var­i­ous crises?

I am afraid nei­ther Mr Jakobs, Prof Dr Schupp nor I can offer a fool-proof recipe for suc­cess here. There is not just one switch that changes every­thing for the bet­ter. The thought is appeal­ing, but utopi­an, unfortunately.

Just as the poly­cri­sis con­sists of sev­er­al sin­gle crises – cli­mate cri­sis, democ­ra­cy cri­sis (because of algo­crat­ic ten­den­cies among oth­er things), eco­nom­ic cri­sis, secu­ri­ty cri­sis – cop­ing with the poly­cri­sis has to pur­sue a mul­ti­modal strat­e­gy. Iso­lat­ed attempts to res­cue the cli­mate, to rec­on­cile all nations for world peace or to restore the integri­ty of democ­ra­cy, are the first steps in the right direc­tion. But with­out going hand in hand, they won’t achieve a sus­tain­able effect.

Ms Borchert, thank you for shar­ing your insights on the dif­fer­ent aspects regard­ing algocracy.

Thank you, Dr Cal­daro­la, and I look for­ward to read­ing your upcom­ing inter­views with recog­nised experts, delv­ing even deep­er into this fas­ci­nat­ing topic.