T anspo a ion Enginee ing 4 (2021) 100064
Con en s lis s a ailable a ScienceDi ec
T anspo a ion Enginee ing
jou nal homepage: www.else ie .com/loca e/ eng
Assessing whe he a ificial in elligence is an enable o an inhibi o o
sus ainabili y a indica o le el
Shi am Gup a
a
, Simone D. Langhans
b
, Sami Domisch
c
, F ancesco Fuso-Ne ini
d
,
Anna Fellände
e
, Manuela Ba aglini
, Max Tegma k
g
, Rica do Vinuesa
h , e , ∗
a
Bonn Alliance o Sus ainabili y Resea ch/Inno a ion Campus Bonn (ICB), Uni e si y o Bonn, Bonn D-53113, Ge many
b
Basque Cen e o Clima e Change (BC3), Leioa 48940, Spain
c
Depa men o Ecosys em Resea ch, Leibniz Ins i u e o F eshwa e Ecology and Inland Fishe ies, Müggelseedamm 310, Be lin 12587, Ge many
d
Uni o Ene gy Sys ems Analysis (dESA), KTH Royal Ins i u e o Technology, B inell ägen, 68SE-1004 S ockholm, Sweden
e
AI Sus ainabili y Cen e , S ockholm SE-114 34, Sweden
T anspa en In e ne , Tå up Bygade 30 Mesinge, DK-5370, Denma k
g
Cen e o B ains, Minds and Machines, Massachuse s Ins i u e o Technology, Camb idge, MA 02139, Uni ed S a es
h
SimEx/FLOW, Enginee ing Mechanics, KTH Royal Ins i u e o Technology, S ockholm SE-10044, Sweden
a i c l e i n o
Keywo ds:
AI
Sus ainabili y
Machine lea ning
T anspo a ion sys em
Clima e change
a b s a c
Since he ea ly phase o he a ificial-in elligence (AI) e a expec a ions owa ds AI a e high, wi h expe s be-
lie ing ha AI pa es he way o managing and handling a ious global challenges. Howe e , he significan
enabling and inhibi ing influence o AI o sus ainable de elopmen needs o be assessed ca e ully, gi en ha
he echnology diffuses apidly and affec s millions o people wo ldwide on a day- o-day basis. To add ess his
challenge, a panel discussion was o ganized by he KTH Royal Ins i u e o Technology, he AI Sus ainabili y Cen-
e and MIT Massachuse s Ins i u e o Technology, ga he ing a wide ange o AI expe s. This pape summa izes
he insigh s om he panel discussion a ound he ollowing hemes: The ole o AI in achie ing he Sus ainable
De elopmen Goals (SDGs); AI o a p ospe ous 21s cen u y; T anspa ency, au oma ed decision-making p o-
cesses, and pe sonal p ofiling; and Measu ing he ele ance o Digi aliza ion and A ificial In elligence (D&AI)
a he indica o le el o SDGs. The esea ch-backed panel discussion was dedica ed o ecognize and p io i-
ize he agenda o add essing he p essing esea ch gaps o academic esea ch, unding bodies, p o essionals,
as well as indus y wi h an emphasis on he anspo a ion sec o . A common conclusion ac oss hese hemes
was he need o go beyond he de elopmen o AI in sec o ial silos, so as o unde s and he impac s AI migh
ha e ac oss socie al, en i onmen al, and economic ou comes. The eco dings o he panel discussion can be
ound a :
h ps://www.k h.se/en/2.18487/e enemang/ he- ole-o -ai-in-achie ing- he-sdgs-enable -o -inhibi o -
1.1001364?da e = 2020–08–20&leng h = 1&o gleng h = 185&o gda e = 2020–06–30
Sho link: h ps://bi .ly/2Kap1 E
Mo i a ion
The as -paced ise o a ificial in elligence (AI) impac s a wide ange
o sec o s, hence i is c ucial o igilan ly assess he oppo uni ies and
challenges AI may pose o sus ainable de elopmen . Fo ins ance, a e-
cen e iew by Vinuesa e al. [59] epo ed bo h posi i e and nega i e
impac s o AI on he 17 Sus ainable De elopmen Goals (SDGs) om
he Uni ed Na ions [55] . The amoun o li e a u e ega ding he use and
applica ions o AI g ows apidly, epo ing on he po en ial AI holds o
∗ Co esponding au ho a : SimEx/FLOW, Enginee ing Mechanics, KTH Royal Ins i u e o Technology, S ockholm SE-10044, Sweden.
E-mail add ess: [email p o ec ed] (R. Vinuesa).
p o ide suppo in a ious c i ical con ex s, such as in he medical do-
main o in disas e managemen [48 , 51] . In addi ion, o he a eas can
significan ly benefi om AI, e.g. he anspo a ion sec o h ough AI
applica ions aimed a educing he d ag o a numbe o ehicles [23 , 46]
as well as he u ban sus ainabili y sec o h ough mo e obus me hods
o p e en high pollu ion le els and u ban-hea -island effec s [53] . De-
spi e hese benefi s i is c ucial o unde s and he p ac ical implica ions
o deploying AI-based echnology and he po en ial nega i e impac s
on SDGs ela ed o example o equali y o clima e change. In he ol-
h ps://doi.o g/10.1016/j. eng.2021.100064
Recei ed 18 Decembe 2020; Recei ed in e ised o m 5 Ma ch 2021; Accep ed 7 Ma ch 2021
2666-691X/© 2021 The Au ho (s). Published by Else ie L d. This is an open access a icle unde he CC BY license ( h p://c ea i ecommons.o g/licenses/by/4.0/ )
S. Gup a, S.D. Langhans, S. Domisch e al. T anspo a ion Enginee ing 4 (2021) 100064
lowing, we discuss he a ious opics co e ed in he panel discussion
1
o ganized by KTH Royal Ins i u e o Technology, he AI Sus ainabili y
Cen e and MIT Massachuse s Ins i u e o Technology. Th oughou his
pape , we used he e ms ‘sus ainable ansi ion’ e e ing o he adi-
cal ans o ma ion owa ds a sus ainable socie y, ‘sus ainable de elop-
men ’ e e ing o he de elopmen which helps in mee ing he cu en
equi emen s wi hou comp omising he abili y o u u e gene a ions,
and ‘sus ainable ans o ma ion’ e e ing o he shi ha may unda-
men ally al e human and en i onmen al in e ac ions. Emphasis will be
placed on (i) disce ning which applica ions can benefi mos om AI,
(ii) which ones may be mos significan ly a isk, as well as (iii) on he
impac o AI in p oducing a social di ide.
The ole o a ificial in elligence in achie ing he Sus ainable
De elopmen Goals
The SDGs p o ide an excellen amewo k o assessing he impac
o AI on diffe en sec o s as well as o iden i ying possible syne gies
among hem. We s a ed by di iding he 17 SDGs in o h ee ca ego ies:
Socie y, Economy, and En i onmen . This di ision is consis en wi h
o he s p e iously epo ed in he li e a u e [45 , 54] . Th ough a de ailed
expe -elici a ion p ocess, we ound ha 79% o he SDG a ge s can
be posi i ely affec ed by AI, while 35% may be nega i ely affec ed by
AI de elopmen [59] . When analyzing he h ee diffe en g oups, we
obse ed ha he En i onmen ca ego y en ails he highes po en ial
wi h 93% o he a ge s being posi i ely affec ed, whe eas Socie y has
he la ges nega i e effec wi h 38% o he a ge s exhibi ing a nega i e
in e ac ion wi h AI. I is impo an o no e ha , when aking in o accoun
he ype o e idence indica ing he connec ion wi h AI, we obse ed
ha he posi i e effec s on he En i onmen and on Socie y we e qui e
obus , whe eas addi ional esea ch is needed o e alua e he possible
posi i e effec s o AI on he Economy. Addi ional de ails can be ound
in he wo k by Vinuesa e al. [59] .
Some examples o posi i e effec s due o AI include SDGs 8 and 9,
whe e AI-enabled echnology may help c ea e new jobs and enhance
p oduc i i y, as well as SDG 1 h ough he use o sa elli e da a o ack
po e y [25] . O he posi i e applica ions o AI a ge SDGs 7 and 13,
specifically in he con ex o mo e efficien ene gy use ( h ough sma -
g id applica ions) and h ough de eloping mo e obus ools o p edic
and manage pollu ion in ci ies [53] . This is o g ea ele ance e.g. in
Eu ope whe e u ban pollu ion is esponsible o app ox. 800,000 dea hs
pe yea [31] . AI can he e o e ha e a posi i e impac on SDG 13 (cli-
ma e ac ion); no e ha he o he wo SDGs in he En i onmen g oup,
i.e. SDGs 14 and 15, can also significan ly benefi om AI h ough im-
p o ed conse a ion and managemen o biodi e si y and na u al e-
sou ces.
AI has he po en ial o h ea en he achie emen o a numbe o appli-
ca ions ele an o he SDGs wi hin Economy and Socie y. Fi s , jobs may
be los [8] , al hough based on al e na i e s udies he ne impac may
no be nega i e [1] . This h ea is eflec ed in une en oppo uni ies o
access AI compu ing esou ces, which ul ima ely may inc ease inequali-
ies. Fu he mo e, AI algo i hms may pola ize socie ies and inc ease dis-
c imina ion [15] . These can be eflec ed in da a-d i en me hods o han-
dle he COVID-19 pandemic [35] , and mo e specifically con ac - acing
sma phone applica ions [60] , whe e he handling is key o a oid dis-
c imina ion and pola iza ion. No e ha pola ized and unequal socie ies
can in u n ha e impo an impac s on peace and s abili y [39] , i.e. o
SDGs 16 and 17. I is impo an o highligh ha AI can ha e la ge effec s
on he global ene gy demand. The o al elec ici y demand o in o ma-
ion and communica ions echnologies (ICT) could equi e up o 20% o
he global elec ici y by 2030, being a 1% oday [26] .
1 h ps://www.k h.se/en/2.18487/e enemang/ he- ole-o -ai-in-
achie ing- he-sdgs-enable -o -inhibi o -1.1001364?da e = 2020-08-
20&leng h = 1&o gleng h = 185&o gda e = 2020-06-30
Technology, indi iduals, and go e nmen s a e h ee agen s in e ac -
ing wi h each o he as well as wi h he En i onmen . The eby, he speed
o change in echnology is so high ha indi iduals (when i comes o
echnology adop ion) and go e nmen s ( ega ding egula ions) signifi-
can ly lag behind. Consequen ly, he e a e la ge esea ch gaps o be a -
ended o, o manage he ansi ion o an AI-based socie y. In addi ion,
ou esul s ha e exposed he significan ulne abili y o in as uc u es.
Subs an ial wo k is needed o o e come AI gaps in anspa ency, sa e y,
and e hical s anda ds, so ha e e yone can benefi om he la ge po-
en ial o AI (see also [59] ).
AI o a p ospe ous 21s cen u y
AI is a double-edged swo d: On he one hand, AI has g ea po en ial
o con ibu e immensely o sol ing exis ing p oblems and he ewi h o
people’s well-being in he nea u u e. As AI could eplace many p o-
cedu es and ou ines ca ied ou by humans, he isk o human ailu e
could be educed and hence nume ous li es could be sa ed by, o in-
s ance, (i) sel -d i ing ca s based on AI- echnology ha educe he isk
o acciden s [27] ; bu see [41] , (ii) diminishing he amoun o mis akes
made in hospi al ca e [36] ; (iii) imp o ing diagnos ics in e.g. lung can-
ce [42] and blindness [40] ; (i ) accele a ing medical science gi en new
AI- echnologies ha also con ibu e owa ds he disco e y o new d ugs
[16] . In addi ion, ( ) educa ion is likely o become mo e cus omizable
and accessible a ound he wo ld. On he o he hand, AI can be pu pose-
ully misused causing ha m, such as done by AI-d i en da a mining (see
Camb idge Analy ica , [13] o ha es da a abou human online beha -
io wi h he aim o influence elec ions. Likewise, AI- echnology is used
o ca y ou d one-flown bomb a acks ha ha e killed people in a -
ious ins ances. O he examples include, o ins ance, he gli ch in he
AI- ading sys em a Knigh Capi al which led o a 440 Million USD loss
[14] , o Boeing ’s sa e y-c i ical sys ems being po en ially ulne able o
being hacked h ough hei en e ainmen sys em so wa e [20] .
The complexi y o AI echnologies and he ac ha decision mak-
e s o en do no unde s and i s powe , a e posing majo challenges o
a ai , e hical and sus ainable use o AI [59] . The key is o ob ain as
much o he benefi s o AI, while concu en ly a oiding misuses, one
ha needs o be add essed u gen ly since AI echnology is ad ancing
s eadily and becoming mo e powe ul apidly. A solu ion o his chal-
lenge is o d aw a clea line be ween he accep able and unaccep able
use o AI. The widesp ead doub ha such an ag eemen is un ealis ic
can be ebu ed by he example o he ban o bioweapons ag eed upon
du ing he 1975 Asiloma con e ence on he egula ion o bio echnol-
ogy, mo e specifically DNA echnologies [6] which since hen is globally
ollowed. As an analogy, such beneficial use o AI is shown e.g. in e ms
o DNA echnologies ha keep playing an essen ial ole in he de el-
opmen o accines. In 2017, he Asiloma Con e ence on Beneficial AI
c ea ed a se o guidelines o AI esea ch – he 23 Asiloma AI P inci-
ples ( h ps:// u u eofli e.o g/ai-p inciples/ ) which a e being endo sed
by a g owing numbe o key playe s in AI esea ch, indus y and de el-
opmen .
The ul ima e d i e o an AI ag eemen is a globally-sha ed posi i e
ision based on equal benefi s o all. Despi e sounding u opian, he e
a e a ious examples o sha ed isions ha ha e led o posi i e changes
in he pas : he in en ion o democ acy, he scien ific, indus ial and
compu e e olu ions, ee heal hca e, ee highe educa ion, o peace
h ough in e dependence in Eu ope. Taking he SDGs as ou baseline,
he global-AI ision may go beyond and aspi e p ospe i y o all, knowl-
edge h ough AI, AI o science, sa ing he clima e wi h AI, cu ing cance
wi h AI, and heal h o all.
Socie al and e hical conside a ions when applying he SDGs o AI
P io o he SDGs he Uni ed Na ions es ablished he Millennium De-
elopmen Goals which p omp ed he “In o ma ion and Communica ion
2
S. Gup a, S.D. Langhans, S. Domisch e al. T anspo a ion Enginee ing 4 (2021) 100064
Technologies (ICTs) o De elopmen ”ini ia i e. The B oadband Com-
mission o Sus ainable De elopmen assessed he ole ha echnology
could play, no jus in enabling he goals, bu in c ea ing echnology as
an accele a o o he goals. The p ima y in en o he Commission was
o add ess he digi al di ide by pai ing digi aliza ion wi h sus ainabili y,
while expanding he ICT indus y o enable access o he newly mobile
and digi al wo ld o billions o people.
Mo e ecen ly, Melinda Ga es and Jack Ma decla ed The Age o Digi-
al In e dependence which included fi e se s o ecommenda ions on how
he in e na ional communi y could wo k oge he o op imize he use o
digi al echnologies and o mi iga e hei isks. The ecommenda ions in-
cluded: inclusi e digi al economy; human and ins i u ional capaci y; hu-
man igh s and human agency; us , secu i y, and s abili y; and global
digi al coope a ion.
These conside a ions a e especially ele an wi h he ad en o AI
gi en he possibili ies o exponen ial benefi s and exponen ial isks.
AI is diffe en om he ICTs o he pas , because i is sel -scaling, sel -
lea ning, and sel -p opaga ing. Wi h he echnology becoming mo e ad-
anced, he lack o p ope go e nance, anspa ency, explainabili y, and
accoun abili y becomes mo e p onounced. This needs o be add essed.
In he con ex o sus ainabili y, AI- echnology esul s in a new ype o
digi al pollu ion in he o m o p i acy in usion, disc imina ion, and
biases. Une hical ou comes a e a esul o he applica ion, and no he
aul o he echnology, since he echnology i sel can be conside ed
neu al. Howe e , wi h he in en ion o d i e echnology as a o ce o
good comes also a esponsibili y o implemen sa egua ds and ensu e
he echnology is no misused. The consequences o ailing his espon-
sibili y a e se e e, esul ing in he amplifica ion o disc imina ion and
inequali y, in e ence o aul y o inco ec conclusions, and iola ion o
p i acy and us . The e a e ou common ypes o pi alls ha lead o
une hical ou comes:
•Imma u e da a and AI (insufficien aining o algo i hms on da ase s
as well as lack o ep esen a i e da a could lead o inco ec and
une hical ecommenda ions).
•Misuse / o e use o da a ( he AI applica ion, o solu ion, could be
o e ly in usi e using oo b oad o oo deep open da a o i could be
used o unin ended pu poses by o he s).
•Bias o he c ea o ( alues and bias a e in en ionally o unin en ion-
ally p og ammed by he c ea o who may also lack knowledge/skills
o how he solu ion could scale in a b oade con ex ).
•Da a bias ( he a ailable da a is no an accu a e eflec ion o eali y,
o he p e e ed eali y, and may lead o inco ec and une hical
ecommenda ions).
To c ea e a sus ainable app oach o AI and o unde s and i s impac
on people and socie y, AI- echnologies mus be assessed h ough a mul-
idisciplina y lens, combining he echnical, legal and socie al pe spec-
i es. Wi h new and un o eseen e en s such as he COVID-19 pandemic,
he Black-Li es-Ma e mo emen , o economic and job- ela ed c ises
(wi h de imen al effec s on e hical ques ions and s anda ds o ou so-
cie y), i is mo e impo an han e e o deploy da a-d i en echnology
in a sa e and us wo hy way.
T anspa ency, au oma ed decision-making p ocesses, and
pe sonal p ofiling
We li e in a big-da a-analy ics e a whe e au oma ed decision mak-
ing, machine lea ning and p ofiling echniques a e inc easingly able o
make assessmen s and p edic ions o indi iduals’ li es based on his o -
ical da a. These echniques can be ha m ul o indi iduals, because cu -
en laws and hei in e p e a ions nei he p o ide a da a subjec wi h
sufficien con ol o e he assessmen s made by au oma ed decision-
making p ocesses no wi h sufficien con ol o e how hese p ofiles a e
used. The main legal and e hical issues a e disc imina ion, biases, and
lack o anspa ency, and indi iduals ha e no access o hei p ofiles.
Consequen ly, people’s au onomy, digni y, and eedom a e in isk.
The e a e many examples o assessmen s being made based on online
au oma ed decision-making p ocesses:
An algo i hm decides whe he one s ays in o ou o he job ma -
ke wi h a simple pe sonali y es . The eby, one would no know ha
she/he is being disc imina ed agains , because i is he algo i hm ha
decides abou he ec ui men . Hence, people’s pe sonali y and job sui -
abili y a e assessessed based on a 30 s ideo. I he candida e does no
pass wi h a ce ain sco e, she/he may e en be ou o he labo ma ke
o good ( h ps://www.8andabo e.com/ ). Despi e such algo i hmic bi-
ases, AI-d i en decisions may be easily accep ed by humans, since he
ejec ion o AI-d i en decisions would be labo -in ensi e and edious.
Mo eo e , an AI-d i en classifica ion o job seeke s as e.g. “hopeless ”
has he po en ial o igge a p ocess whe e esou ce dep i a ion by he
public employmen se ice can ac ually lead o he ealiza ion, and la e
o he (ci cula ) alida ion, o his pa icula p edic ion [3] .
Insu ance companies collec da a om social ne wo ks o p edic
how much use s’ heal h ca e could cos hem.
2 Alexa can p edic a
use ’s heal h s a us h ough analysing oice and coughing, which is ol-
lowed by sending ad e isemen s o so e- h oa p oduc s.
3 Facebook
can p edic you poli ical iews ( h ps://www.ci izenme.com/ acebook-
p edic -poli ical- iew ), you ace, eligion, and sexual o ien a ion, and
e en when you a e going o die.
4 Is i ai o wonde how such au o-
ma ed decision-making and pe sonal p ofiling a e egula ed in he GDPR
(Gene al Da a P o ec ion Regula ion)? Is he e a “Righ o Explana ion ”
in he GDPR? Da a subjec igh s o anspa ency a e desc ibed in A i-
cles 13–15 GDPR. The igh o be no ified (A icles 13–14 GDPR) is a
da a con olle ’s du y and co e s da a p o ided di ec ly by he da a sub-
jec , obse ed da a and da a om a hi d pa y. Also, he igh o access
(A icle 15 GDPR) has o be appealed o by he da a subjec . Rega ding
he igh o be no ified, A icle 13.2 ( ) GDPR in o ms abou no ifica ion
equi emen s when pe sonal da a is collec ed di ec ly om he da a sub-
jec , and a icle 14.2 (g) GDPR in o ms abou no ifica ion equi emen s
when pe sonal da a is ob ained om a hi d-pa y. I has been sugges ed
ha he no ifica ion du ies ou lined in hese wo a icles g an an ex pos
explana ion, which means ha one has he igh o be no ified abou he
exis ence o he logic in ol ed, as well as he significance and he en is-
aged consequences o au oma ed decision-making. Bu his sugges ion is
w ong o a eason. These no ifica ion du ies p ecede decision-making
and apply in he momen da a is collec ed o p ocessing and e e only
o inpu da a.
Rega ding he igh o access, a icle 15.1 (h) GDPR says ha indi-
iduals ha e he igh o access hei pe sonal da a and o he ollowing
in o ma ion: whe he he e will be au oma ed decision-making, includ-
ing p ofiling, meaning ul in o ma ion abou he logic in ol ed in he
decision p ocess, as well as he impac and he en isaged consequences
o such p ocessing o he da a subjec . The eali y is ha , wi h a lack
o an explici deadline o appealing, he igh o access is limi ed o
explana ions o sys ems’ unc ionali ies. This is, again, an ex-an e expla-
na ion.
Fu he mo e, he T ade Sec e s Di ec i e s a es in i s a icle 2.1
5
ha
ade sec e is e e y hing ha is no known, i is any hing ha has com-
me cial alue and is any hing whe e easonable s eps a e aken o keep
i sec e . As a di ec consequence o his defini ion, an indi idual does
no ha e he igh o be no ified, as A icle 13.2 ( ) and 14.2 (g) o he
GDPR conside s, no he igh o access es ablished in A icle 15.1 (h)
2 h ps://choice.np .o g/index.h ml?o igin = h ps://www.np .o g/sec ions/heal h-
sho s/2018/07/17/629441555/heal h-insu e s-a e- acuuming-up-de ails-
abou -you-and-i -could- aise-you - a es? = 1544433231164
3 h ps://www. eleg aph.co.uk/ echnology/2018/10/09/amazon-pa en s-
new-alexa- ea u e-knows-offe s-medicine/
4 h ps://www.independen .co.uk/li e-s yle/gadge s-and-
ech/news/ acebook-pa en -p edic -die-dea h-p edic ion-algo i hm-pe sonal-
da a-p i acy-a8417771.h ml
5 h ps://eu -lex.eu opa.eu/legal-con en /EN/TXT/PDF/?u i = CELEX:32016L0943
3
S. Gup a, S.D. Langhans, S. Domisch e al. T anspo a ion Enginee ing 4 (2021) 100064
o he GDPR o he p ofile ha he companies ha e abou hem and he
pu pose i is used o . In o he wo ds, he GDPR does no con empla e
in an explici way a igh o an explana ion o he da a subjec o au-
oma ed decision-making, which makes he p inciples o anspa ency
and accoun abili y impossible o apply [4] .
Measu ing he ele ance o Digi aliza ion and a ificial
in elligence (D&AI) a he indica o le el o SDGs: Digi ainabili y
Applica ion o Digi aliza ion and AI (D&AI) o sus ainable de el-
opmen is being iewed as a majo mo emen shaping he economy,
en i onmen , and socie y. The union o he wo domains (Digi aliza-
ion and AI), pa icula ly unde he amewo k o SDG and D&AI is less
explo ed ye p omising, since i os e s cohe en effo s owa ds sus-
ainable de elopmen by p o iding new alue-gene a ing oppo uni ies
[18 , 37] . Howe e , due o he mul i ace ed na u e and unde lying com-
plexi ies o he SDG 2030 Agenda, effo s o explo e he meaning ulness
o a D&AI applica ion o SDGs a e s ill spa se [65] . Only a ew s udies
ha e so a in es iga ed he nexus be ween D&AI and he SDGs (e.g.,
[17 , 30 , 59] ), hus limi ing he possibili y o open deba es and a cohe -
en dialog be ween key-s akeholde s .
Conside ing he nume ous hu dles and unce ain ies ela ed o D&AI,
he Bonn Alliance o Sus ainabili y Resea ch/Inno a ion Campus Bonn
is wo king on he p ojec "digi ainable" o unco e he in ica e ela-
ionship be ween D&AI and sus ainabili y. In his con ex , "digi ainabil-
i y" - a combina ion wo d o he "digi aliza ion" and "sus ainabili y"
was amed, e e ing o he c oss- e iliza ion be ween he p ocesses
o digi aliza ion and sus ainable de elopmen . Resea ch is unde way o
comp ehend he ele ance o D&AI a he indica o le el o SDGs. Re-
sea ch which explo es associa ions among SDGs and D&AI om a p ac-
ical iewpoin and os e s he de elopmen o a ious quali a i e and
quan i ies me hods conside ing echnological as well as social pe spec-
i es, a e ongoing o de elop e idence o a D&AI and SDG nexus (e.g.,
[58 , 64] ). Howe e , syne gies and ade-offs among SDGs and hei indi-
ca o s a e c i ical aspec s ha need o be ca e ully assessed. Resea ch on
in e ac ions o p io i ize cohe en ac ions o sus ainable de elopmen
is apidly g owing (e.g., [2 , 22 , 61] ). Howe e , he e a e gaps in e ms
o iewpoin s o iden i ying he in e ac ions and asse ing agmen ed
ou comes, depending on he ocus and con ex . The e o e, o be e un-
de s and and gene a e da a-backed e idence, which is ye no well ec-
ognized and ha dly eflec ed in deba es, u he esea ch owa ds de-
li e ing sys emic-me hodologies is equi ed. Al hough s ill e ol ing, e-
sea ch on a mind ul use o D&AI is s a ing o shape. Fu he wo k and
unding a e needed o con inue explo ing he conscious and sus ainable
applica ions o D&AI echnologies o sus ainable ans o ma ion om
bo h social and echnological pe spec i es.
T ansdisciplina i y o sus ainabili y and AI applica ions
Si ua ions o sus ainable ans o ma ion equi e balancing sus ain-
able and unsus ainable ends, which a y spa io empo ally and com-
p ehend social and echnological ad ancemen s. Sus ainabili y as a con-
cep is d i en by alues, equi ing a ision o wha needs o be sus-
ained and a wha cos . We can obse e his by looking a he chal-
lenges in in eg a ing he h ee pilla s o sus ainabili y – economy, en i-
onmen , and socie y –which in some con ex s a e no mu ually com-
pa ible. Conside ing he need o sus ainable ans o ma ion wi hin he
cu en global change con ex , i is essen ial o unde s and he syn-
e gies and ade-offs be ween sus ainable and economic de elopmen ,
be ween en i onmen and economic de elopmen , economic de elop-
men and social con ex s, e c. The simplis ic, educ i e, and linea
logic behind disciplina y knowledge p oduc ion is limi ed in add ess-
ing complica ions beyond specific scope and me hods. In e p e a ion
challenges also exis be ween disciplines, seeking in en ions in iden-
i ying he key objec i es and hei sub le d i es o using D&AI and
he need o cos -effec i e app oaches o add ess he da a gaps [12] .
Thus, we should be conside ing a holis ic, sys emic, ans-disciplina y,
and o wa d-looking pe spec i e, bo h o D&AI and sus ainabili y
[5 , 43 , 57] . Sys emically quan ifiable and ansdisciplina y pe spec i es
a e needed o accele a ing he p og ess o SDGs, conside ing local and
global pe spec i es and e ical and ho izon al ela ions among key
s akeholde s.
A en ion owa ds sus ainabili y and D&AI applica ions
Achie ing he SDGs is a he isk o a conside able in e up ion due
o eme ging global challenges such as clima e change, en i onmen al
deg ada ion, o e use o mine als, poli ical and financial ins abili y, and
ecen ly he COVID-19 pandemic. SDGs could benefi significan ly om
D&AI suppo , i such suppo is delibe a ed and de eloped wi h hei
capabili ies and limi a ions in mind [66] . To ensu e ha echnological
in e en ions mee he en isaged end equi emen s, i is c ucial o iden-
i y and ealize i al ac o s ha explain he meaning ul applica ion o
echnologies. In o de o gene a e bo h a big pic u e and de ailed in-
sigh s abou he mind ul use o D&AI, p ac ical and esea ch implica-
ions need o be well conside ed. Pa icula ly, o cope wi h go e nance-
ela ed D&AI and sus ainabili y ac ion unce ain ies and con ex , no el
analy ical ools based on p ac ical implica ions o D&AI, conside ing ac-
o s’ pe spec i es, a e highly desi ed [44 , 47] . P ac ical implica ions also
need o demons a e ade-offs, p io i iza ions, and nego ia ions among
SDGs conside ing diffe en con ex s and objec i es. Fo example, he
p e en ion o conflic s be ween SDG indica o a ge s on ce ain occa-
sions may also hampe he implemen a ion o o he SDGs [63] . Some
conflic s may be necessa y, as o ins ance o achie e a long- e m in e-
g a ion as he ela ion be ween Indica o 9.4.1 (CO
2
emission pe uni
o alue added) and Indica o 9.2.1 (Manu ac u ing alue added as a
p opo ion o GDP and pe capi a) [34] . Exis ing li e a u e co obo a es
he impe a i e o measu ing he p og ess by moni o ing and e alua ing
he SDGs o achie emen [11 , 67] . Howe e , a significan challenge e-
mains in unco e ing sui able axonomies o measu e he dynamics o
p og ess and comp ehensi e me hodologies ha can help explo e he
con ex -based assessmen o p ac ical applica ions o D&AI a he in-
dica o le el o SDGs. Me hodologies and analy ical ools beyond goals
and a ge le els, i.e., hose ha a e capable o eflec ing digi ainabili y’s
quali a i e and quan i a i e aspec s while cap u ing he complexi ies o
he SDGs, a e needed o he mind ul applica ion o D&AI o sus ainable
ans o ma ion.
The p ojec “Digi ainable” is dedica ed o se e al o hese aspec s
o add ess gaps iden ified a he in e sec ion o D&AI and sus ainabili y
esea ch. No ably, he esea ch is ocused on add essing he ollowing
o e a ching hemes: 1. Iden i ying he ole D&AI could play o SDGs’
indica o s; 2. De eloping me hods ha help access he esponsible de-
ploymen o D&AI, conside ing bo h social and echnological pe spec-
i es; 3. Scalable D&AI in e en ions o SDGs; 4. Add essing he gaps
be ween esea ch and p ac ice o u ilize D&AI o sus ainable ans o -
ma ion meaning ully. This esea ch is designed o os e he con ex -
awa e, inclusi e, and accoun able applica ion o D&AI in he long- e m,
pa icula ly by enabling collabo a ion, us , and knowledge sha ing
wi h key s akeholde s. The eby, he Digi alisa ion Sus ainabili y Ma-
ix (DSM; [24] ), a pa icipa o y esea ch app oach, se es as a means
in collabo a i e me hods such as pa icipa o y ac ion esea ch (PAR)
o he ansdisciplina y knowledge cu a ion p ocess. PAR se es as a
amewo k o jux apose in e disciplina y s akeholde s in he knowledge
p oduc ion p ocess o co-de elop esea ch, educa ion, and ac ion in e -
linkages be ween science and p ac ice [10] . The DSM connec s D&AI
and sus ainabili y a he SDG indica o le el. I is a wo-dimensional ma-
ix using D&AI hemes and espec i e echnologies connec ed o indica-
o s o a pa icula SDG o seek he pe spec i e on posi i e, nega i e, and
non/unknown ele ance o he di e se s akeholde s who a e pa o he
PAR. The DSM has been es ed in in e ac i e e en s wi h expe s om
di e se sec o s and backg ounds, wi h he objec i e o cap u e ans-
disciplina y knowledge in he ac ion-o ien ed dialogues conce ning he
4
S. Gup a, S.D. Langhans, S. Domisch e al. T anspo a ion Enginee ing 4 (2021) 100064
use o D&AI echnologies o he indica o s o SDGs. DSM can effec-
i ely igge discussions on c ucial aspec s ha need o be conside ed
o D&AI’s p ac ices, which is a s ep owa ds deep- oo ing he ansdis-
ciplina y pe spec i es o meaning ul use o D&AI o SDGs (Gup a e al.
[24] ).
Designed by all o all
A wo ld which hea ily elies on AI echnology will ha e o deal wi h,
in addi ion o hose a o emen ioned, wo p onounced challenges specific
o gende in o de o p o ide equal ou comes and oppo uni ies o all.
The fi s one is fixing he gende da a gap. The gende da a gap is he
phenomenon whe eby he as majo i y o in o ma ion ha has so a
been collec ed globally is on men, anging om economic da a o u ban
planning da a o medical da a [38] . When such da a is used as a sou ce
o help c ea e ideal and unc ioning li ing en i onmen s, hese en i on-
men s only wo k op imally o a selec ed pa o he popula ion – he
men. Fo example, un il ecen ly clinical ials ha e no adequa ely en-
olled women o analyzed sex-specific diffe ences in he collec ed da a
[62] . This is despi e sex diffe ences can be obse ed in a ious disease
s a es in p e alence, diagnosis, se e i y, and ou comes as well as he di -
e en physiologies o he sexes may ansla e in o diffe ences in pha -
macokine ics o pha macodynamics o specific d ugs [33] . The mos
amous example is he one o Ambien, a d ug ea ing insomnia, which
was he fi s one o be p esc ibed in sex-specific doses a e disco e ing
ha women me abolized i much slowe and, he e o e, exhibi ed nex -
day psychomo o impai men wi h he o iginal dosage [21] . Ano he a-
mous example is ca sa e y. Ca s a e s ill mos ly es ed only using a 50 h-
pe cen ile male dummy [32] , esul ing in women d i e s being 47% and
71% mo e likely o be se iously and mode a ely inju ed, espec i ely,
in a ca c ash [7] . Hence, women ha e been disad an aged o millen-
nia. Now, consequen ly, any algo i hm ained on such male-domina ed
da ase s is unlikely o p edic accu a e isks o gi e app op ia e esul s
o e e yone. Examples o ha a e oice and speech ecogni ion sys-
ems, which pe o m wo se o women han o men [49 , 50] o ace
ecogni ion sys ems which p o ide mo e e o s wi h emale aces [9] .
Second, we need o each gende equali y in AI p o essionals. Women
a e highly unde ep esen ed in STEM disciplines (Science, Technol-
ogy, Enginee ing, and Ma hema ics) and he AI field is also a male-
domina ed one. In 2016, women accoun ed o less han a hi d (29.3%)
o hose employed in scien ific esea ch and de elopmen ac oss he
wo ld [56] and only 19.5% o hose held manage posi ions in he so -
wa e echnology indus y [28] . Sys ems and algo i hms c ea ed by male
designe s and enginee s migh en ail biases which a e in oduced uncon-
sciously du ing a ious de elopmen s ages e.g. he sampling o da a,
hei anno a ion, he selec ion o algo i hms, o he e alua ion o me ics
[52] . The consequence a e algo i hms which p ocess da a in a gende -
biased way. Clea ly, da a gaps and algo i hm biases do no only hinde
eaching gende equali y, bu also apply o cul u e and ace. One o he
ew s udies on his opic ound la ge acial dispa i ies in he pe o mance
o fi e comme cial au oma ic speech ecogni ion sys ems, when es ed
by whi e and A ican Ame ican speake s [29] .
Conclusion and ou look
This a icle comp ehensi ely p esen s he mul idisciplina y pe spec-
i es sha ed in he panel discussion by expe s on he in e sec ion o
AI and sus ainable de elopmen . We shed ligh in o he a ious c i ical
aspec s abou whe he , how, and wha necessa y ace s need o be con-
side ed o esponsi ely ha nessing he con enience offe ed by AI o
SDGs. O e all, we p o ide a gene al o e iew o iden i ying and p i-
o i izing he agenda o add essing he p essing esea ch gaps conce n-
ing he ad ancemen o AI o sus ainable ans o ma ion. The signifi-
cance o he nexus be ween he applica ion o AI and he 2030 Agenda
could be sensed as an “endu ing componen ”, which equi es a ho -
ough conside a ion o he h ea s, oppo uni ies, syne gies, and ade-
offs. Thus, a con e ging ans o ma ion is desi ed om mul idisciplina y
pe spec i es aking he SDGs as ou baseline. The dis up ion caused by
AI goes beyond echnology, ep esen ing an in e play be ween echnol-
ogy, socie y, en i onmen , and policy. A be e unde s anding o he
socio- echnical enable s and hei impac on socie y a e equally impo -
an ac o s. Gi en he excep ional p ospec s ha AI may b ing in, we
sugges ha in-dep h explo a ion o he ole o AI o p o ide suppo
and o e come specific limi a ions hinde ing he p og ess o he SDG is
c ucial.
Va ious expe s in he panel discussion highligh ed an u gen need o
add ess gaps, which helps in managing he ansi ion o an AI-assis ed
socie y. Accele a ing p og ess is u gen ly equi ed o manage he sig-
nifican ulne abili y o in as uc u es pe aining o AI. No ably, he
ulne abili ies a ound anspa ency, sa e y, and e hical s anda ds need
o be add essed so ha e e yone can con ibu e wi h su eness and can
benefi om he po en ial ad an ages offe ed by AI. Fu he esea ch is
equi ed o a p o ound unde s anding o he iden ified challenges ak-
ing he SDGs as ou baseline. SDGs can play he ole o a amewo k o
guide he ad ancemen s in AI as a key acili a o o da a-d i en goals
and en ich u he oppo uni ies. Conside ing he new and un o eseen
si ua ions ha impac ed socie y and e hical effec s such as COVID-19,
i is mo e c ucial han e e o deploy sa e and us wo hy da a-d i en
echnologies o sus ainable de elopmen . Fu he mo e, esea ch is e-
qui ed o p o oundly explo e he ex en o which AI can con ibu e o
he agenda o sus ainable de elopmen as well as o ackle po en ial
adeoffs.
Taking in o accoun he unin ended social, e hical, equi y- ela ed,
and en i onmen al conce ns, he e is an u gen need o assess and cu b
social as well as echnical easibili ies o enabling AI-d i en solu ions
o sus ainable ans o ma ion. Me hodologies and analy ical ools o
access he ele ance o AI o sus ainable de elopmen need o be pu -
sued om a sys ema ic and e hical pe spec i e, beyond cu en echno-
cen ic iewpoin s. In-dep h unde s anding beyond goals and a ge le -
els ha a e capable o eflec ing digi ainabili y’s quali a i e and quan i-
a i e aspec s, while cap u ing he complexi ies o he SDGs, a e highly
desi ed o acili a e AI o sus ainable ans o ma ion esponsi ely. As
can be obse ed om he a o emen ioned sec ions, AI discussions ha e
been mainly ocused on e hical and social conside a ions. Howe e ,
since ul ima ely only s aying wi hin he plane a y bounda ies will al-
low o a ansi ion in o a p ospe ous and equi able u u e o all, en-
i onmen al sus ainabili y needs a p ominen place a he o e on o
de eloping and implemen ing AI-d i en echnologies. In pa icula , he
ole o AI in enabling clima e ac ion should be u he in es iga ed, as
clima e change is widely conside ed as he mos p ominen c isis o ad-
d ess in he 21s cen u y, and has wide ipple effec s ac oss all SDGs
(Ne ini [19] ).
Fo p ac i ione s o AI, cu en esea ch on he impac s o AI sug-
ges s he need o go beyond he simple unde s anding o wha migh be
he impac s o an AI applica ion wi hin i s field, bu also unde s anding
i s po en ial posi i e and nega i e effec s ac oss socie al, en i onmen al
and economic ou comes. The SDGs could be used o guide ha e al-
ua ion, as a mul i-disciplina y lens o guide he dialog [59] . Fo ha
o happen, he esea ch and in e na ional communi y should de elop
easy- o-use s anda d ools o guide AI p ac i ione s in such assessmen .
The e is a need o os e dialog and collabo a ion ac oss disciplines
o d aw insigh s ha can accele a e esponsible, e hics-d i en, inclusi e,
and con ex -awa e AI applica ions o sus ainable de elopmen . Hence,
we belie e his a icle p o ides a s ep owa ds a p o ound unde s anding
o he challenges and c i ical aspec s o u ilizing AI o a p ospe ous 21s
cen u y.
Decla a ion o Compe ing In e es
The au ho s decla e ha hey ha e no conflic o in e es .
5
S. Gup a, S.D. Langhans, S. Domisch e al. T anspo a ion Enginee ing 4 (2021) 100064
Acknowledgmen s
The au ho s acknowledge he KTH Sus ainabili y Office and he
KTH Digi aliza ion Pla o m o hei p o ided unding, which enabled
he o ganiza ion o his panel discussion. SG acknowledges he und-
ing p o ided by he Ge man Fede al Minis y o Educa ion and Re-
sea ch (BMBF) o he p ojec “digi ainable ”. SDL acknowledges suppo
h ough he Spanish Go e nmen ’s Ma ía de Maez u excellence acc ed-
i a ion 2018–2022 (Re . MDM-2017–0714). SD acknowledges unding
by he Leibniz Compe i ion (J45/2018).
Re e ences
[1] D. Acemoglu , P. Res epo , A ificial In elligence, Au oma ion, and Wo k. ”, NBER
Na ional Be eau o Economic Resea ch, 2018 Wo king Pape No. 24196 .
[2] C. Allen , G. Me e nich , T. Wiedmann , M. Pede cini , G ea e gains o Aus alia by
ackling all SDGs bu he las s eps will be he mos challenging, Na . Sus ain. 2 (11)
(2019) 1041–1050 .
[3] D. Allhu e , F. Cech, F. Fische , G. G ill, A. Mage , Algo i hmic p ofiling o job
seeke s in Aus ia: how aus e i y poli ics a e made effec i e, F on . Big Da a 3 (2020)
5, doi: 10.3389/ da a .
[4] M. Ba aglini , S. Rasmussen
, T anspa ency, au oma ed decision-making p ocesses
and pe sonal p ofiling, J. Da a P o . P i . 2 (2019) 331–341 .
[5] M.U. Ben-Eli , Sus ainabili y: defini ion and fi e co e p inciples, a sys ems pe spec-
i e, Sus ain. Sci. 13 (5) (2018) 1337–1343 .
[6] P. Be g , D. Bal imo e , S. B enne , R.O. Roblin , M.F. Singe , Summa y s a emen
o he Asiloma con e ence on ecombinan DNA molecules, PNAS 72 (6) (1975)
1981–1984 June .
[7] D. Bose , M. Segui-Gomez , J.R. C andall , Vulne abili y o emale d i e s in ol ed in
mo o ehicle c ashes: an analysis o US popula ion a isk, Am. J. Public Heal h 101
(12) (2011) 2368–2373 .
[8] E. B ynjol sson , A. McA ee , The Second Machine Age: Wo k, P og ess, and P ospe i y
in a Time o B illian Technologies, W. W. No on & Company, 2014 .
[9] J. Buolamwini , T. Geb u , Gende shades: in e sec ional accu acy dispa i ies in com-
me cial gende classifica ion, P oc. Mach. Lea n. Res. 81 (2018) 1–15 .
[10] J.M. Che alie , D.J. Buckles , Pa icipa o y Ac ion Resea ch: Theo y and Me hods o
Engaged Inqui y, Rou ledge, N.p, 2019 .
[11] S. Dalby , S. Ho on , R. Mahon , D. Thomaz , Achie ing he Sus ainable De elopmen
Goals: Global Go e nance Challenges, Rou ledge, N.p, 2019
.
[12] Hai-Anh H. Dang , U. Se ajuddin , T acking he Sus ainable De elopmen goals:
Eme ging measu emen Challenges and Fu he Reflec ions, The Wo ld Bank, N.p.,
2019 .
[13] Eu opean Union. 2019. “Eu opean Pa liamen a a glance. a ificial in-
elligence, da a p o ec ion and elec ions. ” h ps://www.eu opa l.eu opa.eu/
RegDa a/e udes/ATAG/2019/637952/EPRS_ATA(2019)637952_EN.p.
[14] M. Fleu y, How a ificial in elligence is ans o ming he finan-
cial indus y, BBC News (2015) (New Yo k), Sep embe 16, 2015
h ps://www.bbc.com/news/business-34264380 .
[15] D. F ancesca o , Globaliza ion, a ificial in elligence, social ne wo ks and poli ical
pola iza ion: new challenges o communi y psychologis s, Commun. Psychol. Glob.
Pe spec . 4 (2018) 20–41 .
[16] D.H. F eedman , Hun ing o new d ugs wi h AI,
Na u e 576 (2019) S49–S53 .
[17] Global Enabling Sus ainabili y Ini ia i e. 2019. “Digi al wi h Pu pose. ” 2019.
[18] R. Gou ea , D. Kapelianis , S. Kassicieh , Assessing he nexus o sus ainabili y and in-
o ma ion & communica ions echnology, Technol. Fo ecas Soc. Change 130 (2018)
39–44 .
[19] F. Ne ini , e al. , Connec ing clima e ac ion wi h o he Sus ainable De elopmen
Goals, Na . Sus ain. 2 (2019) 674–680 .
[20] A. G eenbe g, A boeing code leak exposes secu i y flaws
deep in a 787 ′ s gu s, WIRED (2019) July 8, 2019
h ps://www.wi ed.com/s o y/boeing-787-code-leak-secu i y-flaws/ .
[21] D.J. G eenbla , J.S. Ha ma z, N.N. Singh, F. S einbe g, T. Ro h, M.L. Moline,
S.C. Ha is, R.P. Kapil, Gende diffe ences in pha macokine ics and pha macody-
namics o zolpidem ollowing sublingual adminis a ion, J. Clin. Pha macol. 54
(2014) 282–290, doi: 10.1002/jcph.220 .
[22] D.J. G iggs , M. Nilsson , A. S e ance , M. Da id , e al. , A Guide o SDG in e ac ions:
om Science o Implemen a ion, In e na ional Council o Science, N.p.Pa is, 2017 .
[23] Guas oni, L., Güemes, A., Iani o, A., Disce i, S., Schla e , P., Azizpou , H., and Vin-
uesa, R.. 2020. “Con olu ional-ne wo k models o p edic wall-bounded u bulence
om wall quan i ies. ”a Xi p ep in a Xi :2006.12483 .
[24] S. Gup a, M. Mo lagh, J. Rhyne , The digi aliza ion sus ainabili y ma ix: a pa ici-
pa o y esea ch ool o in es iga ing digi ainabili y, Sus ainabili y 12 (21) (2020)
9283 (No embe ), doi: 10.3390/su12219283 .
[25] N. Jean , e al. , Combining sa elli e image y and machine lea ning o p edic po e y,
Science 353 (2016) 790–794 .
[26] N. Jones , How o s op da a cen es om gobbling up he
wo ld’s elec ici y, Na u e
561 (2018) 163–166 .
[27] N. Kal a, S.M. Paddock, D i ing o sa e y: How Many Miles o D i ing Would i
Take o Demons a e Vehicle Reliabili y?, RAND Co po a ion, San a Monica, CA,
2016 h ps://www. and.o g/con en /dam/ and/pubs/ esea ch_ epo s/RR1400/
RR1478/RAND_RR1478.pd .
[28] Ke sley, R., Kle k, E., Boussie, A., Longwo h, B.S., Na zkoff, J.A., and Ramji, D..
2019. “The CS gende 3000 in 2019: he changing ace o companies. ”p. 15.
h ps://www.c edi -suisse.com/abou -us-news/en/a icles/news-and-expe ise/cs-
gende -3000- epo -2019-201910.h ml .
[29] A. Koenecke, A. Nam, E. Lake, J. Nudell, M. Qua ey, Z. Mengesha, C. Toups,
J.R. Rick o d, D. Ju a sky, S. Goel, Racial dispa i ies in au oma ed speech ecog-
ni ion, PNAS 117 (14) (2020) 7684–7689, doi: 10.1073/pnas.1915768117 .
[30] O. Kos oska , L. Koca e , A no el ICT amewo k o Sus ainable De elopmen Goals,
Sus ainabili y 11 (7) (2019) 1961 .
[31] J. Lelie eld , e al. , Ca dio ascula disease bu den om ambien ai pollu ion in
Eu ope eassessed using no el haza d a io unc ions, Eu . Hea J. 40 (2019)
1590–1596 .
[32] A. Linde , W. S edbe g, Re iew o a e age sized male and emale occu-
pan models in Eu opean egula o y sa e y assessmen es s and Eu opean
laws: gaps and b idging sugges ions, Accid. Anal. P e . 127 (2019) 156–162,
doi: 10.1016/j.aap.2019.02.030 .
[33] K.A. Liu , N.A. Dipie o Mage , Women’s in ol emen in clinical ials: his o ical
pe spec i e and u u e implica ions, Pha m. P ac . (G anada) 14 (1) (2016) 708
10.18549/Pha mP ac .2016.01.708 .
[34] S. Liu, In e linkages be ween indica o s o Sus ainable De elopmen Goals: e idence
om se en low income and lowe middle-income coun ies, Sus ain. De . Res. 2
(2020) 1June, doi: 10.30560/sd . 2n1p58 .
[35] W. Naudé, R. Vinuesa , Da a, Global de elopmen , and COVID-19: Lessons and Conse-
quences, Wo ld Ins i u e o De elopmen Economic Resea ch
(UNU-WIDER), 2020
No. wp-2020-109 .
[36] D.B. Neill , H.J. Heinz , Using a ificial in elligence o imp o e hospi al inpa ien ca e,
IEEE 28 (2) (2013) 92–95 .
[37] T. Osbu g , C. Loh mann , Sus ainabili y in a Digi al Wo ld, Sp inge , N.p., 2017 .
[38] C.C. Pe ez , In isible Women, Penguin Random House, London, 2019 .
[39] M. Pe i , Towa ds a c i ique o algo i hmic eason. A s a e-o - he-a e iew o a -
ificial in elligence, i s influence on poli ics and i s egula ion, Quad. Del CAC 44
(2018) .
[40] S. Ra ind an, How a ificial in elligence is helping o p e en blindness. ”, Na . Ou -
look The Eye.
(2019), doi: 10.1038/d41586-019-01111-y .
[41] L.P. Robe , A e au oma ed ehicles sa e han manually d i en ca s? AI Soc. 34
(2019) 687–688 .
[42] N. Sa age , Ano he se o eyes o cance diagnos ics, Na . Ou look 579 (2020)
S14–S16 .
[43] R.W. Scholz , E.J. Ba elsman , S. Die enbach , L. F anke , A. G unwald , D. Helbing ,
R. Hill , e al. , Unin ended side effec s o he digi al ansi ion: Eu opean scien is s’
messages om a p oposi ion-based expe ound able, Sus ainabili y 10 (6) (2018)
2001 .
[44] M.S. Smi h , C. Cook , Y. Sokona , T. Elmq is , K. Fukushi
, W. B oadga e , M.P. Ja zeb-
ski , Ad ancing sus ainabili y science o he SDGs, Sus ain. Sci. 13 (6) (2018)
1483–1487 .
[45] SRC, S ockholm Resilience Cen e (SRC)Con ibu ion o he 2016 Swedish 2030
Agenda HLPF Repo , S ockholm Uni e si y, 2017 .
[46] P.A. S ini asan , L. Guas oni , H. Azizpou , P. Schla e , R. Vinuesa , P edic ions o u -
bulen shea flows using deep neu al ne wo ks, Phys. Re . Fluids 4 (2019) 054603 .
[47] K. Sulli an , S. Thomas , M. Rosano , Using indus ial ecology and s a egic manage-
men concep s o pu sue he Sus ainable De elopmen Goals, J. Clean. P od. 174
(2018)
237–246 .
[48] W. Sun , P. Bocchini , B.D. Da ison , Applica ions o a ificial in elligence o disas e
managemen , Na . Haza ds 2020 (3) (2020) 1–59 .
[49] R. Ta man , Gende and dialec bias in YouTube’s au oma ic cap ions, in: P oceedings
o he Fi s Wo kshop on E hics in Na u al Language P ocessing, 2017, pp. 53–59 .
[50] R. Ta man, C. Kas en, Effec s o alke dialec , gende & ace on accu acy
o Bing Speech and YouTube au oma ic cap ions, In e speech (2017) 934–938,
doi: 10.21437/In e speech.2017-1746 .
[51] D.S. Ting , Y. Liu , P. Bu lina , X. Xu , N.M. B essle , T.Y. Wong , AI
o medical imaging
goes deep, Na . Med. 24 (5) (2018) 539–540 .
[52] S. Tolan , Fai and unbiased algo i hmic decision making: cu en s a e and u u e
challenges, JRC Tech. Rep. (2018) .
[53] P. To es , S.Le Clainche , R.. Vinuesa , On he expe imen al, nume ical and da a–
d i en me hods o
s udy u ban flows, Ene gies 14 (2021) 1310 .
[54] UN, Uni ed Na ions Economic and Social Council. 2019. “Sus ainable De elopmen . ”
[55] UN Gene al Assembly (UNGA). 2015. A/RES/70/1T ans o ming ou wo ld: he 2030
Agenda o sus ainable de elopmen . N.p.: Resolu . 25, 1–35.
[56] UNESCO Ins i u e o S a is ics. 2019. “Fac shee No. 55: women
in science. ”
p. 2. h p://uis.unesco.o g/si es/de aul /files/documen s/ s55-women-in-science-
2019-en.pd .
[57] S.C. Valencia , D. Simon , S. C oese , J. No dq is , M. Oloko , T. Sha ma , N.T. Buck ,
I. Ve sace , Adap ing he Sus ainable De elopmen Goals and he new u ban agenda
o he ci y le el: ini ial eflec ions om a
compa a i e esea ch p ojec , In . J. U b.
Sus ain. De . 11 (1) (2019) 4–23 .
[58] M. an de Velden , ICT and sus ainabili y: looking beyond he an h opocene, in:
P oceedings o he IFIP In e na ional Con e ence on Human Choice and Compu e s,
Sp inge , 2018, pp. 166–180. N.p .
[59] R. Vinuesa, H. Azizpou , I. Lei e, M. Balaam, V. Dignum, S. Domisch, A. Fellän-
de , S.D. Langhans, M. Tegma k, F.F. Ne ini, The ole o a ificial in elligence in
achie ing he Sus ainable De elopmen Goals, Na . Commun. 11 (233) (2020) o g/,
doi: 10.1038/s41467-019-14108 .
[60] R. Vinuesa , A. Theodo ou , M. Ba aglini , V. Dignum , A socio- echnical
amewo k
o digi al con ac acing, Resul s Eng. 8 (2020) 100163 .
[61] N. Wei z , Towa ds sys emic and con ex ual p io i y se ing o implemen ing he
2030 agenda, Sus ain. Sci. 13 (2) (2018) 531–548 (Sp inge ) .
[62] H.P. Whi ley , W. Lindsey , Sex-based diffe ences in d ug ac i i y, Am. Fam. Physician
80 (11) (2009) 1254–1258 .
6
S. Gup a, S.D. Langhans, S. Domisch e al. T anspo a ion Enginee ing 4 (2021) 100064
[63] R. Wong , J. an de Heijden , A oidance o conflic s and ade-offs: a challenge o
he policy in eg a ion o he Uni ed Na ions Sus ainable De elopmen Goals, Sus ain.
De . 27 (5) (2019) 838–845 .
[64] J. Wu , S. Guo , H. Huang , W. Liu , Y. Xiang , In o ma ion and communica ions ech-
nologies o sus ainable de elopmen goals: s a e-o - he-a , needs and pe spec i es,
IEEE Commun. Su . Tu o . 20 (3) (2018) 2389–2406 .
[65] T. Yigi canla , F. Cugu ullo, The sus ainabili y o a ificial in elligence: an u banis ic
iewpoin om he lens o sma and sus ainable ci ies, Sus ainabili y 12 (20) (2020)
1–24 (Swi ze land), doi: 10.3390/su12208548
.
[66] T. Yigi canla , K.C. Desouza, L. Bu le , F. Roozkhosh, Con ibu ions and isks o a -
ificial in elligence (AI) in building sma e ci ies: insigh s om a sys ema ic e iew
o he li e a u e, Ene gies 13 (6) (2020), doi: 10.3390/en13061473 .
[67] A. Yoneha a, O. Sai o, K. Hayashi, M. Nagao, R. Yanagisawa, K. Ma suyama, The
ole o e alua ion in achie ing he SDGs, Sus ain. Sci. 12 (6) (2017) 969–973,
doi: 10.1007/s11625-017-0479-4 .
7
