How do we want to shape our future dig­i­tal space?

The dig­i­tal space offers us world­wide, bor­der­less, paper­less, imme­di­ate reach, com­merce and inter­ac­tion with oth­er peo­ple, cul­tures, mar­kets and opin­ions. But the dig­i­tal space has also down­sides like fakes, con­trol and dis­tor­tion. The dig­i­tal space con­sists of dif­fer­ent dig­i­tal tools that we can shape and design accord­ing to our needs and val­ues. Do we have a con­crete aim in what to achieve with the dig­i­tal space, how it should work, and what rules are appro­pri­ate to guar­an­tee free­dom, fair­ness and trust?

In the lat­est of her Duet inter­views, Dr Cal­daro­la, edi­tor of Data Ware­house as well as co-author of Big Data and Law, and Dr Hal­lensleben dis­cuss pos­si­ble designs for our dig­i­tal space.

Dr Hal­lensleben, what oppor­tu­ni­ties does the dig­i­tal space offer us?

Dr Hal­lensleben: Before delv­ing into details, I believe there is a fun­da­men­tal oppor­tu­ni­ty that often gets overlooked.

The dig­i­tal space is some­thing we fun­da­men­tal­ly shape and cre­ate as human­i­ty and soci­ety. We pos­sess an immense degree of con­trol over its struc­ture and function.

I argue that this oppor­tu­ni­ty is neglect­ed because a nar­ra­tive of inevitabil­i­ty often sur­rounds AI and oth­er dig­i­tal top­ics. For instance, phras­es like ‘tech­ni­cal progress is mov­ing in a cer­tain direc­tion,’ ‘a break­through has occurred,’ ‘AI is pro­gress­ing rapid­ly,’ or ‘dig­i­tal plat­forms now enable fric­tion­less trade in mobil­i­ty or oth­er sec­tors’ are com­mon. The implic­it assump­tion is that these devel­op­ments sim­ply hap­pen, and every­one must react or con­form, as if they were nat­ur­al phe­nom­e­na like weath­er or earth­quakes that one can only adapt to or work around.

In real­i­ty, devel­op­ments in the dig­i­tal space are the result of human choic­es, some indi­vid­ual, some sys­temic made at nation­al or large cor­po­rate lev­els. Ulti­mate­ly, these are our choic­es, and they bring with them immense opportunities.

I wish we would make far greater use of these oppor­tu­ni­ties and design pos­si­bil­i­ties. My desire is for us to more fre­quent­ly ask our­selves what kind of dig­i­tal space we tru­ly want. We are per­mit­ted to dream when it comes to this ques­tion. Chances are, it is fea­si­ble if a suf­fi­cient num­ber of peo­ple desire it, and most of it is achiev­able with­in the dig­i­tal space. This implies that an opti­mistic sense of agency is lack­ing, as is the courage to embrace that agency in many areas. I wish we would assert our oppor­tu­ni­ties to design the dig­i­tal space by select­ing and shap­ing our desired outcomes.

I am remind­ed of Vic­tor Hugo’s thoughts con­cern­ing ideas:

‘Noth­ing is as pow­er­ful as an idea, whose time has come’

With regard to our con­ver­sa­tion, I would like to add:

‘The time has come, what is the idea?’

Every medal has two sides. What risks do we face in the dig­i­tal space?

One fun­da­men­tal risk is the scal­a­bil­i­ty of neg­a­tive out­comes. If a dig­i­tal busi­ness mod­el emerges that is not con­struc­tive in the long run or has adverse side effects – for exam­ple, a busi­ness mod­el offer­ing con­tent or a ser­vice in exchange for data, dri­ven by auc­tion­ing user atten­tion for adver­tis­ing – this lim­it­ed, data-dri­ven busi­ness mod­el can eas­i­ly become glob­al. If some­one, some­where, con­cludes that this is a prof­itable approach, it tends to spread glob­al­ly unless there are obsta­cles, deci­sions, or reg­u­la­tions to pre­vent it. If fea­si­ble, such mod­els prop­a­gate very quick­ly across the glob­al dig­i­tal space, and most mod­els that have emerged this way car­ry inher­ent risks.

The sec­ond risk I want to high­light is the diver­gence of agency and account­abil­i­ty. In the phys­i­cal world, if I per­form an action, I can typ­i­cal­ly be held account­able for it.  In the dig­i­tal space, this is less clear-cut. As an exam­ple, the own­ers of large social media plat­form have plen­ty of agency in shap­ing their offer­ings but evade account­abil­i­ty to a sig­nif­i­cant degree…This is a high­ly rel­e­vant risk.

A third risk in the dig­i­tal space is the ease with which pow­er and con­trol can con­cen­trate. The dig­i­tal space lends itself to the for­ma­tion of monop­o­lies sim­ply because the cost of scal­ing from 10% to 100% of a user base, or even from 1% to 10%, is typ­i­cal­ly very small. In the phys­i­cal world, scal­ing to a monop­oly is far more dif­fi­cult, espe­cial­ly to a glob­al monop­oly, which is almost impos­si­ble. The instru­ments orig­i­nal­ly devel­oped for the phys­i­cal space to pre­vent monop­o­lies and dis­man­tle car­tels become more cru­cial and hard­er to enforce in the dig­i­tal space, giv­en the transna­tion­al or multi­na­tion­al nature of its enti­ties and ser­vices. I believe one of the main risks is our assump­tion that the dig­i­tal space will some­how achieve a hap­py mar­ket equi­lib­ri­um on its own. In real­i­ty, it is far more frag­ile than the phys­i­cal econ­o­my and requires much stronger atten­tion to main­tain that sweet spot where healthy com­pe­ti­tion thrives while pre­vent­ing the emer­gence of monop­o­lies. These are some of the inher­ent risks.

I also want to high­light anoth­er risk that is par­tic­u­lar­ly impor­tant to me, which relates to the recent capa­bil­i­ties of arti­fi­cial intel­li­gence. By ‘recent,’ I refer not just to phe­nom­e­na as new as Chat­G­PT, but to devel­op­ments dat­ing back to 2018 when deep­fakes first emerged. This marked a moment when it became pos­si­ble not only to fake every type of con­tent but also to fake peo­ple. Now, a suite of AI tech­nolo­gies is avail­able that allows us to man­u­fac­ture con­vinc­ing­ly decep­tive ‘human beings’ in the dig­i­tal space. The fun­da­men­tal assump­tion that the peo­ple I inter­act with are real is no longer valid.

This is some­thing that can­not hap­pen in the phys­i­cal world; if I meet some­one face-to-face, I know they are a real per­son and a sin­gu­lar enti­ty who can­not simul­ta­ne­ous­ly be anoth­er per­son or pre­tend to be a hun­dred dif­fer­ent peo­ple. It has become the new nor­mal that when I encounter some­one in the dig­i­tal space, I don’t know if they are real. I don’t know if I’m wast­ing my time engag­ing with this ‘per­son’ because I might just be try­ing to con­vince a bot or build a rela­tion­ship with one.

This under­mines a fun­da­men­tal ele­ment of soci­ety: the abil­i­ty to trust inter­ac­tions with anoth­er per­son. I would even go so far as to say that we risk los­ing the dig­i­tal space as a viable place for social­is­ing, con­duct­ing busi­ness, hav­ing dis­cus­sions, and engag­ing in demo­c­ra­t­ic dis­course. When we reach a point where a sig­nif­i­cant per­cent­age of par­tic­i­pants are bots, or must be assumed to be bots, I would dis­en­gage, and we would approach the ‘dead inter­net’ mod­el, a space where bots deceive oth­er bots.

In sum­ma­ry, both the oppor­tu­ni­ties and risks are sub­stan­tial and chal­leng­ing. How can the dig­i­tal space be designed to har­monise these oppor­tu­ni­ties and risks? You advo­cate for ‘dig­i­tal trust’ and ‘dig­i­tal integri­ty.’ What exact­ly do you mean by these terms?

Build­ing on my pre­vi­ous respons­es, we must acknowl­edge that we are not mere­ly fac­ing issues with bots, deep­fakes, or gen­er­a­tive AI. Instead, we con­front a fun­da­men­tal cri­sis of trust with­in the dig­i­tal space. We must address this prob­lem at its foun­da­tion­al lev­el, recog­nis­ing that trust is an essen­tial ele­ment for any func­tion­ing soci­ety. If we lose the abil­i­ty to trust in the dig­i­tal space, it becomes fun­da­men­tal­ly unin­hab­it­able, under­min­ing the great promise it oth­er­wise holds.

Let us define trust in this con­text. I like to illus­trate it with a ques­tion I often ask peo­ple: ‘When were you born?’ Although they typ­i­cal­ly look quite aston­ished and polite­ly respond, ‘Of course, I know when I was born,’ I then ask, ‘How do you know?’ It is a fact that I, as a new­born, was present dur­ing my birth. How­ev­er, at that point in time, I have no rec­ol­lec­tion of the time, place, or cir­cum­stances of my birth. Instead, I trust my par­ents who told me that day was my birth­day. Or I might trust my par­ents’ neigh­bours who told me that on a cer­tain day, my par­ents returned from the hos­pi­tal with me in their arms. Or I trust a per­son in City Hall who hands me a birth cer­tifi­cate with my name and birth­day on it.

The point is, even for sim­ple per­son­al fac­tu­al infor­ma­tion, you must trust oth­er peo­ple. That is why trust is fun­da­men­tal. It is not sole­ly relat­ed to the com­plex­i­ty that aris­es when peo­ple can­not under­stand the ‘black box’ of AI, the data flow of a busi­ness mod­el, or the under­ly­ing tech­nol­o­gy of a dig­i­tal event when access­ing the dig­i­tal space. It is sim­ply con­nect­ed to the fact that there are sur­pris­ing­ly few things we can per­son­al­ly and reli­ably ver­i­fy ourselves.

Soci­ety can­not func­tion with­out trust. This leads us to the ques­tion: How do we jus­ti­fy trust? Or, how do we jus­ti­fy being vul­ner­a­ble? How can we accept not being able to ver­i­fy every­thing, even when com­pli­cat­ed and how do we jus­ti­fy trust­ing cer­tain peo­ple even when we make our­selves vulnerable?

Return­ing to trust in the dig­i­tal space, sys­tems are tech­ni­cal­ly high­ly com­plex, requir­ing us to trust our providers, brows­er man­u­fac­tur­ers, device mak­ers, and so forth. We need enabling tech­nolo­gies to help us jus­ti­fy this trust.

I would like to high­light two aspects that relate to the dig­i­tal integri­ty and trust you men­tioned in your ques­tion. There are three dis­tinct types of trust: (a) trust that the coun­ter­part is indeed a per­son and not a bot, (b) trust that the per­son is hon­est with me, and © the trust or degree of trust in the oth­er person’s judgment.

We need tools to help us nav­i­gate this, giv­en the myr­i­ad peo­ple we encounter online. First, tools that track our trust assess­ments. Trust is high­ly indi­vid­ual; the extent to which I trust some­one dif­fers from how much anoth­er per­son trusts the same indi­vid­ual. Con­verse­ly, know­ing that you trust some­one, and my trust­ing your judg­ment, might lead me to also trust that per­son. How­ev­er, if I lat­er get to know the oth­er per­son myself and con­clude for myself that this per­son is ‘full of air,’ my trust in your judge­ment might decrease. We all main­tain an intu­itive record of how much we trust oth­ers, and this is both indi­vid­ual and dynamic.

In the phys­i­cal world, we man­age to main­tain all three types of these trust rela­tion­ships in our minds because human beings are evo­lu­tion­ar­i­ly wired to han­dle such trusts through ‘gut feelings.’

In the dig­i­tal world, how­ev­er, this is insuf­fi­cient. I believe that an enabling tech­nol­o­gy is need­ed to facil­i­tate the record-keep­ing and track­ing of our indi­vid­ual trust assess­ments, and to allow us to share them with­in our social circles.

Sec­ond­ly, and even more impor­tant­ly, we need an enabling tech­nol­o­gy for pseu­do­ny­mous iden­ti­ties. I can­not dis­cuss trust with­out dis­cussing iden­ti­ty, because trust always involves anoth­er per­son build­ing a track record in my eyes. I can­not make any trust assess­ment based on a sin­gle data point; I need mul­ti­ple data points, either from my own exchanges and expe­ri­ences with a per­son or from the expe­ri­ences of oth­ers. It is always a series of data points, and I can only cor­re­late them if the oth­er per­son main­tains a con­sis­tent iden­ti­ty, not vary­ing iden­ti­ties. I don’t need to know their real iden­ti­ty, but I do need to recog­nise a per­son from one inter­ac­tion to the next. We need pri­va­cy-pro­tect­ing, pseu­do­ny­mous iden­ti­ties that are sin­gu­lar per con­text. We do not need a sin­gle pseu­do­nym to fol­low a per­son across the entire inter­net; that would be unwise, as it cre­ates many vul­ner­a­bil­i­ties that could allow us to trace back to the real iden­ti­ty. That is not what we want. Instead, we need an iden­ti­ty per con­text, per plat­form, per forum, or what­ev­er else.

There­fore, when dis­cussing with some­one in an online forum, I want to be sure (a) that the per­son is real and not a bot, and (b) that this per­son has only one account on this forum (or per­haps two or three, but cer­tain­ly not hun­dreds or thou­sands of accounts).

Thus, that per­son needs to have an iden­ti­ty on the forum. No one needs to know their real iden­ti­ty, but a pseu­do­ny­mous iden­ti­ty is nec­es­sary. The inter­est­ing aspect is that while this might sound ambi­tious, the pro­to­cols and even the tech­nol­o­gy already exist. This func­tion­al­i­ty is even inte­grat­ed into ID cards in Ger­many and is part­ly described in the Euro­pean eIDAS Reg­u­la­tion, though it is not cur­rent­ly being utilised.

On one hand, one could argue, ‘Oh, that is frus­trat­ing.’ But on the oth­er hand, one could also argue, ‘Oh, great, we have a sig­nif­i­cant prob­lem to solve, and we already pos­sess the tech­nol­o­gy to do so. It’s mere­ly a ‘roll-out problem.’

Do we still trust peo­ple and infor­ma­tion in the dig­i­tal space? AI, fake news, and sim­i­lar phe­nom­e­na hijack atten­tion, dis­tort demo­c­ra­t­ic dis­course, and com­pli­cate trans­par­ent mar­kets. How would dig­i­tal trust and dig­i­tal integri­ty need to be imple­ment­ed con­crete­ly to effec­tive­ly and sus­tain­ably pro­tect the dig­i­tal space against attacks, dis­tor­tions, and oth­er threats?

This builds on my pre­vi­ous answer. The key term is avoid­ing dis­tor­tion. To phrase it pos­i­tive­ly: the key is to cre­ate fair­ness. We are accus­tomed to the notion or prin­ci­ple of democ­ra­cy, which is wide­ly accept­ed to ensure fair­ness of rep­re­sen­ta­tion: ‘One per­son, one vote.’

We now need a relat­ed prin­ci­ple, which I would call ‘One per­son, one voice,’ so that the atten­tion, or the chance of gain­ing atten­tion, is also fair­ly distributed.

You do not have an inher­ent right to be lis­tened to, as oth­ers might con­clude they are not inter­est­ed or that a per­son is ‘sparkling lots of bull­shit.’ How­ev­er, every­one needs to have a fair chance to be heard. If peo­ple deploy armies of bots, that fun­da­men­tal­ly destroys fair­ness. If I oper­ate thou­sands of bots and you mere­ly post as your­self, my opin­ions will receive a thou­sand times more atten­tion and a thou­sand times more oppor­tu­ni­ty to be heard, which is fun­da­men­tal­ly unfair. There­fore, we tru­ly need fair­ness of attention.

To me, that is a far bet­ter approach than attempt­ing to fil­ter con­tent to label cer­tain things as fake or untrue. Address­ing val­ues is a much hard­er task because it requires sig­nif­i­cant mod­er­a­tion effort and robust gov­er­nance struc­tures, both city-wide and board​-lev​el​.at con­sti­tutes true or not true.

A good start­ing point is to make sure that every­one that the same chance to be heard. That is by far the most impor­tant thing. I pre­fer this approach over fil­ter­ing because it leaves oppor­tu­ni­ty for evo­lu­tion.  Many of the opin­ions that are main­stream today – things like equal­i­ty between men and women, like con­ti­nen­tal drift in geol­o­gy – start­ed as fringe opin­ions against the received wis­dom at the time. And yet over time these fringe opin­ions became main­stream. I think it would be very opti­mistic to assume that all of today’s main­stream opin­ions are per­fect and the end of the evo­lu­tion and that is all how it should be. That is a strange assump­tion. We have to assume that some of the opin­ions that appear fringe today will be gain trac­tion over time and will become main­stream and accept­ed wis­dom in the future, even though this prob­a­bly only applies to a small pro­por­tion of these fringe opin­ions. We need to pre­serve the oppor­tu­ni­ty for soci­ety and human­i­ty to evolve in that way.

Infor­ma­tion­al self-deter­mi­na­tion, demo­c­ra­t­ic dis­course, and fair­ness are val­ues not uni­ver­sal­ly shared across the globe. How do we achieve har­mon­i­sa­tion and uni­form glob­al rules for sta­bil­i­ty and sustainability?

The short and flip­pant answer is cer­tain­ly: ‘We don’t.’ I believe it is very healthy for the world to have a com­pe­ti­tion regard­ing the best sys­tem to live in. what is most eco­nom­i­cal­ly and social­ly suit­able. Through this com­pe­ti­tion, we might find some sys­tems to be wrong or dis­taste­ful, and oth­ers appro­pri­ate or man­age­able. It is per­fect­ly accept­able to pro­mote the sys­tem we believe is best and to sup­port peo­ple world­wide who are work­ing towards it.

How­ev­er, we can­not assume that the entire world will migrate to the sys­tem we con­sid­er best, even in our glob­alised world. To me, the best approach is to max­imise the ben­e­fits of our sys­tem and make those ben­e­fits vis­i­ble, effec­tive­ly say­ing: ‘Look at how we live, our polit­i­cal sys­tem, our demo­c­ra­t­ic sys­tem, soci­ety, human­i­ty. See how effec­tive­ly we cre­ate pros­per­i­ty, ensure par­tic­i­pa­tion for every­one, accom­plish deci­sion-mak­ing process­es, and achieve hap­pi­ness.’ The most effec­tive way to suc­ceed in glob­al com­pe­ti­tion is to put our val­ues into prac­tice and demon­strate what life is like with­in these val­ues. This serves as a ‘walk­ing adver­tis­ing bill­board’ for our sys­tem, val­ues, and per­cep­tions. Then, oth­er soci­eties might fol­low, or they might fal­ter. Both out­comes are accept­able and must be acknowledged.

There is one excep­tion to the prin­ci­ple of ‘live and let live and make the sys­tem attrac­tive and work well’: sit­u­a­tions where we lack clear ter­ri­to­r­i­al bound­aries. This cov­ers sev­er­al aspects:

First­ly, it encom­pass­es the entire aspect of sus­tain­abil­i­ty and cli­mate change, because we can­not allow, for exam­ple, Amer­i­cans to ‘burn their coun­try with oil’ if it harms the entire plan­et. Sim­i­lar­ly, we can­not per­mit inhab­i­tants in Brazil to destroy their forests with­out con­cern, as it dam­ages essen­tial ele­ments of the plan­et on which we all live.

Sec­ond it cov­ers the aspect of glob­al plat­forms which are under the con­trol of glob­al com­pa­nies that have size and rev­enues that are big­ger than the mon­ey of many coun­tries. Here again we can­not say “Live and let live”. Here we need to pro­tect our nation­al ter­ri­to­ries by reg­u­lat­ing the mar­ket through laws like the Euro­pean Dig­i­tal Mar­ket Act or the Euro­pean Dig­i­tal Ser­vice Act. These are rea­son­ably potent tools if enforced prop­er­ly. It is cer­tain­ly messy to draw lines of juris­dic­tion in the dig­i­tal space. There is no way around it as long as the gen­er­al ideas of what the dig­i­tal space should look like start to diverge more and more. The chal­lenge is to avoid the Chi­nese or Russ­ian approach­es where they have big fire walls and shut off the inter­net for its cit­i­zens by cre­at­ing lit­tle islands. That does not cor­re­spond to our Euro­pean mind­set. The inter­ac­tion with the rest of the world is for the Euro­peans impor­tant because we want to con­tin­ue to inter­act, to be chal­lenged, to be trans­par­ent. The task is to have bor­ders and bound­aries in some respect but not in others.

AI assur­ance and qual­i­ty are also among your key inter­ests. Does this con­cept obscure the require­ments of the Euro­pean AI Reg­u­la­tion? What is the rela­tion­ship between AI assur­ance, dig­i­tal trust, and dig­i­tal integrity?

The approach to AI assur­ance, par­tic­u­lar­ly the assur­ance of AI qual­i­ty, is built upon the Euro­pean AI Act. The AI Act pri­mar­i­ly focus­es on mit­i­gat­ing risks and estab­lish­es only a min­i­mum thresh­old that must be met. In this sense, it is bina­ry: either you com­ply, or you do not. This con­sti­tutes the base­line. How­ev­er, it is not some­thing that will cre­ate a func­tion­ing, vibrant, and com­pet­i­tive mar­ket. For exam­ple, if I exam­ine the mar­ket for chat­bot or face recog­ni­tion sys­tems, I can assume that any prod­uct I find is com­pli­ant with the law; oth­er­wise, the ven­dor would be on ‘thin ice.’ Com­pli­ance does­n’t help me decide which prod­uct to buy. Sim­i­lar­ly, from the ven­dor or man­u­fac­tur­er’s per­spec­tive, com­pli­ance with the EU AI Act can­not be used as a sales argu­ment, nor does it help them address dif­fer­ent mar­ket segments.

So, the ques­tion is, what do we actu­al­ly need for a func­tion­ing and trans­par­ent mar­ket? I like to com­pare it with a very com­mon and estab­lished exam­ple: pur­chas­ing a car. When I select a car, I con­sid­er var­i­ous met­rics (fuel con­sump­tion, boot space, crash test rat­ings, stop­ping dis­tances, top speed, etc.). Depend­ing on my use case and what I need the car for, I pri­ori­tise dif­fer­ent aspects. It is cru­cial that I have met­rics to pri­ori­tise and make an informed deci­sion. This is how a trans­par­ent, com­pet­i­tive, and func­tion­ing mar­ket emerges, one that fos­ters innovation.

That is what we should also desire for the AI mar­ket. As long as the sole cri­te­ri­on is com­pli­ance, there will be no com­pe­ti­tion and, con­se­quent­ly, no mar­ket dri­ve for inno­va­tion. There­fore, we need to ele­vate the descrip­tion of rel­e­vant char­ac­ter­is­tics of AI sys­tems (qual­i­ty, per­for­mance, etc.) in a non-bina­ry man­ner. It is insuf­fi­cient to have met­rics like ‘secure/insecure’ or ‘multilingual/monolingual chat­bot.’ Mul­ti­ple lev­els of secu­ri­ty, lin­guis­tic capa­bil­i­ty, and oth­er fac­tors are nec­es­sary. Then, depend­ing on my use case, I can make deci­sions accord­ing to my spe­cif­ic needs.

These types of use-case-spe­cif­ic, mul­ti-lev­el met­rics are nec­es­sary to accu­rate­ly express a sys­tem’s char­ac­ter­is­tics. Only then will we have the oppor­tu­ni­ty or rather, the con­se­quence of a trans­par­ent, com­pet­i­tive, and inno­v­a­tive mar­ket that allows for trade-offs. I view qual­i­ty assur­ance for AI as build­ing upon com­pli­ance with cer­tain risk-mit­i­gat­ing requirements.

Dr Hal­lensleben, thank you for shar­ing your insights on the oppor­tu­ni­ties and risks of the dig­i­tal space and what shape it might need.

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