Benke, I o e al.
A icle — Published Ve sion
Hyb id Adap i e Sys ems
Business & In o ma ion Sys ems Enginee ing
P o ided in Coope a ion wi h:
Sp inge Na u e
Sugges ed Ci a ion: Benke, I o e al. (2024) : Hyb id Adap i e Sys ems, Business & In o ma ion
Sys ems Enginee ing, ISSN 1867-0202, Sp inge Fachmedien Wiesbaden GmbH, Wiesbaden, Vol. 66,
Iss. 2, pp. 233-247,
h ps://doi.o g/10.1007/s12599-024-00861-y
This Ve sion is a ailable a :
h ps://hdl.handle.ne /10419/315718
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 p://c ea i ecommons.o g/licenses/by/4.0/
DISCUSSION
Hyb id Adap i e Sys ems
I o Benke •Michael Knie im •Ma c Adam •Michael Beigl •Ve ena Do ne •
Ul ich Ebne -P ieme •Man ed He mann •Ma in Kla mann •
Alexande Maedche •Julia Na zige •Pe a Nieken •Jella P ei e •
Clemens Puppe •Felix Pu ze •Benjamin Scheibehenne •Tanja Schul z •
Ch is o Weinha d
Published online: 6 Ap il 2024
ÓThe Au ho (s) 2024
1 Adap i e Sys ems – The Need o an In e disciplina y
Pe spec i e
I o Benke, Alexande Maedche, Jella P ei e , Ch is o
Weinha d
Adap i e sys ems based on in o ma ion echnologies (IT)
and a i icial in elligence (AI) capabili ies ha e become
ubiqui ous in almos all a eas o ou p i a e and business
li es. Think o social media eeds ha adjus hei con en
o ou beha io o obo-ad iso s ha adap hei buying
beha io acco ding o he s ock ma ke and ou inancial
p e e ences. Simila ly, i ness, li es yle, and heal hca e
apps con inuously ack humans o moni o and assess
pe sonal i ness, nu i ion, and heal h s a us in o de o
adap con ex -speci ic ecommenda ions (Alawneh e al.
2023; Noo be gen e al. 2019). Adap i e sys ems a e also
inding hei way in o companies, e.g., shi plans dynam-
ically adap o he equi emen s o manu ac u ing si es such
I. Benke (&)M. Knie im M. Beigl U. Ebne -P ieme
M. Kla mann A. Maedche P. Nieken C. Puppe
B. Scheibehenne C. Weinha d
Ka ls uhe Ins i u e o Technology, Ka ls uhe, Ge many
e-mail: [email p o ec ed]
M. Knie im
e-mail: [email p o ec ed]
M. Beigl
e-mail: [email p o ec ed]
U. Ebne -P ieme
e-mail: [email p o ec ed]
M. Kla mann
e-mail: [email p o ec ed]
A. Maedche
e-mail: [email p o ec ed]
P. Nieken
e-mail: [email p o ec ed]
C. Puppe
e-mail: [email p o ec ed]
B. Scheibehenne
e-mail: [email p o ec ed]
C. Weinha d
e-mail: [email p o ec ed]
M. Adam
Uni e si y o Newcas le, Newcas le, Aus alia
e-mail: [email p o ec ed]
V. Do ne
Vienna Uni e si y o Economy and Business, Vienna, Aus ia
e-mail: [email p o ec ed]
M. He mann F. Pu ze T. Schul z
Uni e si y o B emen, B emen, Ge many
e-mail: [email p o ec ed]
F. Pu ze
e-mail: [email p o ec ed]
T. Schul z
e-mail: [email p o ec ed]
J. Na zige
Aa hus Uni e si y, Aa hus, Denma k
e-mail: [email p o ec ed]
J. P ei e
Uni e si y o Giessen, Giessen, Ge many
e-mail: [email p o ec ed].de
123
Bus In Sys Eng 66(2):233–247 (2024)
h ps://doi.o g/10.1007/s12599-024-00861-y
as p oduc ion demand a a gi en ime o blue-colla
wo ke s’ physical abili ies.
O e all, adap i e sys ems hold eno mous po en ial o
c ea e indi idual, o ganiza ional, and socie al bene i s.
Howe e , hei ad anced adap i e IT capabili ies may be
con o e sial since hey also ep esen social, e hical, and
economic h ea s o indi iduals, g oups, o ganiza ions, and
he en i e socie y. Fo example, hey can lack anspa ency
on how hey a i e a hei adap a ion beha io (e.g., o
inancial ad ice), o hey can os e addic i e usage
beha io s h ough highly engaging in e aces (e.g., on
social media apps). Besides he in o ma ion sys ems (IS)
discipline, o he disciplines such as economics, manage-
men , psychology and neu oscience, and compu e science
a e esea ching adap i e sys ems om a social and ech-
nological poin o iew. In doing so, hey ha e de eloped
di e en pe spec i es on his socio- echnical phenomenon,
wi h a di e se unde s anding o he concep , i s cha ac-
e is ics, and i s bounda ies. Gi en he impo ance and
ubiqui y o adap i e sys ems in ou li es, he e is a need o
a sha ed unde s anding o guide u u e esea ch on adap i e
sys ems ac oss single disciplina y bounda ies o con ibu e
o he g ea e bene i o indi iduals, g oups, o ganiza ions,
and socie y.
S epping back in ime, he concep o adap a ion o igi-
na es om he ield o biology. The amous e olu ionis
Cha les Da win desc ibed he concep o Da winian
adap a ion as an o ganism’s ea u e ha was unc ionally
designed by he selec ion p ocess o e olu ion ac ing in
na u e (Tho nhill 1997). In gene al, i desc ibes an o gan-
ism’s abili y o adap i s cha ac e is ics and beha io in an
en i onmen wi hin a gi en ime ame by na u al selec-
ion. Wi h he apid p og ess o IT and he p oli e a ion o
adap i e sys ems, a ange o di e en disciplines ha e
adop ed he gene ic adap a ion concep om biology,
o med hei own discipline-speci ic unde s anding o
adap i e sys ems, and employed i wi hin hei subjec s o
in es iga ion and hei disciplina y knowledge con ibu-
ions. Unsu p isingly, howe e , he espec i e disciplines
ha e de ined he concep o adap i e sys ems e y di e -
en ly depending on hei own pe spec i es and sel -image.
Agains his backd op, he goal o his discussion a icle
is o highligh he di e en pe spec i es o he abo e-
men ioned disciplines when in es iga ing adap i e sys ems
om a social and echnological poin o iew. I becomes
ob ious ha he di e en disciplines, depending on hei
indi idual oo s, no su p isingly place mo e impo ance on
ei he he social o echnical dimensions o adap i e sys-
ems. Building on his insigh , we p opose a i s s ep
owa ds a sha ed unde s anding o hyb id adap i e sys ems
ha equally balances he social and he echnological
dimension. This concep ualiza ion may lay he ounda ion
o es ablishing an in e disciplina y esea ch pe spec i e
o his socio- echnical phenomenon. We a gue ha in e -
disciplina y esea ch on hyb id adap i e sys ems will ul i-
ma ely con ibu e o he g ea e bene i o indi iduals,
g oups, o ganiza ions, and socie y.
2 Managemen
Pe a Nieken, Ma in Kla mann
Managemen sciences con ibu e o he knowledge o
adap i e sys ems by s udying hei implica ions on o ga-
niza ions (i.e., ins i u ions o companies) – which a e an
impo an elemen in ou li es. Speci ically, mos o us
expe ience o ganiza ions in a leas wo oles: as employees
and as cus ome s. To illumina e a managemen pe spec i e
on adap i e sys ems, we, he e o e, u n in he ollowing o
Human Resou ce Managemen and Ma ke ing as wo
managemen disciplines ha ocus on humans in he ole o
employees and cus ome s.
2.1 Human Resou ce Managemen Pe spec i e
Adap i e sys ems challenge classical managemen pa a-
digms and call o a ho ough in es iga ion o hyb id
managemen p ac ices, since hei implica ions o man-
age s and o ganiza ions a e mani old. Adap i e sys ems
wi h a sys em-d i en adap a ion need a clea and ans-
pa en o mula ion o objec i es and goals. Gi en ha he
o ganiza ion’s and i s membe s’ goals o en di e , his is
no an easy and s aigh o wa d ask. How do o ganiza ions
balance con lic ing goals and pe spec i es and align hem
wi hin an adap i e IT sys em? Wi hou anspa ency and
clea guidelines, such sys ems will be p one o manipula-
ions and po en ial biases leading o dis us and ai ness
conce ns. Howe e , e en wi h clea objec i es and goals,
humans will adap o he adap i e IT sys em and y o
challenge he unde lying mechanisms. Thus, we will
obse e a co-e olu ion and co-e ol emen o humans and
echnology in adap i e sys ems in managemen . Today, we
see a apid in eg a ion o echnological de ices and assis-
ance sys ems eco ding biosignals in o de o be e
unde s and and assis humans in he wo kplace. I is o
u mos impo ance o unde s and he unde lying mecha-
nisms and beha io al e ec s o help design e icien and
ai adap i e (managemen ) sys ems.
Human Resou ce Managemen ocuses on he human
and aims o es ablish a mo i a ing, sa e, and inspi ing wo k
en i onmen . Sys ems ha adap o indi idual needs can
help ul illing his goal. On he o he hand, handing o e
con ol o an adap i e sys em can spa k ea and educe
us in manage ial decisions. Conside , o ins ance, he
co e o ganiza ional asks pe o mance e alua ion and
123
234 I. Benke e al.: Hyb id Adap i e Sys ems, Bus In Sys Eng 66(2):233–247 (2024)
leade ship. Adap i e sys ems can po en ially educe human
bias and p ocess la ge amoun s o in o ma ion leading o a
ai e pe o mance e alua ion. By combining pe o mance
da a wi h physical o psychological pa ame e s o he
employees, hese sys ems can e alua e employees and
ma ch hem wi h he mos i ing posi ion (see e.g.,
Ba holomeyczik e al. 2022 o he usage o EEG in wo k
con ex s). Howe e , employees migh eel inclined o al e
hei beha io s a egically (Haus eld e al. 2021). Adap i e
sys ems can also assis leade s o ailo hei communica-
ion o he indi idual’s needs (see, e.g., Nieken 2022b and
Nieken 2022a). Gi en ha leade ship is ypically seen as a
ask ha needs o be assigned o a human, being e alua ed
and led ‘‘by an algo i hm’’ migh lead o lowe employee
engagemen and us in he o ganiza ion. Thus, ans-
pa ency and a clea de ini ion o goals, powe , and decision
igh s is c ucial in human esou ce managemen o ensu e
he accep ance o adap i e sys ems.
2.2 Ma ke ing Pe spec i e
In ma ke ing, adap i e sys ems play an e e -g owing ole.
A p ominen example e e yone is amilia wi h a e ec-
ommenda ion engines. Online e aile s lea n he p e e ence
s uc u es o hei cus ome s om hei sea ch beha io and
ea lie pu chases. Based on hese p e e ences, hey cus-
omize which p oduc s he cus ome sees in a gi en sea ch
and in wha o de . Ano he example is ‘‘p og amma ic
ad e ising’’ whe e algo i hms bid o he possibili y o
showing ce ain con en o po en ial cus ome s. The bids
a e based on an ad’s expec ed pe o mance, gi en he
po en ial cus ome ’s cu en sea ch his o y (among o he s).
Mo e will become possible. Fo ins ance, we ha e ound
ha i is possible o p edic cus ome mood based on small
audio snippe s, which could make oice shopping mood-
adap i e (Halbaue and Kla mann 2022).
I is a key ene o ma ke ing heo y (and well sub-
s an ia ed h ough empi ical esul s) ha ma ke ing (and
companies) will only be success ul i hei o e ings sa is y
cus ome needs. Exagge a ed claims (o wo se, manipula-
ion, (Kla mann 2020)) can easily back i e. Fo he design
o adap i e sys ems ha a e o be used by cus ome s, his
implies ha – a hei co e – hey need o be adap i e o
cus ome needs. The e o e, om a ma ke ing pe spec i e,
adap i e sys ems need o ul ill a numbe o equi emen s
ha may di e (a leas in e ms o hei p io i iza ion).
Technically, hey should be based on alida ed measu es
and causal in e ences (ins ead o me e p edic ions).
Rega ding he measu es used, human–compu e in e -
aces o en ely on measu es ha a e ela i ely a om
measu ing he needs ha eally d i e cus ome beha io
(e.g., eye mo emen s, mouse mo emen s, physiological
eac ions). The isk ha comes wi h hese measu es is ha
– al hough hey may be co ela ed o needs (which would
need o be es ablished) – hey a e likely o be con ounded
wi h a numbe o o he a iables (and possibly e en o he
needs). F om he cus ome s’ pe spec i es, his may c ea e
su p ising (and seemingly andom) adap i e esponses
om he sys em.
Rega ding causal in e ence, sys ems ha a e me ely
based on p edic ion (i.e., co ela ional da a) will always be
highly dependen on con ex and ime. This may quickly
c ea e unplanned (and possibly inapp op ia e) cus ome
expe iences i he ci cums ances change. E e yone who
has shopped using ecommenda ion engines has aced
coun less si ua ions whe e he ecommenda ions seem o be
comple ely useless. This is o en based on algo i hms ha
a e simply co ela ional.
3 Economics
Clemens Puppe, Julia Na zige
Digi aliza ion has an impac on nea ly e e y aspec o
human in e ac ion. Ex ending he esea ch e i o y o he
in e sec ion o digi aliza ion and economics hus appea s o
be a na u al and wo hwhile endea o . And indeed, bo h
esea ch ields y o unde s and and design complex
in e ac ion ne wo ks by de eloping app op ia e mecha-
nisms and sys ems (Blume e al. 2015). Mo eo e , when
in eg a ing ou unde s anding o adap i e sys ems in o
economic esea ch, i becomes clea ha nume ous eco-
nomic p oblems a e ins ances o complex adap i e sys ems
hemsel es.
3.1 Adap i e Sys ems and In e ac i e Decisions
Complexi y a ises, among o he hings, om he ne wo k
o in e ac ing economic agen s. Thei agg ega e dynamic
beha io can be desc ibed by assessing indi idual beha -
io , s a egies, and decisions (Holland and Mille 2023).
Howe e , as hese au ho s also a gue, adap i e agen
beha io canno pe o m a an op imum le el du ing all
i e a ions o decision making. Indeed, e en i adap i e
sys ems a e designed o help agen s pe o m be e , non-
op imal s a es a e empo a ily una oidable due o he ime
needed o adap . Impo an economic examples a e he
in e ac ion be ween agen s in ne wo ks, o ins ance,
be ween i ms ha compe e on pla o ms. The e y idea o
he mechanics o adap a ion necessi a es an analysis o
sa is icing a he han op imizing sys ems (Simon 1957).
Imp o emen a he han op imali y is he goal ha adap-
i e sys ems need o se e. This poses a numbe o chal-
lenges o modeling as we ha e o ocus on non-linea
eedback, s a egic beha io , as well as indi idual and
123
I. Benke e al.: Hyb id Adap i e Sys ems, Bus In Sys Eng 66(2):233–247 (2024) 235
spa ial he e ogenei y in changing ime scales (Le in e al.
2013). The analyses also mus accoun o agen s’ abili y o
change and o lea n om expe ience. Examples a e he
expanding inno a ions o in o ma ion echnologies and
no el applica ions ha a e con inuously in eg a ed in o ou
daily li es and impac s a egies in decision making p o-
cesses. I equi es opening he pe spec i e on uly in e -
disciplina y esea ch o add ess hese challenges in
economics.
3.2 The Use o Adap i e Sys ems o Design Digi al
Nudges in Complex Si ua ions
Complexi y does no only cha ac e ize ne wo ks o in e -
ac ing economic agen s, bu also indi idual decisions: E en
when making a e y simple decision, an indi idual mus
ake many decision- ele an ac o s in o accoun and each
a decision wi hin a e y sho pe iod. Indi iduals who pay
limi ed a en ion o ace o he cogni i e limi a ions may
no conside all hese ac o s ( o an o e iew o models o
limi ed a en ion see Gabaix (2019)) and hence make
subop imal decisions. While adap i e sys ems migh no
achie e op imali y in such complex en i onmen s, hey can
help o imp o e decisions. Fo example, when indi iduals
pay limi ed a en ion, adap i e sys ems can be used o
design nudges, ha is, ools ha help o s ee indi idual
decisions in he ‘‘ igh ’’ di ec ion wi hou in e e ing wi h
he eedom o choice (Thale and Suns ein 2009).
Speci ically, in ecen yea s, digi al nudges ha e become
popula ( o an o e iew o di e en ypes o nudges see
Hummel and Maedche (2019) – encompassing, o exam-
ple, adap i e digi al ools such as app-based goal se ing
(e.g., Loeschel e al. 2020), o compu e games (Koch e al.
2023)). Such sys ems allow o send indi idualized eed-
back o eminde s and adap o expe ience. Challenges in
designing such nudges in complex en i onmen s a ise om
he p esence o beha io al o a en ional spillo e s on non-
a ge beha io s (c ., Al mann e al. 2022; Koch e al. 2023;
T ach man 2021) – a challenge ha adap i e sys ems migh
be be e sui ed o ackle han adi ional ools.
4 Psychology and Neu ocogni i e Sciences
Ul ich Ebne -P ieme , Man ed He mann, Benjamin
Scheibehenne
Psychology and neu ocogni i e science allow us o explo e
he use as he (in e )ac ing subjec in adap i e sys ems
h ough heo e ical concep s and empi ical p oo o why
and how human subjec s beha e in adap i e IT sys ems. On
a concep ual le el, his migh be achie ed by inse ing he
cogni i e a chi ec u e a he in e sec ion be ween humans
and machines. This means ha machine beha io migh
os e (ins ead o jus subs i u e) human decision making
p ocesses and ha human decision beha io will ansla e
o he design o adap i e sys ems. On a me hodological
le el, we a gue ha psychology will deli e he basic
building block we need o make he whole machine y
wo k. This co e s he ollowing aspec s: Fi s , p o iding
beha io al and biosignal da a o ob ain in o ma ion on
whe he a sys em is adap i e (o e en needs adap a ion).
Second, deli e ing he cogni i e and s a is ical models o
build a amewo k o expe imen al esea ch on adap i e
sys ems. Thi d, e alua ing he e ec and e ec sizes o
adap i e sys em in e en ions in indi idual and social
beha io using human ac i i y measu es. Gene ally, we
a gue ha success ul adap i e in o ma ion echnology
equi es a (be e ) unde s anding o human pe cep ion and
beha io (e.g., how much in o ma ion can people p ocess,
which ep esen a ion o ma os e s comp ehension, how
do si ua ional ac o s and emo ions in luence pe cep ion
and p e e ences). The ollowing h ee sub-sec ions p esen
h ee di e en pe spec i es on he ole o psychology and
neu ocogni i e sciences in adap i e sys ems esea ch.
4.1 Measu ing Psychological Phenomena o Adap i e
Sys ems: Validi y and E hical Conce ns
The ask o measu ing psychological phenomena can ange
om sligh o ex eme di icul y. This o igina es om he
b ead h o psychological phenomena, namely beha io ,
emo ions, and cogni ion. Heal h psychology is an example
o whe e adap i e sys ems can be applied easily, as heal h
beha io can be measu ed con inuously in daily li e.
Seden a iness, de ined as p olonged si ing, causes no only
physiological heal h issues bu also nega i ely a ec s
mood and he abili y o wo k (Giu giu e al. 2021).
Adap i e sys ems can use accele a i e senso s o measu e
body pos u e in e e yday li e, classi y p olonged si ing
pe iods in eal- ime, and igge ala ms o s udy he
unde lying mechanisms o o implemen change (Giu giu
e al. 2020). Such adap i e sys ems equi e he possibili y
o measu e and analyze seden a iness di ec ly, in his case
ia accele a i e senso s. In o he wo ds, we can measu e
and analyze he beha io i sel in a alid way in e e yday
li e. In psychology, hose me hods o en sha e he e m
Jus -In-Time Adap i e In e en ions (Nahum-Shani e al.
2018).
Un o una ely, o he psychological phenomena a e mo e
di icul o moni o in daily li e. An example om he
men al heal h con ex is he au oma ed eal- ime p edic ion
o upcoming illness episodes. Such an adap i e sys em
would ack psychological phenomena o in e es , analyze
hem in eal- ime, ac i a e ala ms in case o eme gency, as
well as adap h esholds o ala ms in case o alse o missed
123
236 I. Benke e al.: Hyb id Adap i e Sys ems, Bus In Sys Eng 66(2):233–247 (2024)
ala ms (Mu
¨hlbaue e al. 2018). In such a scena io, mobile
sensing o bipola diso de is a p ime candida e (Ebne -
P ieme and San angelo 2020), as mobile sensing pa am-
e e s a e closely ela ed o he psychopa hology o in e es
(e.g., al e ed communica i eness). Fo una ely, commu-
nica i eness, ope a ionalized ia mul iple pa ame e s,
including incoming and ou ing phone calls and ex , o
example, has e ealed a signi ican , empi ical ela ion o
illness episodes. Howe e , he alidi y o he eal- ime
measu ed p oxy o communica i eness emains unsa is-
ac o y. Human communica i eness encompasses much
mo e han jus communica ion wi h he sma phone,
al hough his p opo ion has emendously inc eased o e
he las decade. Ne e heless, di ec communica ion wi h
o he s, communica ion in g oups, communica ion ia
phones o (non-sma phone-based) ideo con e ences,
acial exp ession, ges u es as well as eye-con ac du ing
communica ion a e in eg al componen s o communica-
i eness. Sma phone communica ion, as a p oxy wi h
limi ed alidi y, ep esen s jus a po ion o human com-
munica i eness and does no measu e he cons uc i sel ,
in con as o ou example o seden a iness. O he psy-
chological phenomena, especially cogni ion and emo ions,
a e much mo e di icul o e en impossible o ack in daily
li e in eal- ime. Whe eas pe iphe al physiology has been
success ully used o decades, i s alidi y mus be ques-
ioned i always he same physiological measu e is used as
a p oxy o di e en cons uc s, such as hea a e ( a i-
abili y) o s ess, ange , elaxa ion, low, engagemen ,
a ec i i y, jus o name a ew.
E hical and p i acy conce ns can also hampe some
assessmen possibili ies which would be easible om a
echnological poin o iew. Con inuous audio eco dings
a e highly limi ed in se e al coun ies (e.g., Ge many),
esul ing om es ic ions speci ied, e.g., by Gene al Da a
P o ec ion Regula ion (GDPR) and he Telecommunica-
ions Ac s. Howe e , psychological phenomena, such as
mood, s ess, o mo i a ion, could be app oxima ed using
he au oma ic ex ac ion o ea u es om audio signals, a
me hod mainly applied in he a ec i e compu ing esea ch
communi y. Recen ad ances in AI migh enable eal- ime
analyses and ea u e ex ac ion on he mobile de ices
hemsel es, s o ing jus he ex ac ed pseudonymized
ea u es.
4.2 Decision-Making Pe spec i e
Human decision making can be unde s ood as an in e ac-
ion be ween indi idual cogni ion, such as pe cei ing and
p ocessing in o ma ion on he one hand, and he s uc u e
o he en i onmen on he o he hand (Simon 1990). This
in e ac ion implies ha beha io and hus e ealed p e -
e ences a e inhe en ly con ex dependen (e.g., T e sky and
Simonson 1993). I he decision con ex i sel changes as a
unc ion o indi idual beha io , as is he case o adap i e
sys ems, in e es ing eedback-loops occu . This may yield
malign consequences such as il e -bubbles, g oup pola -
iza ions, o he ding beha io , bu i may also os e sound
decision making, o example h ough decision suppo
sys ems ha p o ide in o ma ion in a anspa en way and
hus help people cu h ough he clu e (e.g., Ruo e al.
2022). Adap i e sys ems can also help decision make s o
o e come inhe en cogni i e capaci y limi s, e.g., wi h
espec o wo king memo y o in o ma ion p ocessing
(Olschewski e al. 2018). As an example, one may hink o
adap i e so ing and sc eening algo i hms ha na ow
down inc easingly la ge (online) asso men s and hus
allow consume s o quickly ind op ions (p oduc s, se -
ices, oman ic pa ne s, e c.) ha mee hei p e e ences
(e.g., Scheibehenne e al. 2010).
4.3 Going Beyond he Beha io al E alua ion: The
Impac o Human B ain Ac i i y Pa e ns
Concep s de i ed om beha io al economics, such as
nudging ood choices, migh p o ide he g ound o
adap i e sys ems whe eby hey can be able o shape a
subjec ’s beha io by combining bo h echnical and soci-
e al cueing. As sugges ed by He cbe g e al. (2022), one o
hese heal h nudges is a Nu i-Sco e, consis ing o a colo -
coded on -on-package nu i ion label, which was imple-
men ed as a olun a y se ice in Eu opean g oce y se ing
en i onmen s. I has been demons a ed ha his 5-colou
label e ec i ely guides consume s die a y decisions
owa ds heal hie choices a he poin o pu chase
(And ee a e al. 2021). I a colo label p omo es die a y
decisions owa ds heal hie ood i ems, i is in e es ing o
unde s and which pa o he decision making p ocess is
a ec ed by he label and how i con ibu es o he espec-
i e decision o adop a heal hie li es yle. Thus, i has o be
in es iga ed which a ibu es o a ood i em, such as he
as e o he p oduc , he gi en p ice, o i s subjec i e al-
ua ion o di e en heal h a ibu es a e a ec ing he indi-
idual beha io o make a choice o he espec i e ood
i em. Whe he o no adap i e sys ems will lead o sus-
ainable beha io al changes highly depends on how an
in e en ion is able o change he neu al a chi ec u e
unde lying he in ended beha io al change. By p o iding a
cogni i ely inspi ed and neu obiologically g ounded
amewo k, he neu ocogni i e pe spec i e migh signi i-
can ly add o he unde s anding o echnically based
adap i e sys ems. Wi h espec o he abo e in oduced
labeling o ood i ems o induce heal hie die a y decisions,
da a om neu ocogni i e expe imen s using unc ional
magne ic esonance imaging ( MRI) ha e demons a ed
ha colo -coded ( a ic ligh ) nu i ion labels signi ican ly
123
I. Benke e al.: Hyb id Adap i e Sys ems, Bus In Sys Eng 66(2):233–247 (2024) 237
al e he subjec s’ alua ion o heal hie ood i ems medi-
a ed by an inc eased WTP (Enax e al. 2015). The beha -
io al change ansla es in o a s onge ac i i y and
unc ional coupling in neu onal ne wo ks associa ed wi h
con ol o inhibi o y beha io (such as he en omedial
p e on al co ex) and sel -con ol in decision making (such
as he in e io and do sola e al pa s o he on al co ex).
A signi ican change in he unc ional connec i i y o b ain
ne wo ks is associa ed wi h he alua ion and ewa d
expec a ion o ood i ems and wi h he inhibi ion o
unheal hy choices signaled ia colo codes. This da a
demons a es how adap i e sys ems in e en ions migh be
e lec ed in he neu al a chi ec u e.
Al hough he add-on alue o neu ocogni i e app oa-
ches o he unde s anding o how adap i e sys ems shape
human beha io is s ill unde deba e and s ongly a ec ed
by di e ging heo e ical concep s, s a is ical models o
human decision making as well as expe imen al designs o
decision p ocesses, he e is inc easing e idence ha ana-
lyzing b ain beha io will signi ican ly add o making he
design o adap i e sys ems success ul. Unde s anding he
neu onal amewo k will no only help o unde s and how
adap i e sys ems wo k bu will also help o ge an idea o
how and why human beha io such as adop ing heal hie
ood choices sys ema ically escapes a ional choice models
ha a e de i ed om pu e echnically inspi ed scien i ic
app oaches (Mobbs e al. 2018).
5 Compu e Science
Michael Beigl, Felix Pu ze, Tanja Schul z
In compu e science, he e m ‘‘adap a ion’’ e e s o a
p ocess in which a echnical sys em adap s i s beha io o
indi idual use s based on h ee sou ces o in o ma ion: he
con ex o use, he en i onmen , and in o ma ion abou i s
use s (Feigh e al. 2012). The la e may include accumu-
la ed big da a o a ge ed use popula ions as well as cu -
en and endu ing cha ac e is ics o indi idual use s.
Compu e science is conce ned wi h he implemen a ion o
he p ocessing o inpu , he implemen a ion o he adjus -
men s a egy o change and adjus he inpu , and he
gene a ion o ou pu om he inpu and he adjus men
s a egy. While hese h ee componen s can be concep u-
alized as indi idual pa s o an adap i e sys em, hey sha e
a lo o me hods and challenges. Fo example, machine
lea ning is a key concep in all pa s o he in e p e a ion
o he he e ogeneous inpu signals, o he decision making
in he adjus men p ocess, and o he gene a ion o ou pu .
A join challenge is he handling o di e se pla o ms, om
mobile sys ems o mixed eali y.
The design o adap i e sys ems is in es iga ed in many
esea ch communi ies in compu e science, e.g., in he ield
o speech communica ion sys ems ha adap o speake s
(Legge e and Woodland 1995), domains (Sama akoon
e al. 2018), and languages (Schul z and Waibel 2001), in
human–compu e in e ac ion (HCI) (A alos-Vi e os e al.
2018; Benyon and Mu ay 1993; an de Zwaag e al.
2010), in obo ics and embodied agen s (Billa d e al.
2007), in so wa e enginee ing (Kepha and Chess 2003;
K ame and Magee 2007), in ubiqui ous compu ing (Cou-
inho e al. 2021), and in in e ac i e sys ems wi h appli-
ca ions ha ange om clinical suppo (Su on e al. 2020)
and heal h and nu sing ca e (Hu e e al. 2020) o e e yday
assis ance a home (G a e al. 2009) and e e yday ac i -
i ies a la ge (Rami ez-Ama o e al. 2017), jus o name a
ew. In he ollowing we b ie ly in oduce a cogni i e
sys ems and pe asi e compu ing pe spec i e.
5.1 Cogni i e Sys ems Pe spec i e
In he ield o cogni i e sys ems and AI, cogni ion-enabled
agen s a e coined o pe cei e hei con ex and en i onmen
h ough senso s and ac upon ha en i onmen h ough
ac ua o s (Russel and No ig 2003), and such agen s a e
cha ac e ized by he abili y o lea n abou , om, and wi h
humans. This abili y o lea n abou , om, and wi h humans
is c ucial o cogni i e sys ems in e ac ing wi h di e en
use s in di e en con ex s o en i onmen s and doing so
o e ex ended pe iods o ime. In such cases, he cogni i e
sys em mus deal wi h bo h in a-indi idual a ia ions in
use s a es and ai s, and in e -indi idual di e ences
be ween use s.
In he ollowing, we will highligh selec ed ends and
ad ancemen s in he ield o cogni i e sys ems. In pa ic-
ula , we ocus on Biosignal-adap i e Cogni i e Sys ems.
S a e-o - he-a sys ems as adap i e sys ems apply he
human-sys em-in e ac ion-loop concep , in which he
cogni i e sys em as de ined as an adap i e IT sys em
ecei es and in e p e s mul imodal biosignals o he human
use (Schul z e al. 2013). In such biosignal-adap i e
cogni i e sys ems, biosignals a e p ocessed and classi ied
in o use s a es (e.g., emo ions, s ess, wo kload, men al
asks) and ai s (e.g., iden i y, gende , pe sonali y) o adap
he in e ac ion s a egy such ha i bes suppo s he indi-
idual use s as adap i e social en i ies in hei aims and
needs. A he same ime, he cogni i e sys ems p o ide
anspa en and low-la ency eedback o he use s such ha
use s may in en ionally mode a e hei own biosignals o
ailo he sys em ou pu o hei needs. We e e o his
bila e al adap a ion p ocess as ‘‘co-adap a ion’’. We will
in oduce wo examples o biosignal-adap i e cogni i e
sys ems: (1) biosignal-enabled speech communica ion
sys ems ha p ocess speech- ela ed biosignals om b ain
123
238 I. Benke e al.: Hyb id Adap i e Sys ems, Bus In Sys Eng 66(2):233–247 (2024)
ac i i y wi h e y low la ency such ha use s can hea
hemsel es speaking e en when hey a e only imagining o
do so, and (2) biosignal-adap i e mixed eali y in e aces
o which adap a ion is c ucial o a oid use dis ac ion as
well as o cope wi h apid changing con ex and limi ed
bandwid h o explici in e ac ion.
5.1.1 Biosignal-Enabled Speech Communica ion
Speech is a complex p ocess ha s a s in he b ain and
ends wi h espi a o y, la yngeal, and a icula o y ges u es
o c ea ing acous ic signals o communica ion. Speech-
ela ed biosignals can be measu ed a he le el o he
ne ous sys em, muscula ac ion po en ials, speech kine-
ma ics, and sound p essu e. Biosignal-enabled speech
communica ion sys ems con e hese biosignals in o ex
o syn he ic oices by eplacing he acous ic signal p o-
cessing on end o au oma ic speech p ocessing wi h
me hods ailo ed o biosignals while lea ing he modeling
back-end o na u al language p ocessing unchanged
(Schul z e al. 2017). This app oach opens up new a enues
in speech communica ion sys ems. I makes speech p o-
cessing possible in he absence o an in elligible acous ic
signal, allowing o silen speech in e aces (Denby e al.
2010) ha gi e a oice o mu e people (e.g., la yngec-
omies), and i p ocesses speech when acous ic sound is no
desi able (e.g., o a oid dis u bance o p ese e p i acy in
public spaces), o subjec o noise (e.g., ad e se en i on-
men s, unde wa e , masks). Fu he mo e, since speech-
ela ed biosignals p ecede he acous ic ou pu by ens o
milliseconds, hey can p o ide ins an eedback o a human
in he sys em-in e ac ion-loop.
In ou wo k, we ake ad an age o hese bene i s by
cap u ing a icula o y muscle ac i i ies using su ace
Elec omyog aphy (EMG) and encoding he esul ing
elec ical biosignals in o ep esen a ions di ec ly decoded
in o audible speech signals. Since he EMG signal appea s
app oxima ely 60 ms p io o a icula o y mo emen s, he
la ency be ween sys em inpu and ou pu can be sho ened,
he eby imp o ing ace- o- ace communica ion quali y.
The almos ins an audi o y eedback inc eases a icula o y
awa eness, which is use ul o speech he apy and language
lea ning. Also, since EMG cap u es muscle ac i i y a he
han acous ics, EMG-based sys ems can handle silen
speech whe e a speake mo es he a icula o s as i p o-
ducing no mal modal speech bu supp esses he pulmona y
ai s eam so ha no sound is emi ed (Janke and Diene
2017).
5.1.2 Biosignal-Adap i e Mixed Reali y In e aces
Capabili ies o adap a ion become especially impo an in
sys ems ha exis on he mixed eali y con inuum
(Milg am e al. 1995), including augmen ed eali y (AR)
and VR. The easons o inc eased need o adap a ion a e
h ee old: Fi s , when ich, imme si e con en is gene a ed
he e is an ine i abili y o s imuli in mixed eali y ha can
o e whelm o dis ac use s and ha canno easily be
a oided by looking away om he sc een. Second, he e is
he dynamically changing con ex h ough he con inuous
gene a ion o new con en in he case o VR o he mobili y
in case o AR. Thi d, he e is he limi ed bandwid h o
explici use in e ac ion, as he ypical modali ies o use
inpu , ges u es, o speech canno con ey all ypes o
in o ma ion e y quickly and may equi e o ms o in e -
ac ion which a e no always a ailable.
Fo adap i e VR, esea che s ha e explo ed ways o
using adap a ion o ailo an expe ience o he use p o ile
(Bake and Fai clough 2022). A ypical aspec o he use
p o ile, besides cha ac e is ics such as gende o age, en ail
he use ’s men al wo kload le el, which can be moni o ed
h ough EEG (T emmel e al. 2019), NIRS (Pu ze e al.
2019), o o he physiological signals (Chiossi e al. 2022)
and which can be used o manipula e he isual complexi y
o he i ual en i onmen acco dingly (i.e., educing he
numbe o elemen s). O he ypes o adap a ion ely mo e
on he de ec ion o indi idual e en s, such as esponding o
di e en ypes o e o s in he VR scene ( acking e o s,
ende ing gli ches, e c.), which can be de ec ed om he
b ain ac i i y ollowing he occu ence o such e en s (Si-
Mohammed e al. 2020).
Fo adap i e AR, a s ong ocus has been on a en ion as
a a ge a iable, as his limi ed cogni i e esou ce espe-
cially comes unde s ess in AR se ings whe e he com-
plexi y o he eal wo ld is complemen ed by i ual s imuli
ha can be equally complex. Fo example, Vo mann and
Pu ze (2020) demons a ed he easibili y o EEG-based
a en ion adap a ion o a sma home con ol sys em ha
could imp o e usabili y signi ican ly. While he esea ch-
g ade de ices o such applica ions a e cu en ly s ill
unwieldy and no sui ed o ex ended use, u he esea ch
has al eady shown ha he same p inciple can be applied o
mobile sys ems: Vo mann e al. (2022) demons a ed how
consume -g ade EEG could be le e aged in a mobile AR-
based ansla ion applica ion o add a en ion adap a ion in
o de o u n o isual clu e when no a en ion was being
paid. Ano he domain in which eal-wo ld applicabili y is
in es iga ed a e AR sys ems in lea ning en i onmen s, o
example he use o a en ion-measu ing glasses (Kosmyna
e al. 2018).
5.2 Pe asi e Compu ing Pe spec i e
Today, AI plays a c i ical ole in adap i e sys ems by
enabling he de elopmen o in elligen algo i hms ha can
lea n and adap o changing ci cums ances. AI echniques
123
I. Benke e al.: Hyb id Adap i e Sys ems, Bus In Sys Eng 66(2):233–247 (2024) 239
such as machine lea ning, neu al ne wo ks, and e olu-
iona y algo i hms a e used o c ea e in elligen sys ems
ha can adap o di e en si ua ions, ul ima ely achie ing
hyb id in elligence (Aka a e al. 2020). A p ominen
example is Cha GPT, whe e he con ibu ion o esea ch on
adap i e sys ems lies in i s abili y o p o ide esea che s
wi h access o a as amoun o in o ma ion o gene a e
new insigh s. Al hough he sys em does no adap , he la ge
and complex con ex ea u e makes i appea o be adap i e
– hus, due o he dynamic and la ge con ex o GPT (Yang
e al. 2022) bo h sys em ou pu and human in e p e a ion
adap , making he sys em appea magically in elligen .
A mo e es ablished domain is con ol sys ems. Con ol
sys ems a e used in adap i e sys ems o manage inpu and
ou pu a iables and ensu e ha he sys em adap s co ec ly
o changing ci cums ances. These me hods can be com-
bined, o example, wi h so wa e enginee ing me hods o
c ea e sel -adap i e p og ams (B un e al. 2013)o o
de elop hem (K upi ze e al. 2015). HCI is conce ned
wi h he design and e alua ion o in e ac ion be ween a
human and a compu e , including he in e ace design, he
en i onmen (e gonomics), and he in e ac ion p ocess.
Wi h he ad en o ubiqui ous compu ing, he HCI com-
muni y is adding o i s esea ch agenda he concep o
dynamicy esponse o he use ’s con ex . This con ex ual-
iza ion in ol es adap ing use in e aces and in e ac ions
based on he use ’s con ex (Abowd e al. 1999). Wi h he
addi ion o senso s o compu e sys ems, he possibili y o
adap a ion goes beyond he di ec inpu he use gi es he
sys em and includes, o example, he en i onmen o
biological condi ions (Gelle sen e al. 2002). Beyond pe -
sonal compu e (PC)-based a i ac s (lap op, sma phones,
and wa ches, e c.), adap a ion is e en mo e necessa y in
obo ics. Robo ics esea che s use adap i e sys ems o
design obo s ha can ope a e in dynamic and unce ain
en i onmen s by using a mix u e o he abo e concep s.
Social obo s ep esen a signi ican ad ancemen in he
ield, allowing use s o in e ac wi h an h opomo phized
obo s (B eazeal 2003).
F om he abo e-men ioned scien i ic wo k, a se ies o
cha ac e is ics can be de i ed ha can be used o classi y
adap i e sys ems, such as:
Adap a ion Aspec Desc ip ion
Deg ee o sys em adap a ion
s. use adap a ion
The deg ee o which he sys em can
adap o he use ’s needs e sus he
use ’s need o adap hei beha io o
success ully ope a e he compu e
sys em
Use inpu and ou pu
a iables o adap a ion
Which and how many a iables as
pa ame e s, ou pu a iables, and an
adjus men mechanism de ine he
adap a ion om inpu o ou pu
Adap a ion Aspec Desc ip ion
Senso o sys em inpu
modali ies used o
adap a ion
Which inpu pa ame e s a e sensed
by senso s o sys em inpu modali ies
o enable he adap a ion
Adap a ion mechanism The adap a ion mechanism con ains
he adap a ion goals and a ans e
unc ion as an abs ac ins an ia ion
o he de ailed p ocedu al ans e
s eps om inpu o ou pu
Dynamic changes An adap i e sys em dynamically
changes he logic o he adap a ion
mechanism o is gi en one which
changes he ou pu a iables
acco ding o ex e nal o in e nal
con ex s
Con inuous s. s a ic
adap a ion
An adap i e sys em con inuously
execu es he adjus men mechanism
(i.e., he p ocess om inpu o ou pu
pa ame e s ia he ans e unc ion)
Flexibili y The e is a ange o en i onmen s and
si ua ions which he sys em can adap
o
Robus ness An adap i e sys em can con inue o
unc ion e en i some o i s
componen s ail o do no unc ion
op imally. I can also eco e om
unexpec ed changes in i s
en i onmen
Scalabili y An adap i e sys em can be scaled up
o down o handle la ge o smalle
asks wi hou losing i s adap i e
capabili ies
Abili y o lea n An adap i e sys em can lea n om i s
pas expe iences and imp o e i s
pe o mance o e ime
In eg a ion An adap i e sys em may in eg a e
di e en ypes o adap i e sys ems,
such as ule-based sys ems, neu al
ne wo ks, and e olu iona y
algo i hms, o achie e op imal
pe o mance, o i may be based on a
single ype o adap i e sys em
Responsi eness An adap i e sys em can espond
quickly o changes in i s en i onmen
o inpu s and adjus i s beha io
acco dingly
6 In o ma ion Sys ems
Ma c Adam, I o Benke, Ve ena Do ne , Michael Knie im,
Alexande Maedche, Jella P ei e , Ch is o Weinha d
F om an IS pe spec i e, adap i e sys ems a e an in e es ing
phenomenon as hey ep esen a con empo a y class o IS
ha le e ages s a e-o - he-a AI and senso echnologies o
pe o m sys em-d i en adap a ions and in pa allel has
123
240 I. Benke e al.: Hyb id Adap i e Sys ems, Bus In Sys Eng 66(2):233–247 (2024)
ela ed po en ials. In: IEEE con e ence on i ual eali y and 3D
use in e aces (VR). IEEE, A lan a, pp 653–661. h ps://doi.o g/
10.1109/VR46266.2020.00088
Simon H (1957) Models o Man; social and a ional. Wiley
Simon HA (1990) In a ian s o Human Beha io . Annu Re Psychol
41:1–19
S eil J, Hages ed I, Huang MX, Bulling A (2019) P i acy-awa e eye
acking using di e en ial p i acy. In: p oceedings o he 11 h
ACM symposium on eye acking esea ch and applica ions.
ACM, Den e . h ps://doi.o g/10.1145/3314111.3319915
S e n J, Schild C, A slan RC, Jones BC, DeB uine LM, Hahn A, Pu s
DA, Ze le I, Ko dsmeye TL, Feinbe g D, Penke L (2021) Do
oices ca y alid in o ma ion abou a speake ’s pe sonali y?
J Res Pe sonal. h ps://doi.o g/10.1016/j.j p.2021.104092
Su on RT, Pincock D, Baumga DC, Sadowski DC, Fedo ak RN,
K oeke KI (2020) An o e iew o clinical decision suppo
sys ems: bene i s, isks, and s a egies o success. Npj Digi Med
3:17. h ps://doi.o g/10.1038/s41746-020-0221-y
Thale RH, Suns ein CR (2009) Nudge: imp o ing decisions abou
heal h, weal h and happiness. Penguin
Tho nhill R (1997) The concep o an e ol ed adap a ion. In:
cha ac e izing human psychological adap a ions. Ciba ounda-
ion symposium 208:4–22
Tom o de S, P o hmann H, B anke J, Ha
¨hne J, Mni M, Mu
¨lle -
Schloe C, Rich e U, Schmeck H (2011) Obse a ion and
con ol o o ganic sys ems. In: Mu
¨lle -Schloe C, Schmeck H,
Unge e T (eds) O ganic compu ing — a pa adigm shi o
complex sys ems. Sp inge , Basel, pp 325–338. h ps://doi.o g/
10.1007/978-3-0348-0130-0_21
T ach man H (2021) Does p omo ing one beha io dis ac om
o he s? e idence om a ield expe imen . SSRN Elec on J.
h ps://doi.o g/10.2139/ss n.3941884
T emmel C, He C, Sa o T, Rechowicz K, Yamani Y, K usienski DJ
(2019) Es ima ing cogni i e wo kload in an in e ac i e i ual
eali y en i onmen using EEG. F on Hum Neu osci 13:401.
h ps://doi.o g/10.3389/ nhum.2019.00401
T e sky A, Simonson I (1993) Con ex -Dependen P e e ences
Manag Sci 39:1179–1189
Unbehauen H (2000) Adap i e Regelsys eme. Regelungs echnik III.
Sp inge , Heidelbe g, pp 133–261
Vo mann L-M, Weidenbach P, Pu ze F (2022) A AwAR ansla e:
a en ion-awa e language ansla ion applica ion in augmen ed
eali y o mobile phones. Senso s 22:6160. h ps://doi.o g/10.
3390/s22166160
Vo mann L-M, Pu ze F (2020) A en ion-awa e b ain compu e
in e ace o a oid dis ac ions in augmen ed eali y. In: ex ended
abs ac s o he 2020 CHI Con e ence on human ac o s in
compu ing sys ems. ACM, Honolulu h ps://doi.o g/10.1145/
3334480.3382889
Wand Y, Webe R (1995) On he deep s uc u e o in o ma ion
sys ems. In Sys J 5:203–223. h ps://doi.o g/10.1111/j.1365-
2575.1995. b00108.x
Weiß T, Me kl L, P ei e J (2023) Cus ome decision-making
p ocesses e isi ed: insigh s om an eye acking and ECG s udy
using a hidden Ma ko model. In: P oceedings o he Neu oIS
Weyns D (2019) So wa e enginee ing o sel -adap i e sys ems.
Handbook o so wa e enginee ing. Sp inge , Cham, pp 399–443
Wickens C, Hollands J, Banbu y S, Pa asu aman R (2015) Enginee -
ing psychology and human pe o mance. Rou ledge, New Yo k
Wiene M, Ma
¨h ing M, Remus U, Saunde s C (2016) Con ol
con igu a ion and con ol enac men in in o ma ion sys ems
p ojec s: e iew and expanded heo e ical amewo k. MIS Q
40:741–774. h ps://doi.o g/10.25300/MISQ/2016/40.3.11
Yang Z, Gan Z, Wang J, Hu X, Lu Y, Liu Z, Wang L (2022) An
empi ical s udy o GPT-3 o ew-sho knowledge-based VQA.
In: P oceedings o he AAAI Con e ence on a i icial in elli-
gence. 36:3081–3089. h ps://doi.o g/10.1609/aaai. 36i3.20215
an de Zwaag D, an den B oek D, Janssen JJ (2010) Guidelines o
biosignal d i en HCI. In: CHI, A lan a
123
I. Benke e al.: Hyb id Adap i e Sys ems, Bus In Sys Eng 66(2):233–247 (2024) 247
