Sax, Ma ijn; Wang, Hao
A icle
F om h ea o oppo uni y: Gaming he algo i hmic
sys em as a se ice
In e ne Policy Re iew
P o ided in Coope a ion wi h:
Alexande on Humbold Ins i u e o In e ne and Socie y (HIIG), Be lin
Sugges ed Ci a ion: Sax, Ma ijn; Wang, Hao (2025) : F om h ea o oppo uni y: Gaming he
algo i hmic sys em as a se ice, In e ne Policy Re iew, ISSN 2197-6775, Alexande on Humbold
Ins i u e o In e ne and Socie y, Be lin, Vol. 14, Iss. 2, pp. 1-33,
h ps://doi.o g/10.14763/2025.2.2007
This Ve sion is a ailable a :
h ps://hdl.handle.ne /10419/321975
S anda d-Nu zungsbedingungen:
Die Dokumen e au EconS o dü en zu eigenen wissenscha lichen
Zwecken und zum P i a geb auch gespeiche und kopie we den.
Sie dü en die Dokumen e nich ü ö en liche ode komme zielle
Zwecke e iel äl igen, ö en lich auss ellen, ö en lich zugänglich
machen, e eiben ode ande wei ig nu zen.
So e n die Ve asse die Dokumen e un e Open-Con en -Lizenzen
(insbesonde e CC-Lizenzen) zu Ve ügung ges ell haben soll en,
gel en abweichend on diesen Nu zungsbedingungen die in de do
genann en Lizenz gewäh en Nu zungs ech e.
Te ms o use:
Documen s in EconS o may be sa ed and copied o you pe sonal
and schola ly pu poses.
You a e no o copy documen s o public o comme cial pu poses, o
exhibi he documen s publicly, o make hem publicly a ailable on he
in e ne , o o dis ibu e o o he wise use he documen s in public.
I he documen s ha e been made a ailable unde an Open Con en
Licence (especially C ea i e Commons Licences), you may exe cise
u he usage igh s as speci ied in he indica ed licence.
h ps://c ea i ecommons.o g/licenses/by/3.0/de/deed.en
Volume 14 |
F om h ea o oppo uni y: Gaming he
algo i hmic sys em as a se ice
Ma ijn Sax Uni e si y o Ams e dam
Hao Wang Wageningen Uni e si y & Resea ch
DOI: h ps://doi.o g/10.14763/2025.2.2007
Published: 6 May 2025
Recei ed: 12 Janua y 2024 Accep ed: 27 Ma ch 2024
Funding: This esea ch was unded by a g an om he Resea ch P io i y A ea
‘Human(e) AI’ o he Uni e si y o Ams e dam.
Compe ing In e es s: The au ho has decla ed ha no compe ing in e es s exis ha
ha e in luenced he ex .
Licence: This is an open-access a icle dis ibu ed unde he e ms o he C ea i e
Commons A ibu ion 3.0 License (Ge many) which pe mi s un es ic ed use,
dis ibu ion, and ep oduc ion in any medium, p o ided he o iginal wo k is p ope ly
ci ed. h ps://c ea i ecommons.o g/licenses/by/3.0/de/deed.en
Copy igh emains wi h he au ho (s).
Ci a ion: Sax, M., & Wang, H. (2025). F om h ea o oppo uni y: Gaming he
algo i hmic sys em as a se ice. In e ne Policy Re iew, 14(2). h ps://doi.o g/10.14763/
2025.2.2007
Keywo ds: Gaming, Algo i hmic anspa ency, E hics, Manipula ion, Algo i hmic
decision-making
Abs ac : Gaming he sys em – i.e., s a egic a emp s o manipula e he inpu (s) o , o one’s
in e ac ions wi h, an algo i hmic sys em o y o secu e a be e ou come han in ended by he
sys em’s design – is commonly po ayed as a h ea o online pla o ms and se ices. Tech
companies o en use his gaming conce n o jus i y hei eluc ance o p o ide algo i hmic
anspa ency. In his pape , howe e , we will explo e a new business model in he digi al economy
we call gaming- he-sys em-as-a-se ice (GaaS). In his model, anspa ency p omises a e w apped
in o an assis ed gaming se ice and sold as a p emium ea u e. This way, he alleged isk o
anspa ency – gaming he sys em – is u ned in o a mone isa ion ea u e o se ice p o ide s. As
such, GaaS is a ypical example o how ech companies can a emp o u n egula o y p essu es
(e.g., o p o ide mo e insigh in o how i s algo i hmic cu a ion and ecommenda ion sys ems wo k)
in o a comme cial oppo uni y. To begin o e hink ou no ma i e and egula o y app oaches o he
in e ace o anspa ency and gaming, we pe o m a i s explo a ion o se e al po en ial challenges
posed by his new business model. Fi s , GaaS is en wined wi h an incen i e s uc u e ha is
hos ile o consume s and exploi a i e in na u e. Second, GaaS is essen ially a pay- o-win ea u e,
aising ques ions o equali y and ai ness. Thi d, he commodi ica ion o anspa ency h ough GaaS
can ‘ ain ’ and e ode anspa ency as an impo an democ a ic alue.
Issue 2
In oduc ion
Call i a pe manen mud w es ling ma ch, o an e e nal game o ug o wa : law-
make s demanding mo e anspa ency and online pla o ms and se ice p o ide s
in he digi al economy pushing back agains hose demands.1Lawmake s o en ely
on anspa ency obliga ions as a co e ea u e o hei s a egy o ame he powe
o digi al echnology companies (Diakopoulos, 2020). Fo example, he ecen Eu-
opean Union’s legisla i e agenda aimed a he echnology sec o – he Digi al
Se ices Ac (DSA), he Digi al Ma ke s Ac (DMA), and he A i icial In elligence Ac
(AIA) – con ains se e al anspa ency obliga ions. Companies ypically do no wel-
come (addi ional) anspa ency obliga ions. An o en hea d a gumen agains
anspa ency measu es is he gaming- he-sys em-a gumen : anspa ency o (o en
algo i hmic) co e sys ems used o decision-making, cu a ion, and/o anking can
p o ide use s wi h help ul in o ma ion o game hose sys ems (K oll e al., 2017, p.
639, Bambaue & Za ksy, 2018: 15). In his con ex , ‘gaming’ e e s o s a egic a -
emp s o manipula e he inpu (s) o , o one’s in e ac ions wi h, an algo i hmic
sys em o y o secu e a be e ou come han in ended by he sys em’s design. Pe-
e e al. (2019) poin ou ha online pla o ms and se ices “ ou inely denig a e
hese ac i i ies as sys em-gaming o manipula ion” o signal ha such p ac ices
a e unde s ood as p oblema ic unde mining o he ‘p ope ’ unc ioning o he pla -
o ms o se ices (p. 1). Following he gaming- he-sys em-a gumen , anspa ency
obliga ions a e hus mainly po ayed as isks o he business ope a ions o pla -
o m and se ice p o ide s in he digi al economy.
The gaming a gumen agains anspa ency may, howe e , need o be nuanced in
answe o new comme cial s a egies in he digi al economy. In his space whe e
anspa ency obliga ions a e o en po ayed as isks by se ice p o ide s hem-
sel es, one can obse e hose same se ice p o ide s o e ing p emium ea u es
ha look suspiciously simila o – i onically – se ices ha p omise gaming o al-
go i hmic anking and ou comes. In wha ollows, we explo e a new business mod-
el we call gaming- he-sys em-as-a-se ice (GaaS). The co e idea behind GaaS is
ha use s can be cha ged a p emium o help hem game he sys em o a se ice
hey use o secu e (be e chances o ) be e ou comes. Mos in e es ingly, GaaS
and algo i hmic anspa ency a e closely ela ed. The p omise o GaaS can be un-
de s ood as being p edica ed on p omises ela ed o anspa ency. To see why, con-
side he ac ha gaming wo ks bes when one has some deg ee o p i ileged in-
sigh in o how he sys em wo ks. In he con ex o GaaS, howe e , one o en deals
1. Special hanks o Saman ha B adshaw, Camille Gi a d-Chanude , F édé ic Dubois, and F ancesca
Musiani o hei help ul commen s and sugges ions.
2 In e ne Policy Re iew 14(2) | 2025
wi h indi ec anspa ency a bes . The anspa ency is indi ec because he insigh
in o he wo kings o he sys em is pa ial and o e ed h ough a se ice ha is de-
signed and con olled by he sys em owne . So, a he han p o iding comple e
anspa ency2on he en i e sys em which allows use s o exploi ha in o ma ion
o de ise s a egies o gaming hemsel es, he o e ed GaaS se ice p e-s uc u es
he oppo uni ies o gaming. This p e-s uc u ing can ake shape by he gaming
se ice con aining limi ed in o ma ion in combina ion wi h speci ic ools, o by
s uc u ing he se ice a ound gaming wi h he manda o y ‘help’ o employees o
he se ice. Pu simply, he alleged isk o algo i hmic anspa ency – i.e. gaming
he sys em – is ac ually u ned in o a mone isa ion ea u e.
In his pape , we unde s and algo i hmic anspa ency in wo ways. Typically, i is
seen om a echno-cen ic pe spec i e, ocusing on he openness and disclosu e
o how algo i hms wo k om a echnical pe spec i e. This in ol es examining da a
se s, pa ame e s, and models o p o ide insigh s in o hei inne wo kings, ela ing
o concep s like explainable and in e p e able algo i hms (Diakopoulos, 2020).
This ype o anspa ency is o en aimed a expe s, egula o s, o s akeholde s
who need o unde s and he p ecise mechanics o he sys em o secu e ai ness
and accoun abili y. Howe e , as Mi els ad e al. (2019) show, his echnical pe -
spec i e on anspa ency is no he only ype o anspa ency ha ma e s; he e is
also ano he ype ha in ol es e e yday explana ions and is mo e embodied o
humans. Fo his ype, ma hema ically- ounded algo i hms a e ansla ed in o hu-
man e ms (La sson & Hein z, 2020), and anspa ency eme ges om he use ’s
p ac ical expe iences and in e ac ions wi h he pla o m (Ha esamud am e al.,
2023). This is wha Ha esamud am e al. call ‘in e ac ion anspa ency,’ o embod-
ied anspa ency, which e e s o he way anspa ency is expe ienced and unde -
s ood h ough di ec in e ac ion wi h a sys em o pla o m, a he han h ough ab-
s ac o pu ely echnical explana ions (Ha esamud am e al., 2023). This o m o
anspa ency o en c ea es “a nuanced unde s anding” and “ ich, con ex ual, si ua -
ed explana ions” o pla o ms (Ha esamud am e al., 2023, pp. 97-98). Fo exam-
ple, conside how including a pe in he pho os on you p o ile may (seemingly)
imp o e he pe o mance o you p o ile in Tinde ’s ecommenda ion algo i hm.
(Wang, 2023). Al hough Tinde ’s algo i hm emains mys e ious in e ms o i s p e-
2. Full o comple e anspa ency is, o cou se, an (almos ) unin elligible no ion o begin wi h. T ans-
pa ency is always anspa ency o some hing and in he digi al economy i is di icul o see how,
o ins ance, an online pla o m can be comple ely anspa en . E en i a pe son gains access o all
exis ing in e nal documen a ion o said pla o m, he e a e s ill ways in which he pla o m is no
ully anspa en . Top-le el execu i es may ha e made decisions on he basis o in o mal mee ings
ha a e no documen ed and in en ions s a ed in documen s may no co ela e wi h he eal in en-
ions exis ing only in he heads o senio managemen .
3 Sax, Wang
cise echnical ope a ion, use s can s ill gain a con ex ual unde s anding o how i
wo ks h ough hei in e ac ions, o ming nuanced and e e yday insigh s. This em-
bodied anspa ency is i al o use s (o en non-expe s) who may no g asp algo-
i hmic de ails bu ca e deeply abou how pla o ms a ec hem and hei in e ac-
ions.
This pape builds on wo examples – one al eady ope a ional, and one ha was
announced bu la e abandoned – o explo e and explain how hese wo ypes o
algo i hmic anspa ency can be mone ised. The FICO Sco e example demon-
s a es how pla o ms package i s algo i hmic anspa ency as a p emium se ice,
e ealing o c edi consume s how he FICO Sco e wo ks and how hei sco es a e
in luenced by di e en da a poin s and weigh s. Ano he example is Tinde
Concie ge, which was announced (and la e abandoned) as a p emium se ice o-
cusing less on e ealing echnical de ails o algo i hms and da a se s, bu ins ead
p omised a paid- o coaching se ice o help use s o gain nuanced insigh s and
con ex ual unde s anding abou how i s algo i hm wo ks (like how ce ain pho os
o ac ions may enhance hei p o ile isibili y) o hope ully secu e mo e (and ‘be -
e ’) ma ches. In p ac ice, when pla o ms mone ise algo i hmic anspa ency, hey
usually blend hese wo ypes and engage wi h hem o a ying ex en s.
These wo examples se e as a i s explo a ion, bu hey e lec a mo e gene al
phenomenon o mone ising algo i hmic anspa ency as a gaming se ice. The FI-
CO case shows how anspa ency as a gaming se ice is al eady ope a ional, while
he Tinde Concie ge case ells us some hing abou he ypes o ini ia i es big
pla o m p o ide s a e ac i ely conside ing and expe imen ing wi h. We ind i in-
s uc i e o no only discuss examples o se ices ha ha e al eady been imple-
men ed, bu also ake se iously se ices ha a e ‘only’ conside ed by pla o ms. In
a as de eloping pla o m economy, discussions on he e hics and egula ion o
pla o m se ices equi es an ongoing an icipa o y mindse and wi h i a willing-
ness o discuss he ongoing expe imen s o pla o ms; a me ely eac i e mindse
will unde mine ou abili y o de elop u u e-p oo and c ea i e analyses o he
pla o m economy.
As we will discuss in de ail la e , we inc easingly li e in wha Ci on and Pasquale
e e ed o as a “sco ed socie y,” whe e p edic i e algo i hms ank c ucial aspec s o
indi iduals' li es (Ci on & Pasquale, 2014). These algo i hm-d i en anking sys-
ems o en include ewa d and punishmen mechanisms, encou aging use s o op-
imise hei posi ions. Howe e , hei opaque na u e makes i challenging o use s
o unde s and how o imp o e hei ankings, c ea ing an oppo uni y o gaming
se ices o eme ge. Whe he h ough echnical o embodied explana ions, algo-
4 In e ne Policy Re iew 14(2) | 2025
i hmic anspa ency can be mone ised as a p emium o comme cial p oduc , en-
abling use s o game he sys em o hei bene i . This migh sound like a win-win
si ua ion. Use s bene i om gaming as a se ice, while companies balance he isk
o sys em manipula ion wi h addi ional e enue om use s paying om p emium
gaming se ices and con inued engagemen wi h hei se ices. This seemingly
mu ual bene i c ea es a po en ially ad an ageous business model.
Howe e , i hese business models which lip anspa ency and gaming isks upside
down ( om alleged isk o business oppo uni y) indeed s a o ma e ialise and
p oli e a e, we may ha e o e hink ou no ma i e and egula o y app oaches o
hem. In his a icle we make a s a by also explo ing se e al possible challenges
posed by gaming- he-sys em-as-a-se ice. Fi s o all, GaaS is en wined wi h an in-
cen i e s uc u e ha is hos ile o consume s and exploi a i e in na u e. I a se -
ice p o ide wan s o o e GaaS, his in oduces he incen i e o he se ice
p o ide o ac i ely make o keep he wo kings o hei sys ems opaque o unp e-
dic able p ecisely because unce ain y conce ning how he sys em wo ks se es as
a p econdi ion o o e ing GaaS. Second, GaaS is essen ially a pay- o-win ea u e,
aising ques ions o equali y and ai ness. Depending on he con ex whe e GaaS is
in oduced, g an ing ad an ages o hose who can pay a p emium can lead o un-
ai and unjus (ma ke ) ou comes. Thi d and las , he commodi ica ion o ans-
pa ency h ough GaaS can ‘ ain ’ and e ode anspa ency as an impo an democ a-
ic alue. When anspa ency is educed o a commodi y, i no only opens he doo
o manipula ing i s p esen a ion o comme cial in e es s bu also weakens peo-
ple’s mo i a ion o ac i ely engage in c i ical hinking o esis un ai p ac ices in
GaaS.
This a icle is s uc u ed as ollows. In Sec ion 2 we discuss anspa ency as a eg-
ula o y philosophy and unpack he alleged h ea o gaming he sys em esul ing
om anspa ency obliga ions. He e we also discuss how GaaS can posi ion i sel
in he space ypically occupied by discou ses on ( he need o ) anspa ency oblig-
a ions. In Sec ion 3 we u n o he phenomenon o GaaS. We use he examples o
FICO Sco e (al eady implemen ed) and Tinde Concie ge (announced bu no im-
plemen ed) o illus a e co e ea u es o GaaS. In Sec ion 4 we discuss po en ial
challenges posed by GaaS, namely he exploi a i e incen i e s uc u e i in o-
duces o pla o ms and se ice p o ide s, he un ai ness o pay- o-win, and he
e osion o anspa ency as a democ a ic alue.
5 Sax, Wang
T anspa ency as an obliga ion and he alleged h ea o
gaming he sys em
2.1 T anspa ency as a egula o y philosophy
Be o e we u n o GaaS, we i s wan o b ie ly discuss anspa ency and he al-
leged isk o gaming he sys em in he digi al economy. The p inciple o ans-
pa ency has, o cou se, a long his o y in he con ex o policy and egula ion. Al-
eady in 1913 B andeis coined he now amous ph ase ha “sunligh is said o be
he bes o disin ec an s” (B andeis, 1913, p. 1). Su ice i o say, he co e idea ha
anspa ency can se e as an impo an p econdi ion o accoun abili y by making
in o ma ion a ailable ha allows o he inspec ion and e alua ion o en i ies o
ac o s is no a new one. As Ananny and C aw o d (2018, p. 974) summa ise i : “The
implici assump ion behind calls o anspa ency is ha seeing a phenomenon
c ea es oppo uni ies and obliga ions o make i accoun able and hus o change
i ”.
So, he p inciple o anspa ency and i s link o accoun abili y is no a ecen in-
en ion. Wha is a ela i ely new de elopmen hough, is he mo e explici em-
b ace by legisla o s o anspa ency as a p onounced egula o y p inciple in e-
sponse o he inc easing use o , gene ally pu , algo i hmic decision making sys-
ems he unc ioning o which seem opaque o he ou side (Pasquale, 2015;
Lee ssen, 2023). Acco ding o Mo ozo ai e (2024) “ he ecu ing heme in all exam-
ined (p oposed) legal ins umen s [i.e., he DSA, DMA, and AIA] is a s ong empha-
sis on anspa ency obliga ions” (p. 253). One can clea ly see his in he DSA,
whe e anspa ency obliga ions a e a he co e o he legisla i e philosophy be-
hind he ac . The DSA does no only con ain mo e gene al p o isions on, o in-
s ance, anspa ency epo ing obliga ions o p o ide s o in e media y se ices
(A icle 15) and p o ide s o online pla o ms (A icle 24), bu also speci ic ecom-
mende sys em anspa ency obliga ions (A icle 27). A icle 27 manda es online
pla o ms o “se ou in hei e ms and condi ions, in plain and in elligible lan-
guage, he main pa ame e s used in hei ecommende sys ems, as well as any
op ions o he ecipien s o he se ice o modi y o in luence hose main pa ame-
e s”. Because he ecommende engines used by pla o ms a e algo i hmically d i-
en, A icle 27 can igh ly be unde s ood as a p o ision aimed a wha is o en
called algo i hmic anspa ency.
2.2 T anspa ency (claims) as a s a egic ool and he space i
a o ds o GaaS
To unde s and he p ecise ela ion be ween anspa ency and accoun abili y, i can
6 In e ne Policy Re iew 14(2) | 2025
be help ul o dis inguish be ween wha can be called he openness dimension and
he epis emic dimension o anspa ency. The openness dimension e e s o mak-
ing in o ma ion public and inspec able; b inging in o ma ion ou in he open so o
say. Bi kinshaw, o ins ance, (2006) emphasises ha “openness is e y simila o
anspa ency” (p. 190). One can hink o he go e nmen eleasing documen s a e
a eedom o in o ma ion eques , o o a leak such as he Panama Pape s whe e
la ge amoun s o p e iously inaccessible documen a ion suddenly become a ail-
able. The openness o in o ma ion is, howe e , no he same as he comp ehensi-
bili y, explainabili y, and/o usabili y o in o ma ion. Ei he he na u e o he in o -
ma ion (e.g., highly echnical documen a ion) o he amoun o in o ma ion can e-
sul in he di icul ies o uly unde s and o p ocess he now open in o ma ion.
This is why anspa ency is o en hough o ha e an impo an epis emic dimen-
sion as well: “ anspa ency also equi es ex e nal ecep o s capable o p ocessing
he in o ma ion made a ailable” (Heald, 2006, p. 25). When he epis emic dimen-
sion is aken se iously, ue anspa ency also equi es ha public/open in o ma-
ion mus be unde s andable o i s a ge audience.
T anspa ency’s openness dimension and i s epis emic dimension can, o cou se, be
misaligned – no all in o ma ion ha is made public is also unde s andable, and
no e e y hing ha is explained in an unde s andable manne is backed by pub-
licly accessible and e i iable in o ma ion. I is p ecisely in his po en ial o mis-
alignmen , we a gue, ha one can ind he inhe en poli ical and s a egic na u e
o anspa ency (Wang, 2022). T anspa ency is always a o ded by an ac o wi h
pa icula in e es s and incen i es, and o an ac o (o se e al ac o s) wi h pa icu-
la in e es s and incen i es. One ac o ’s anspa ency can be ano he ac o ’s incom-
p ehensibili y. Because ou a gumen ocuses on how he space o anspa ency
discou se as i is shaped by (especially) ecen egula ion can po en ially be mone-
ised by se ice p o ide s wi h gaming- he-sys em-as-a-se ice, we a e mos ly in-
e es ed in how anspa ency claims can be pu o s a egic use. A se ice p o ide
can, o ins ance, claim ha p o iding some addi ional explana ion o how ce ain
unc ions/sys ems wo k coun s as p ac ising anspa ency, in an a emp no o
make ac ual documen a ion public. O , ice e sa, an ac o can make public la ge
amoun s o highly echnical documen a ion which can be e y di icul o make
sense o wi hou addi ional explana o y guidance. We a e, he e o e, no in e es ed
in being he a bi e s o wha de ines eal o ue anspa ency. Fo ou a gumen i
is much mo e impo an o acknowledge how di e en ypes o ac o s end o
make di e en anspa ency claims o achie e di e en – o en sel -se ing – ou -
comes.
7 Sax, Wang
2.3 P oxies and gaming
I is p ecisely in he con ex o sweeping anspa ency egula ion such as he EU’s
DSA, DMA, AIA package ha online se ice p o ide s and pla o ms dec y he isk
o hei sys ems being gamed. Co one and S andbu g (2019) p o ide a e y help-
ul o e iew o he deba e on algo i hmic anspa ency and he conce n o ‘gaming
he sys em’. They d aw on li e a u e in (empi ical) legal s udies, compu e science,
and game heo y o explain how gaming can occu and when i is and isn’ a eal-
is ic isk. “Fundamen ally, he gaming h ea s ems om decision make eliance
on p oxies o c i e ia” (Co one & S andbu g, 2019, p. 626). As a use o a sys em,
knowledge o which p oxies a e used o decision making can allow one o ( y o)
exploi hose p oxies o ( y o) s ee he e en ual ou come/decision in he desi ed
di ec ion. The use o p oxies in algo i hmic decision-making con ex s is ine i able,
because hey a e used “when he ideal decision-making c i e ia a e unasce ain-
able as a p ac ical ma e o simply unknowable” (Co one & S andbu g, 2019, p.
635). Take, o ins ance, a da ing app. I we assume ha he da ing app decides on
ma ches based on which ma ches ha e he highes chance o de elop in o a suc-
cess ul ela ionship3, i immedia ely becomes clea ha he ideal decision-making
c i e ia – i.e., a success ul ela ionship – is si ua ed in he u u e and unknowable
a he ime o ma ching people. So, p oxies ha e o be used o app oxima e as i
we e he ideal decision-making c i e ia.
No all p oxies unc ion in he same manne hough and no all ypes o p oxies
a e equally sui able o gaming. Co one & S andbe g (2019, pp. 636-640) desc ibe
h ee laye s o p oxies. The i s laye conce ns inpu da a ha goes in o an algo-
i hmic decision-making p ocedu e. In a da ing app, his can be da a conce ning
one’s age, sexual p e e ences, and in e es s. The second laye uses he inpu (and
possibly o he ) da a “ o compu e a p edic ed alue o he ou come a iable ha is
only a p oxy o ha indi idual’s “ ue” ou come alue” (Co one & S andbu g,
2019, p. 638). In he da ing app example, his would conce n he ways in which in-
pu da a will be used o compu e a p edic ed alue o an ou come alue such as
‘p edic ed chance ha a ma ch leads o a da e’. The hi d laye conce ns how he
ou come alue chosen by he designe o he sys em a he second le el i sel
se es “as a p oxy o he ideal decision c i e ion” (Co one & S andbu g, 2019, p.
638). Fo a da ing app, he ideal decision c i e ion – which is no di ec ly measu -
able and/o knowable, hence he need o he use o p oxies – would be some-
hing like ‘ he wo people ma ching will de elop a success ul ela ionship’.
3. The e m ‘success ul ela ionship’ should be ead as ‘a ype o ela ionship o in e ac ion ha he
pe sons ha a e da ing conside o be sa is ying ela i e o wha e e s anda d hey hemsel es
deem ele an ’.
8 In e ne Policy Re iew 14(2) | 2025
Hence, he FICO Sco e case shows how gaming he sys em can happen h ough
mone ising h ee laye s o echnical anspa ency. C edi use s who buy an
in o ma ional epo on how he algo i hm echnically wo ks and pe o ms, can
imp o e hei c edi sco es and he eby gain mo e economic and social bene i s.
3.2 Tinde Concie ge, and o he da ing apps
Ha ing discussed he FICO sco es case, we now u n o a sligh ly mo e specula i e
example, namely Tinde Concie ge – a coaching se ice announced by Tinde in
2020 which p omised Tinde use s who we e willing o pay o i help om Tinde
employees o c a be e pe o ming p o iles (B own, 2020). The se ice has no
(ye ) been eleased in i s o iginally ad e ised o m, bu some o i s ea u es ha e
been in eg a ed in o some o Tinde ’s p emium membe ship package. Despi e i s
somewha specula i e na u e, we ind his example especially use ul because 1)
he announced ea u es a e almos an ideal ype o he ype o coaching-based
GaaS we a e in e es ed in, and 2) he ac ha Tinde announced his ea u e
publicly clea ly shows ha he indus y is in ac al eady hinking along hese
lines. In a as -paced pla o m economy, an icipa ing u u e de elopmen s by
sc u inising he expe imen s pla o ms a e engaged in is a good way o ensu e
one’s c i ical analyses will no only be eac i e in na u e. Mo eo e , o he se ices
in he da ing app ecosys em a e also mo ing in simila di ec ions wi h p emium
membe ships and ea u es ha p omise o help one pe o m be e in he
(algo i hmic cu a ion and anking o he) da ing app in ques ion. Whe e Tinde
p omised he expe ise o eal employees o hose willing o pay a p emium, o he
da ing apps ha e elied on AI-d i en coaching bo s a he han human expe s. Fo
example, Ma ch.com has de eloped an AI da ing cha bo named “La a” ha se es
as a pe sonal lo e coach by using na u al language p ocessing (Li, 2019). Some
o he da ing companies, like eHa mony, Happn, and Lo e lu e , ha e also
de eloped AI-d i en lo e coaches (Tu ley, 2021; Ghosh, 2017; Sil a, 2018). These
coaches p omise o help use s na iga e da es and op imise p o iles o ge mo e
da ing oppo uni ies (Tu ley, 2021).
Le us now look a Tinde and Tinde Concie ge in some de ail. Algo i hmic
anking and cu a ion ob iously plays a cen al ole on Tinde . Use s c ea e a p o ile
which con ains se e al pic u es, pe sonal in o ma ion (e.g., age, gende , wha one
is looking o on Tinde ) and in o ma ion on p e e ences (e.g., hobbies, music).
When using he app, one p o ile a a ime is shown o you based on an algo i hmic
anking p ocedu e ha is la gely opaque. I seems ob ious ha he in o ma ion
one has p o ided onesel plays a signi ican ole, bu one’s in e ac ion his o y wi h
Tinde p o iles shown o onesel is also hypo hesised o play a ole. We
delibe a ely w i e ‘hypo hesised’ because as a use you can only guess how he
algo i hmic anking wo ks. The e is a whole ecosys em o da ing websi es and
communi ies ying o e e se-enginee Tinde ’s algo i hmic anking by modelling
15 Sax, Wang
i on he concep o Elo a ing sys ems used mo e gene ally in game
heo y.7Academics, in u n, ha e also no ed how Tinde ’s algo i hmic opaqueness
poses no only me hodological challenges in e ms o esea ch bu also in oduces
unce ain ies o which use s can espond in a a ie y o ways (see, e.g., Duguay,
2017; Cou ois & Timme mans, 2018; Wang, 2023). A co e p esumed ea u e o
Tinde ’s algo i hm – o en discussed online – is an indi ec a ac i eness sco e
assigned o p o iles by p oxy, based on how o he s ha e in e ac ed wi h ha
p o ile (e.g., i many people swipe igh on you p o ile you a e p esumed o be
a ac i e). Ano he o en discussed assump ion8is ha one’s own swiping
beha iou also se es as a p oxy o one’s a ac i eness, whe e being a mo e
‘picky’ swipe is assumed o be a p oxy o being mo e a ac i e and being a e y
eage swipe is assumed o be a p oxy o being a less desi able membe o he
da ing pool.
Now, his is no he place o explo e hese ma e s in (mo e) de ail. Wha is
in e es ing hough is he ac ha he “algo i hmic imagina ies” (Buche , 2017)
which a e being discussed on da ing websi es and communi ies a e also, di ec ly
o indi ec ly, cons i u i e o unde s anding Tinde as an algo i hmic se ice ha
can – o maybe e en should – be gamed. He e we also see he di e en p oxy
laye s om Sec ion 2.3 again. One can y o game he ma ches one will ge
di ec ly by changing in o ma ion one pu s on one’s p o ile onesel (e.g., i you
change you age on you p o ile, you p o ile will be shown o people who ha e
indica ed hey a e looking wi hin a speci ic age ange). Bu as he Tinde Elo
communi y shows, he e a e also mo e indi ec ways o ying o in luence he
ele an p oxies by, o ins ance, expe imen ing wi h di e en ypes o swiping
beha iou o – hope ully – indi ec ly in luence he p oxies ha de e mine one’s
pe o mance in he algo i hm.
Wi hin his con ex o Tinde as a pe manen objec o gaming a emp s by he
communi y, Tinde Concie ge was announced in Ma ch 2020 as a p emium se ice
ha would cos be ween$20 and$50 a mon h (di e en Tinde use s go di e en
pop-ups announcing he se ice a di e en p ice poin s9). The pop-up shown o
some use s ead as ollows:
“Ou Concie ge se ice may be headed you way. Fo $20, you’ll ge access o
ou eam o expe s who will help you c a he pe ec p o ile. Go on, ha e a
as e o he good li e.”10
The GaaS o e ed he e clea ly doesn’ e ol e a ound he o e ing o ex ensi e
echnical anspa ency on Tinde ’s algo i hm di ec ly o he use . Ra he , Concie ge,
in i s announced o m, could be a g ea example o embodied anspa ency. The
16 In e ne Policy Re iew 14(2) | 2025
7. Elo a ing sys ems (named a e physics p o esso A pad Elo) a e used in, o ins ance, chess o
assign alues o indi idual playe s o p edic hei pe o mance in ou namen s (see, e.g., Olmeda,
2022). Al eady in 2019 The Ve ge epo ed ha Tinde s opped using ‘desi abili y sco es’ in i s
algo i hmic anking (Ca man, 2019). The ac ha an alleged li e al desi abili y sco e is no longe
used, does no imply ha ce ain p oxies o desi abili y a e no s ill impo an o Tinde ’s
algo i hmic anking o p o iles. A he momen o w i ing, he e a e s ill many da ing websi es and
communi ies w i ing de ailed b eakdowns o he hypo hesised wo king o Tinde ’s Elo sco e (see,
e.g., Bailey (2025) . The e a e also websi es ha p omise o help you calcula e (o be e :
app oxima e) you own Tinde Elo sco e (see, e.g., Bayley, 2020)
8. See o ins ance his Reddi h ead on he sub eddi / /SwipeHelpe ( a ne 100, 2021)
9. On Reddi , his led se e al Tinde use s o compa e he p ices hey we e shown o he se ice and
ask whe he he p ice hey we e shown said some hing abou hei a ac i eness (m8keup, 2020).
10. The sc eensho is included in a Fo bes a icle on Tinde Concie ge (B own, 2020).
17 Sax, Wang
se ice p omises access o Tinde employees who p esumably ha e a good
unde s anding o how he algo i hm wo ks ( hey a e “expe s”) and who can make
he algo i hm’s pe o mance anspa en o he use indi ec ly as i we e by making
sugges ions o changes o he p o ile o be e pe o mance o said p o iles on
Tinde . As o he speci ic sugges ions he Tinde expe s can make, we can only
guess wha hose could be because a he momen o w i ing Tinde has no
o icially launched he se ice in i s ad e ised o m.11
I he se ice would e e ma e ialise in i s ull ad e ised o m, we could imagine
se e al GaaS ea u es. Fi s o all, one could imagine he Tinde expe s o e ing
ad ice on how o build one’s p o ile: which pho os end o wo k well, which p o ile
ex s and s a ed in e es s wo k well. Such ad ice would basically come down o
ad ice on gaming he i s inpu da a laye . I swiping beha iou is indeed also an
impo an p oxy in he ma ching algo i hm one could also imagine wha could be
called ‘beha iou al ad ice’ on swiping o be pa o he Concie ge se ice.
3.3 The phenomenon o GaaS
These wo examples e lec a gene al phenomenon ha algo i hmic anspa ency
can be mone ised as pa o a gaming se ice. As a gued by Danielle Ci on and
F ank Pasquale, we a e inc easingly li ing in a “sco ed socie y” whe e p edic i e
algo i hms a e applied o ank indi iduals’ in impo an pa s o hei li es (Ci on
& Pasquale, 2014). Fo ins ance, di e en algo i hms a e c ea ed o ank
indi iduals who a e mos likely o ge a job, commi a c ime, de aul on bills, and
ind a da e. The cases o Tinde (o o he da ing apps) and FICO Sco e a e ypical
examples o his gene al end owa d a sco ed socie y. On he one hand, hese
algo i hm-d i en anking sys ems a e o en embedded wi h a se ies o ewa d and
punishmen mechanisms, which may encou age use s o game he sys em o hei
bes bene i s. Fo example, i Ube d i e s a e algo i hmically anked wi h low
sco es, hey will be immedia ely punished by ha ing hei ecommended
passenge s educed o e en being emo ed om he pla o m (Muldoon &
Raeks ad, 2023). On he o he hand, he opaque na u e o hese anking sys ems
discou ages use s om gaming he sys ems o hei own bene i s. As we
men ioned, he h ee laye s o gaming make i a he di icul o know how
algo i hms ac ually wo k and how o imp o e hei pe o mance, especially when i
comes o he second and hi d laye s o gaming.
This gap be ween use s’ needs o imp o e hei anking and he ac ual di icul y in
knowing how o do so opens oom o gaming as a comme cial se ice. This GaaS
can wo k h ough some deg ee o algo i hmic anspa ency, as imp o ing
pe o mance o en equi es some knowledge o how he algo i hm wo ks. Jus like
wha is shown in he example o FICO Sco e, which discloses i s sco ing
algo i hm’s b eakdown in o i e componen s, le ing c edi consume s know how
18 In e ne Policy Re iew 14(2) | 2025
11. Tinde Concie ge se ice did, howe e , esu ace as a pa o an e en mo e expensi e $500 a mon h
‘Tinde Vaul ’ se ice ha was being pilo ed in he Sp ing o 2023 (Ba , 2023) .
19 Sax, Wang
o imp o e hei sco es acco ding o hese i e ac o s. Simila ly, coaching se ices
in da ing apps can p o ide use s wi h some mo e embodied knowledge o how
hei beha io as a use in e ac s wi h he (la gely) opaque algo i hm(s) o he
se ice hey a e using. Coaches can use e e yday language o explain o indica e
how, oughly speaking, ac o s such as p o ile in o ma ion and pho o selec ion
in luence ma ches. Besides p o iding some gene al in o ma ion abou how
pla o m algo i hms unc ion, GaaS can also o e use s wi h some pe sonalised
knowledge abou how hei pa icula si ua ion, such as no anking high in he
da ing ma ke , is in luenced by he pla o m’s algo i hms. Based on hese
pe sonalised insigh s, coaching-based GaaS could sugges ailo ed me hods o
imp o e pe o mance on he pla o m.
GaaS can hus be unde s ood as a p ocess o mone ising algo i hmic anspa ency
in he sco ed socie y, whe e se ice p o ide s can cha ge i s use s a p emium o
ecei e ‘ anspa ency bene i s’ which he use can use o make one’s p o ile o
con en pe o m be e in algo i hmic anking o cu a ion p ocesses.12In an
en i onmen ha is con olled by he se ice p o ide , howe e , such a p emium
se ice will ypically no o e ‘ eal’, blanke anspa ency, bu a he a mo e
s a egic, pa ial anspa ency w apped in an addi ional se ice whe e he same
se ice p o ide o e s ‘ ools’ and ‘ad ice’ o help ‘op imise’ one’s con en o p o ile
o algo i hmic anking. Pu simply, he alleged isk o anspa ency – i.e., gaming
he sys em – is ac ually u ned in o a mone isa ion ea u e. In he ollowing
sec ion, we will explo e he po en ial e hical challenges o his GaaS.
4. Explo ing e hical challenges o GaaS
In his sec ion we discuss some o he possible e hical challenges posed by GaaS.
Because GaaS as we desc ibe i is a ecen phenomenon, a ie ies o which we
expec o become mo e equen in he nea u u e, his sec ion is mean as a i s
explo a ion.
4.1 Consume -un iendly incen i e s uc u e and exploi a ion
Fi s , GaaS o e ed by he se ice i sel in oduces (o ein o ces) an incen i e
s uc u e ha is hos ile o consume s. To see why he incen i e s uc u e is hos ile,
i should i s be no ed ha GaaS can only be o e ed agains a backg ound o
su icien opaqueness. I a se ice o pla o m elies on algo i hmic cu a ion/
anking which is comple ely anspa en and unde s andable, GaaS will quickly
lose i s alue p oposi ion. Conside he announced (bu la e abandoned) Tinde
Concie ge se ice; his se ice would only be an in e es ing se ice o he public
because he Tinde algo i hm emains o be seen as enigma ic. So, i GaaS is o be
pu sued as a comme cial s a egy, i in oduces an incen i e o main aining (o
e en in oducing new) le els o opaqueness as a necessa y p econdi ion o
20 In e ne Policy Re iew 14(2) | 2025
12. The e a e some o he mo e gene al ypes o GaaS. Fo ins ance, use s can be cha ged a p emium o
ecei e mo e di ec ‘gaming bene i s’ which se e as de ac o pay- o-win ea u es o inc ease one’s
chances o pe o m be e in ( anking) sys ems one in e ac s wi h. Bu o his pape , we’ e mo e
in e es ed in he pa icula phenomenon o GaaS ela ed o he mone isa ion o algo i hmic
anspa ency.
21 Sax, Wang
o e ing GaaS. Such an ou come would be disappoin ing gi en he ecen push o
(e en mo e) anspa ency obliga ions in, o ins ance, he Eu opean legisla i e
agenda o he digi al economy (DSA, DMA, AI Ac ). The s a egic ma ke eac ion
o inding ways o mone ise a desi e o anspa ency in ways ha , i any hing,
in oduce incen i es o no p ac ise genuine anspa ency, could indeed be seen as
bo h i onic and cynical. I emains o be seen, howe e , whe he a iances o GaaS
will ac ually be complian wi h e.g. anspa ency p o isions in he DSA.
Besides he hos ile incen i e s uc u e i sel , one could also ques ion wha ype o
ela ionship be ween he se ice p o ide and he use GaaS esul s in. GaaS is
p emised on ac i ely os e ing opaqueness o be able o o e a se ice ha has
use s pay an addi ional ee o alle ia e/o e come nega i e ex e nali ies esul ing
om he delibe a e opaqueness. One possible way o cha ac e ising such a
ela ionship is as exploi a i e. Philosophically, he e is no consensus on he
p ecise meaning and de ini ion o exploi a ion (see Zwolinski e al., 2022 o an
ex ensi e o e iew o di e en iews and deba es). A minimal, uncon o e sial
unde s anding o exploi a ion is ha exploi a ion in ol es aking
un ai 13ad an age o one’s a ge by using a ulne abili y o he a ge (ei he a
pe sonal cha ac e is ic, o some hing in he a ge ’s en i onmen ) o one’s own
bene i . When we conside Tinde Concie ge as an example o GaaS, a case can
indeed be made o he se ice ha ing signi ican exploi a i e cha ac e is ics.
Use s use Tinde o ind some hing ha is e y impo an o hem: da es, lo e,
in imacy, companionship. To ‘gain access’ o hose ‘goods’ use s a e made e y
awa e o he ac ha hey ha e o pe o m in a compe i i e da ing ma ke ha is
s uc u ed a ound he illus ious p op ie a y Tinde algo i hm (o which he e
exis s a li ely algo i hmic imagina y as we discussed in Sec ion 3.1). So, wha we
ha e is 1) a la ge g oup o use s wi h a s ong desi e o pa icula ou comes
(ma ches, da es, lo e, in imacy), and 2) a la gely opaque (algo i hmic) ga e-keeping
mechanism. Combined, hese wo ci cums ances cons i u e he e ile soil o
delibe a ely exploi a i e comme cial p ac ices. The s ong desi es o ma ches,
da es, lo e, and in imacy se e as an exploi able ulne abili y. Tinde ’s p ojec ed
p emium se ice Tinde Concie ge would ce ainly quali y as an exploi a i e
s a egy ha can be used o ake ad an age o hose ulne abili ies o he bene i
o Tinde . And e en hough Tinde Concie ge ul ima ely did no ge implemen ed,
al eady exis ing Tinde p emium se ices such as Plus, Gold, and Pla inum also
bea he ma ks o a simila exploi a i e logic. These p emium se ices p omise a
ange o ‘powe -ups’ o make you mo e compe i i e is-à- is you Tinde
compe i o s (Tinde , n.d.-a). Fo example, o he mos expensi e subsc ip ion
called Pla inum, Tinde w i es: “Inc ease you ma ch-making po en ial and enjoy
mos o Tinde ’s p emium ea u es wi h Tinde Pla inum™! Da ing online jus go
easie . See someone you’d lo e o mee and can’ wai o ma ch? As a Pla inum
subsc ibe , you can a ach a no e o e e y Supe Like you send, inc easing you
ma ch-making po en ial by up o 25%” (emphasis added, Tinde , n.d.-b). The same
22 In e ne Policy Re iew 14(2) | 2025
13. By using he wo d ‘un ai ’ he e, he p oposed minimal unde s anding o exploi a ion is a no ma i e
one: exploi a ion in his sense is unde s ood as in p inciple w ong (o e iding easons o deeming
he exploi a ion accep able all hings conside ed can s ill exis ). A minimal unde s anding which
does no inco po a e he wo d ‘un ai ’ is o cou se also possible, which would lead o a mo e
neu al unde s anding o exploi a ion. In ha case a oo ball playe making use o a weakness in
he opponen ’s de ensi e posi ioning would also ‘exploi ’ ha pa icula ulne abili y o hei own
bene i . Bu no one would call ha ype o exploi a ion w ong; i ’s pa o he game. We would,
howe e , gene ally conside i un ai i he oo ball playe would – o ins ance – eign an inju y o
cause he opponen s o huddle a ound him o check on him. I he oo ball playe hen exploi s he
de ensi e diso ganiza ion he caused by aking an inju y, we would say he exploi ed his opponen s’
spo smanship in an un ai (i.e., e hically p oblema ic) manne .
23 Sax, Wang
analysis o exploi a ion applies in an e en mo e s aigh o wa d manne o FICO
Sco es. C edi sco es clea ly de e mine one’s abili y o access a wide ange o
essen ial se ices, as well as he condi ions (e.g., in e es a es) unde which one
can access hose essen ial se ices. Selling p emium anspa ency se ices which
a e p emised on people’s eal ea o bad/wo sening c edi quali ies as an
exploi a i e p ac ice unde mos concep ions o exploi a ion.
One may objec ha Tinde is ‘jus a da ing app’ which use s do no need o use a
all i hey don’ wan o. Deciding o use he app also means consen ing o Tinde
a emp ing o p essu e you in o pu chasing p emium se ices ha can help you
‘game’ Tinde ’s algo i hmic ma ch making. The e a e h ee b ie answe s o his
possible objec ion. Fi s , he idea ha as long as consume s a e no ac i ely o ced
o use a se ice any hing goes, is simply w ong. In he EU he e is, o ins ance,
un ai comme cial p ac ice law (Di ec i e 2005/29/EC) which o bids many ypes o
misleading and agg essi e comme cial p ac ices ha ake un ai ad an age o
consume s. Second, he sen imen ha Tinde is ‘jus a da ing app’ is misguided.
Lo e, in imacy, and companionship a e basic human needs and people a e in ac
inc easingly u ning o apps like Tinde o help ul il hose needs. Mo eo e , being
in a ela ionship wi h someone also ends o p i ilege one socie ally since i
allows one o, o ins ance, apply o a mo gage oge he . The e a e socio-
economic implica ions o one’s ‘da ing s a us’, meaning ha a popula da ing app
canno be dismissed as ‘jus a da ing app’.
Thi d and las , e en i one we e o (s ill) hink ha Tinde is no impo an and he
exploi a i e cha ac e is ics o he use -GaaS-p o ide ela ionship a e he e o e
no wo isome, hen we should s ill hink o o he con ex s GaaS could be
in oduced and whe e i would be conside ed (mo e) p oblema ic. The FICO Sco es
example wi h he paid- o se ices ha help use s who can a o d i o op imise
hei c edi sco es comes o mind. C edi sco es can play such a decisi e ole in
people's li es – i can de e mine whe he you can ge he igh mo gage o no –
ha GaaS(-like) se ices in he c edi sco ing con ex a e a legi ima e conce n.
4.2 Pay- o-win and equali y
I one uses GaaS in he hope o pe o ming be e in an algo i hmic anking/
ma ching/cu a ion scena io in o de o, in he end, secu e be e ou comes, one is
basically engaging in wha in he ideogame con ex is known as pay- o-win. Pay-
o-win en ails paying o a compe i i e ad an age (e.g., by ecei ing be e
weapons o ecei ing heal h upg ades), o en wi hou he absolu e gua an ee o
ac ually winning; you s ill ha e o ac ually de ea opponen s wi h you bough
ad an ages helping you. So, echnically speaking pay- o-win means pay- o-be-
mo e-likely- o-win in mos cases. Paid- o ad an ages ypically ‘s ack’, so he mo e
you pay, he mo e ad an ages you can ac i a e and/o he s onge hose
24 In e ne Policy Re iew 14(2) | 2025
Cou ois, C., & Timme mans, E. (2018). C acking he Tinde code: An expe ience sampling app oach
o he dynamics and impac o pla o m go e ning algo i hms. Jou nal o Compu e -Media ed
Communica ion, 23(1), 1–16. h ps://doi.o g/10.1093/jcmc/zmx001
Diakopoulos, N. (2020). T anspa ency. In M. D. Dubbe , F. Pasquale, & S. Das (Eds.), The Ox o d
handbook o e hics o AI (pp. 196–213). Ox o d Uni e si y P ess. h ps://academic.oup.com/edi ed- o
lume/34287/chap e /290661457
Duguay, S. (2017). D essing up Tinde ella: In e oga ing au hen ici y claims on he mobile da ing
app Tinde . In o ma ion, Communica ion & Socie y, 20(3), 351–367. h ps://doi.o g/10.1080/1369118
X.2016.1168471
Eubanks, V. (2018). Au oma ing inequali y: How high- ech ools p o ile, police, and punish he poo . S .
Ma in’s P ess.
Eu opean Pa liamen and Council. (2005). Di ec i e 2005/29/EC o he Eu opean Pa liamen and o he
Council o 11 May 2005 conce ning un ai business- o-consume comme cial p ac ices in he in e nal
ma ke . h p://da a.eu opa.eu/eli/di /2005/29/oj
Eu opean Pa liamen and he Council. (2024). Regula ion (EU) No 2024/1689 o he Eu opean
Pa liamen and he Council o 13 June 2024 laying down ha monised ules on a i icial in elligence. h
p://da a.eu opa.eu/eli/ eg/2024/1689/oj
a ne 100. (2021, June 15). Being picky wo ks like a cha m o my iend’s accoun . Reddi . h ps://ww
w. eddi .com/ /SwipeHelpe /commen s/o0phz /being_picky_wo ks_like_a_cha m_ o _my_ iends/
F eeman, G., Wu, K., Nowe , N., & Wohn, D. Y. (2022). Pay o win o pay o chea : How playe s o
compe i i e online games pe cei e ai ness o in-game pu chases. P oceedings o he ACM on
Human-Compu e In e ac ion, 6(CHI PLAY), 1–24. h ps://doi.o g/10.1145/3549510
Ghosh, S. (2017, Ap il 14). Da ing app Happn is ge ing paid subsc ip ions and will use AI o
ecommend ma ches. Business Inside . h ps://www.businessinside .com/happn-is-ge ing-paid-subs
c ip ions-and-will-use-ai- o- ecommend-ma ches-2017-4?in e na ional= ue& =US&IR=T
Habe mas, J. (1987). The heo y o communica i e ac ion, olume 2: Li ewo ld and sys em: A c i ique o
unc ionalis easoning (T. McCa hy, T ans.). Beacon P ess.
Ha esamud am, K., La sson, S., & Hein z, F. (2023). Th ee le els o AI anspa ency. Compu e , 56(2),
93–100. h ps://doi.o g/10.1109/MC.2022.3213181
Helbe ge , N., Sax, M., S ycha z, J., & Mickli z, H.-W. (2022). Choice a chi ec u es in he digi al
economy: Towa ds a new unde s anding o digi al ulne abili y. Jou nal o Consume Policy, 45(2),
175–200. h ps://doi.o g/10.1007/s10603-021-09500-5
Hood, C., & Heald, D. (Eds.). (2006). T anspa ency: The key o be e go e nance? (1s ed.). B i ish
Academy. h ps://doi.o g/10.5871/bacad/9780197263839.001.0001
K oll, J. A., Huey, J., Ba ocas, S., Fel en, E. W., Reidenbe g, J. R., Robinson, D. G., & Yu, H. (2017).
Accoun able algo i hms. Uni e si y o Pennsyl ania Law Re iew, 165(3), 633–705.
La sson, S., & Hein z, F. (2020). T anspa ency in a i icial in elligence. In e ne Policy Re iew, 9(2). h
ps://doi.o g/10.14763/2020.2.1469
31 Sax, Wang
Laue , J. (2017). C edi wo hy: A his o y o consume su eillance and inancial iden i y in Ame ica.
Columbia Uni e si y P ess.
Lee ssen, P. J. (2023). Seeing wha o he s a e seeing. S udies in he egula ion o anspa ency o social
media ecommende sys ems [Doc o al disse a ion, Uni e si y o Ams e dam]. h ps://hdl.handle.ne
/11245.1/18c6e9a0-1530-4e70-b9a6-35 b37873d13
Li, H. (2019, June 15). The online da ing indus y lo es a i icial in elligence. Synced. h ps://synced e
iew.com/2019/06/15/ he-online-da ing-indus y-lo es-a i icial-in elligence/
m8keup. (2020, Ma ch 3). Wha is inde concie ge? Reddi . h ps://www. eddi .com/ /Tinde /comme
n s/ d2 5m/wha _is_ inde _concie ge/
Mi els ad , B., Russell, C., & Wach e , S. (2019). Explaining explana ions in AI. P oceedings o he
Con e ence on Fai ness, Accoun abili y, and T anspa ency, 279–288. h ps://doi.o g/10.1145/328756
0.3287574
Mo ozo ai ė, V. (2024). Taming digi al in luence: Hype nudging and he ole o Eu opean compe i ion
law [Doc o al disse a ion, U ech Uni e si y]. h ps://doi.o g/10.33540/2080
Muldoon, J., & Raeks ad, P. (2023). Algo i hmic domina ion in he gig economy. Eu opean Jou nal o
Poli ical Theo y, 22(4), 587–607. h ps://doi.o g/10.1177/14748851221082078
myFICO. (n.d.-a). Choose a c edi epo . myFICO. h ps://www.my ico.com/p oduc s/ ico-sco e-c edi -
epo s
myFICO. (n.d.-b). How i wo ks. myFICO. h ps://www.my ico.com/p oduc s/ ico-sco e-how-i -wo ks
myFICO. (n.d.-c). Wha ’s he di e ence be ween FICO sco es and non-FICO c edi sco es? myFICO. h p
s://www.my ico.com/c edi -educa ion/ ico-sco es- s-c edi -sco es#:~: ex =FICO%20Sco es%20a
e%20used%20by,looking%20a %20you %20FICO%20Sco e
myFICO (Di ec o ). (2021, Ap il 19). myFICO 20 h anni e sa y – imeline o ou i s 20 yea s [YouTube
ideo]. h ps://www.you ube.com/wa ch? =ATwgHpOQ6bw
Olmeda, F. (2022). Towa ds a s a is ical physics o da ing apps. Jou nal o S a is ical Mechanics:
Theo y and Expe imen , 2022(11), 113501. h ps://doi.o g/10.1088/1742-5468/ac9bed
Pasquale, F. (2015). The black box socie y: The sec e algo i hms ha con ol money and in o ma ion.
Ha a d Uni e si y P ess.
Pe e, C., Du y, B. E., & Hund, E. (2019). “Gaming he sys em”: Pla o m pa e nalism and he poli ics
o algo i hmic isibili y. Social Media + Socie y, 5(4). h ps://doi.o g/10.1177/2056305119879995
Rawls, J. (1971). A heo y o jus ice. Ha a d Uni e si y P ess.
Sax, M. (2021). Be ween empowe men and manipula ion. Kluwe .
Sax, M., & Ausloos, J. (2021). Ge ing unde you skin(s): A legal-e hical explo a ion o Fo ni e’s
ans o ma ion in o a con en deli e y pla o m and i s manipula i e po en ial. In e ac i e
En e ainmen Law Re iew, 1–24. h ps://doi.o g/10.4337/iel .2021.0001
Sil a, D. (2018). Da ing apps use a i icial in elligence o help sea ch o lo e. In Phys.o g. h ps://ph
32 In e ne Policy Re iew 14(2) | 2025
ys.o g/news/2018-11-da ing-apps-a i icial-in elligence.h ml
Tinde . (n.d.-a). Subsc ip ion ie s. Tinde . h ps:// inde .com/ ea u e/subsc ip ion- ie s
Tinde . (n.d.-b). Tinde pla inum. Tinde . h ps:// inde .com/ ea u e/pla inum
T egel, T., Schwab, M. C., Nguyen, T. T. L., Mülle , P. N., & Göbel, S. (2020). Cos s o compe e –
Analyzing pay o win aspec s in cu en games. In M. Ma, B. Fle che , S. Göbel, J. Baals ud Hauge, &
T. Ma sh (Eds.), Se ious games (pp. 177–192). Sp inge In e na ional Publishing. h ps://doi.o g/10.1
007/978-3-030-61814-8_14
Tu ley, D. (2021, Janua y 17). Lo e in he ime o algo i hms: Would you le you a i icial in elligence
choose you pa ne ? The Con e sa ion. h ps:// hecon e sa ion.com/lo e-in- he- ime-o -algo i hms-
would-you-le -a i icialin elligence-choose-you -pa ne -152817
Walze , M. (1983). Sphe es o jus ice: A de ense o plu alism and equali y. Basic Books.
Wang, H. (2022). T anspa ency as manipula ion? Unco e ing he disciplina y powe o algo i hmic
anspa ency. Philosophy & Technology, 35(3), 69. h ps://doi.o g/10.1007/s13347-022-00564-w
Wang, H. & Philosophy Documen a ion Cen e . (2023). Algo i hmic coloniza ion o lo e: The e hical
challenges o da ing app algo i hms in he age o AI. Techné: Resea ch in Philosophy and Technology,
27(2), 260–280. h ps://doi.o g/10.5840/ echne202381181
Welle , A. (2017). T anspa ency: Mo i a ions and challenges (Ve sion 2). a Xi . h ps://doi.o g/10.4855
0/ARXIV.1708.01870
Zubo , S. (2019). The age o su eillance capi alism. P o ile Books.
Zwolinski, M., Fe guson, B., & We heime , A. (2022). Exploi a ion. In E. N. Zal a & U. Nodelman
(Eds.), S an o d encyclopedia o philosophy. h ps://pla o.s an o d.edu/a chi e
33 Sax, Wang
