Haup , Ma in; F eidank, Jan; Haas, Alexande
A icle — Published Ve sion
Consume esponses o human-AI collabo a ion a
o ganiza ional on lines: s a egies o escape algo i hm
a e sion in con en c ea ion
Re iew o Manage ial Science
P o ided in Coope a ion wi h:
Sp inge Na u e
Sugges ed Ci a ion: Haup , Ma in; F eidank, Jan; Haas, Alexande (2024) : Consume esponses o
human-AI collabo a ion a o ganiza ional on lines: s a egies o escape algo i hm a e sion in
con en c ea ion, Re iew o Manage ial Science, ISSN 1863-6691, Sp inge , Be lin, Heidelbe g, Vol.
19, Iss. 2, pp. 377-413,
h ps://doi.o g/10.1007/s11846-024-00748-y
This Ve sion is a ailable a :
h ps://hdl.handle.ne /10419/318868
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/
Vol.:(0123456789)
h ps://doi.o g/10.1007/s11846-024-00748-y
1 3
ORIGINAL PAPER
Consume esponses ohuman‑AI collabo a ion
a o ganiza ional on lines: s a egies oescape algo i hm
a e sion incon en c ea ion
Ma inHaup 1,2 · JanF eidank2· Alexande Haas1
Recei ed: 19 Janua y 2023 / Accep ed: 14 Feb ua y 2024
© The Au ho (s) 2024
Abs ac
Al hough A i icial In elligence can o e signi ican business bene i s, many con-
sume s ha e nega i e pe cep ions o AI, leading o nega i e eac ions when com-
panies ac e hically and disclose i s use. Based on he pe asi e example o con-
en c ea ion (e.g., ia ools like Cha GPT), his esea ch examines he po en ial o
human-AI collabo a ion o p ese e consume s’ message c edibili y judgmen s and
a i udes owa ds he company. The s udy compa es wo dis inc o ms o human-
AI collabo a ion, namely AI-suppo ed human au ho ship and human-con olled
AI au ho ship, wi h adi ional human au ho ship o ull au oma ion. Building on
he compensa o y con ol heo y and he algo i hm a e sion concep , he s udy
e alua es whe he disclosing a high human inpu sha e (wi hou explici con ol) o
human con ol o e AI (wi h lowe human inpu sha e) can mi iga e nega i e con-
sume eac ions. Mo eo e , his pape in es iga es he mode a ing ole o consum-
e s’ pe cei ed mo ali y o companies’ AI use. Resul s om wo expe imen s in di -
e en con ex s e eal ha human-AI collabo a ion can alle ia e nega i e consume
esponses, bu only when he collabo a ion indica es human con ol o e AI. Fu -
he mo e, he e ec s o con en au ho ship depend on consume s’ mo al accep ance
o a company’s AI use. AI au ho ship o ms wi hou human con ol lead o mo e
nega i e consume esponses in case o low pe cei ed mo ali y (and no e ec s in
case o high mo ali y), whe eas messages om AI wi h human con ol we e no
pe cei ed di e en ly o human au ho ship, i espec i e o he mo ali y le el. These
indings p o ide guidance o manage s on how o e ec i ely in eg a e human-AI
collabo a ion in o consume - acing applica ions and ad ises o ake consume s’ e hi-
cal conce ns in o accoun .
Keywo ds A i icial in elligence· Human-AI collabo a ion· AI augmen a ion· AI
e hics· Con en c ea ion· Algo i hm a e sion
JEL classi ica ion C91· L86· M31· O33
Ex ended au ho in o ma ion a ailable on he las page o he a icle
Re iew o Manage ial Science (2025) 19:377–413
/Publishedonline:4 Ap il2024
M.Haup e al.
1 3
1 In oduc ion
A i icial In elligence (AI) cu en ly eshapes business and ma ke ing s a egies
as companies inc easingly ely on he use o AI sys ems (Kanbach e al. 2023).
Pa icula ly AI-powe ed ools such as Cha GPT ha e seen emendous in e es
as hey a e inc easingly able o c ea e compelling con en ha can ba ely be dis-
inguished om human-au ho ed ex s (Köbis and Mossink 2021; Waddell 2018),
and schola s iden i ied con en gene a ion as a key applica ion a ea o AI in ma -
ke ing, legal, inance and o he business ields (Dwi edi e al. 2023; G ae e and
Bohlken 2020; Kahn 2019). Al hough mo e and mo e companies use AI, con-
sume s ha e a nega i e pe cep ion o AI and indica ed a he an unwillingness o
us in AI. A ecen su ey om Sales o ce among 11,000 consume s e ealed
ha nea ly h ee qua e s o consume s (74%) a e conce ned abou he une hical
use o AI and only hal o hem a e open o use AI o imp o e hei expe iences
(Sales o ce 2023). Schola s acknowledged his phenomenon in a ious s ud-
ies and e med his nega i e pe cep ion o AI as algo i hm a e sion, which was
obse ed e en when algo i hms we e objec i ely ou pe o ming humans (Bu on
e al. 2020; Cas elo e al. 2019; Die o s e al. 2015, 2016; Yeomans e al. 2019).
Reasons o his a e ha indi iduals a e igh ened ha AI will a ain oo much
powe and ge beyond human con ol (Al onseca e al. 2021; Bu on e al. 2020;
Siau and Wang 2020). These nega i e pe cep ions a e c i ical o companies ha
use AI a o ganiza ional on lines (e.g., con en c ea ion o company websi es),
as anspa en AI decla a ion will become a legal obliga ion in many coun ies in
he nea u u e (e.g., “EU AI Ac ”; Eu opean Pa liamen 2023). Thus, companies
using AI- ools in consume - acing applica ions a e inc easingly con on ed wi h
he ques ion o how o in eg a e AI anspa en ly wi hou su e ing om nega i e
consume esponses.
The e is a clea need o u he esea ch on his ques ion. Resea ch has s a ed
o in es iga e how o le e age he e iciency o AI while a oiding nega i e con-
sume esponses esul ing om i s use (Huang and Rus 2022; Zanzo o 2019).
As human-AI collabo a ion seems pa icula ly ui ul o his end, i has ecei ed
inc easing schola ly a en ion la ely (Hassani e al. 2020; Lange and Lande s
2021; Ra opoulos e al. 2023; Zhou e al. 2021). Fo example, his s eam o
esea ch ound ha collabo a i e wo k be ween humans and AI inc eased us in
AI sys ems and manage s’ pe cep ions o empowe men (i.e., he abili y o adap
o change) (Schlei h e al. 2022). Despi e he g owing body o esea ch ega ding
human-AI collabo a ion in a ious ields, empi ical s udies ega ding he use and
decla a ion o human-AI collabo a ion a o ganiza ional on lines emain sca ce,
pa icula ly in he ields o managemen and ma ke ing. This lack o empi ical
s udies is su p ising gi en he high po en ial o c ea e e iciencies a he o gani-
za ional on line and he high pe o mance le el o mode n ex -gene a ing ools
such as Cha GPT (Dwi edi e al. 2023) in combina ion wi h he challenges due
o legisla i e equi emen s o AI anspa ency (Eu opean Pa liamen 2023). In
addi ion, he li le esea ch om o he ields (Waddell 2019; Wölke and Pow-
ell 2018) p o ides con lic ing e idence on consume esponses o human-AI
378
1 3
Consume esponses ohuman‑AI collabo a ion a o ganiza ional…
collabo a ion. O e all, he ques ion o how o anspa en ly in eg a e AI a o gan-
iza ional on lines wi hou su e ing om nega i e consume esponses emains
open.
Ou s udy add esses his ques ion and in es iga es consume esponses o human-
AI collabo a ion a o ganiza ional on lines. Speci ically, wi h a ocus on con en
c ea ion o company homepages, we use he concep o algo i hm a e sion (Bu -
on e al. 2020) and compensa o y con ol heo y (Landau e al. 2015) o de elop a
concep ual model in which wo o ms o human-AI collabo a ion (i.e., AI-suppo ed
human au ho ship and human-con olled AI au ho ship; Baile e al. 2022) ela e
o consume s’ a i ude owa ds he company media ed by message c edibili y. As
he ising use o AI in managemen and ma ke ing has spa ked discussions on co -
po a e esponsibili y and consume s’ pe cep ions o mo ali y o companies’ AI use
(C eme and Kaspa o 2021; Hagendo 2020; Siau and Wang 2020; Wi z e al.
2022), he concep ual model also includes possible mode a o e ec s o consume -
pe cei ed mo ali y o companies’ AI use. We es ou concep ual model wi h da a
om wo expe imen al s udies execu ed wi h ic i ious scena ios and company p o-
iles on he pla o ms mTu k and P oli ic.
Theo e ically, his esea ch con ibu es o a be e unde s anding o human-AI
collabo a ion e ec s o consume - acing applica ions. Ou esul s e eal ha AI use
a o ganiza ional on lines does no gene a e nega i e consume esponses ( ela i e
o human au ho ship) when con en c ea ion by AI is con olled by humans o pe -
cei ed mo ali y is high. Wi h ou indings, we add o he deba e whe he AI should
augmen o eplace humans in managemen (Hassani e al. 2020; Huang and Rus
2022) and o e insigh s o he new ield o AI e hics and i s links o ma ke ing s a -
egy (Siau and Wang 2020). Fo manage s, his esea ch o e s a solu ion o escape
he dilemma be ween e hical (and upcoming legal) misconduc by hiding AI use and
nega i e consume eac ions o anspa en AI use. The indings p o ide hem wi h a
deepe unde s anding o consume esponses o company’s AI use, as well as ac ion-
able guidance o he highly ele an ques ion o how o manage human-AI collabo-
a ion. By doing his, ou esea ch also add esses calls ega ding he op imal design
o human-AI join wo k o ces (Huang and Rus 2022; Zhou e al. 2021).
2 Theo e ical backg ound andhypo heses de elopmen
2.1 Pe o mance andpe cep ions o AI
Rela ed o consume esea ch, AI can be de ined as “any machine ha uses any kind
o algo i hm o s a is ical model o pe o m pe cep ual, cogni i e, and con e sa-
ional unc ions ypical o he human mind” (Longoni e al. 2019, p.630). Since i s
incep ion in he 1950s, AI has unde gone ema kable de elopmen . Ea ly yea s saw
symbolic AI app oaches, ocusing on ule-based sys ems and expe sys ems. In he
ecen yea s, key echnologies such as machine lea ning o neu al ne wo ks e olu-
ionized AI applica ions in a eas like na u al language p ocessing o image p ocess-
ing (Da enpo e al. 2020; Hassani e al. 2020). Pa allel o o he disciplines, he
p ecision and e ec i eness o AI in con en c ea ion is apidly de eloping (Dwi edi
379
M.Haup e al.
1 3
e al. 2023). Quali y and p ecision o AI a e ising d as ically and AI con en is
o en no dis inguishable om human-w i en con en (Köbis and Mossink 2021).
A ecen me a-s udy om G ae e and Bohlken (2020) showed ha AI-based ex s
achie ed compa able e alua ions o human-w i en ex s in a ious s udies—as long
as con en au ho ship was hidden. Howe e , when he use o AI is anspa en , con-
sume s we e ound o eac di e en ly.
2.2 T anspa en AI igge s algo i hm a e sion
When companies anspa en ly decla e hei use o AI, schola s widely obse ed he
phenomenon o algo i hm a e sion, ha is consume s’ eluc ance o use AI (com-
pa ed o humans) (Cas elo e al. 2019; Die o s e al. 2015). This phenomenon was
ound in a ious ins ances, including p oduc - and se ice- ecommenda ions (Lon-
goni and Cian 2022; Wien and Peluso 2021), pe o mance- ela ed o ecas s (Die -
o s e al. 2015), and inancial ad ice (Önkal e al. 2009). A sys ema ic li e a u e
e iew o Bu on e al. (2020) e ealed ha algo i hm a e sion has been consis -
en ly documen ed since he 1950’s and can be a ibu ed o se e al causes: Schola s
asse ed ha humans a ed AI gene ally as less us wo hy, less empa he ic, and less
compe en (Chan-Olms ed 2019; Luo e al. 2019).
Mo eo e , an AI-d i en digi al agen (i.e., cha bo ) was equally e ec i e as
a compe en human sales agen in e ms o con e sion a es—bu only as long as
he cha bo ’s iden i y was hidden. By disclosing he AI iden i y, he pu chase a e
d opped by o e 75% because consume s pe cei ed he cha bo as less knowledge-
able and less empa he ic han a human salespe son (Luo e al. 2019). Simila ly,
Cas elo e al. (2019) showed ha consume s assume ha AI is incapable o suc-
cess ully comple e subjec i e asks, leading o lowe us and eliance on AI. How-
e e , a ecen s udy o Longoni and Cian (2022) showed ha p oduc and se ice
a ibu es (i.e., hedonic o u ili a ian con ex s) de e mine whe he people p e e AI
o human ad ice, and hus ac as a bounda y condi ion o he algo i hm a e sion
e ec . Second, humans seemed o expec mo e pe ec esul s om an AI han om
a human, and seeing AI making a mis ake led o lowe con idence owa ds he AI
and an AI ejec ion o u he asks (Die o s e al. 2015). Thi d, many p ocesses
o AI, such as machine lea ning, a e ha d o explain—e en o hei c ea o s, and
hus a e o en conside ed as inhe en ly in anspa en o as “black box” (Siau and
Wang 2020). This de ici o unde s anding AI c ea es in o ma ion asymme ies and
uels ea s and dis us (Pun oni e al. 2021).
In line wi h hese indings, algo i hm a e sion has also been ound ela ed o AI
con en c ea ion. Indi iduals o en assigned highe a ings ega ding c edibili y,
eadabili y, o quali y o human—( s. AI-) gene a ed con en when au ho ship was
anspa en (G ae e and Bohlken 2020; Waddell 2018). G ae e and Bohlken (2020)
showed ha hese a ings we e e en made ega dless o he ac ual sou ce. Tha
means, despi e an iden ical ex , he assignmen o an AI ( s. human) au ho ship
sys ema ically leads o mo e nega i e a ings.
Schola s consen ha algo i hm a e sion seems o be mainly d i en by a low sub-
jec i e sou ce c edibili y o AI a he han a lack o objec i e AI quali y (G ae e and
380
1 3
Consume esponses ohuman‑AI collabo a ion a o ganiza ional…
Bohlken 2020; Luo e al. 2019). Essen ially, sou ce c edibili y could be de ined as
“quali ies o an in o ma ion sou ce which cause wha i says o be belie able” (Wes
1994, p.159). Acco ding o he sou ce c edibili y heo y (Ho land e al. 1953), indi-
iduals a e mo e likely o be pe suaded when he sou ce is e alua ed as c edible
(i.e., expe ul and us wo hy). Mani old s udies h oughou he las decades sup-
po his p oposi ion ( o an o e iew, see Ismagilo a e al. (2020)). Mo e ( s. less)
c edible sou ces we e ound o c ea e a o able ou comes including enhanced mes-
sage e alua ions, a i udes, and beha io al in en ions. Fo ins ance, high sou ce c ed-
ibili y leads o highe b and us o pu chase in en ions (Ha mon and Coney 1982;
Luo e al. 2019; Ohanian 1990; Visen in e al. 2019). Mo eo e , sou ce c edibil-
i y signi ican ly inc eases message c edibili y pe cep ions (Ismagilo a e al. 2020;
Visen in e al. 2019) and hus, e en he same con en could be pe cei ed di e en ly
due o di e en sou ces. Essen ially, message c edibili y e e s o “an indi idual’s
judgmen o he e aci y o he con en o communica ion” (Appelman and Sunda
2016, p.63).
While mos s udies p o ided e idence o he phenomenon o algo i hm a e -
sion (see G ae e and Bohlken 2020), some s udies ound no e ec o e en a posi i e
e ec o AI au ho ship on pe cei ed con en c edibili y, e.g., in spo s news (Wölke
and Powell 2018). Thus, al hough algo i hm a e sion domina es human pe cep ions
o AI, he e ec was no ully consis en h oughou con en opics o AI asks.
2.3 T anspa en AI igge s pe cei ed loss o con ol
In addi ion o ha , many people ea ha AI could ake o e con ol in se e al
domains o h ea en human jobs (Huang and Rus 2022). These eelings a e no
unjus i ied. When AI akes o e a ask, i o en eplaces human in elligence and
ine i ably akes away human con ol and jobs as a long- e m consequence (e.g.,
au onomous ca s eplace axi d i e s (F ey and Osbo ne 2017; Huang and Rus
2022; Osbu g e al. 2022) o AI agen s eplace jou nalis s (Ye ushalmy 2023)).
Schola s consen ha al eady “ he me e ecogni ion o AI’s capabili y o ac as a
subs i u e o human labo can be psychologically h ea ening” (Pun oni e al. 2021,
p.140).
The desi e o con ol is an essen ial human need and e e s o people’s desi e
o be able o manage p ocesses and ou comes o e en s in li e (Bu on e al. 2020;
Chen e al. 2017; Pun oni e al. 2021). He ein, con ol e e s o he abili y o in lu-
ence ou comes in one’s en i onmen (Skinne 1996). When his need o con ol is
h ea ened o emains unme , people expe ience nega i e a ec , including discom-
o , us a ion, demo i a ion, and helplessness, and espond wi h nega i e beha io
such as mo al ou age o eac ance (Chen e al. 2017; Landau e al. 2015; Pun oni
e al. 2021). Fu he mo e, acco ding o he compensa o y con ol heo y (Landau
e al. 2015), indi iduals who expe ience a educed le el o con ol espond wi h
compensa o y s a egies o es o e hei pe cei ed con ol. As adi ional s a -
egy, people bols e hei pe sonal agency, which is hei belie ha hey possess
he esou ces needed o pe o m a speci ic ac ion (Lange 1975). Acco ding o a
ecen li e a u e e iew o Cu igh and Wu (2023), pe cep ions o low con ol shape
381
M.Haup e al.
1 3
consume s’ beha io ei he by mo i a ing hem o look o a sense o con ol and
o de in hei consump ion en i onmen ; o by mo i a ing hem o use consump ion
as a unc ion o egain con ol. A g owing body o li e a u e examines ha p oduc
acquisi ion could sa is y consume s’ need o con ol (Billo e and Anisimo a 2021;
Chen e al. 2017; Cu igh and Wu 2023).
When i comes o con en ma ke ing, consume s migh ea ha AI becomes so
sophis ica ed ha hey will no be able o dis inguish an AI om a human au ho
(which would be suppo ed by esea ch esul s like he s udy om Köbis and Mos-
sink 2021), esul ing in a lack o con ol o e he message p o ide . This, in u n,
c ea es he ea ha companies migh use AI and manipula e consume s’ ac i i ies
and pe cep ions (Jobin e al. 2019). Fo ins ance, consume s eel unce ain whe he
a con en is genuine human o no —and who con ols i (G ae e and Bohlken 2020).
People do no only ely on hemsel es bu also on o he humans o es o e con ol.
The e o e, a u he s a egy men ioned in he compensa o y con ol heo y is he
so-called seconda y con ol, which is a pe son’s belie o ha e access o an ex e nal
agen who possesses a desi ed o needed abili y (Landau e al. 2015). Tha means,
a pe son o ins i u ion ou side o one’s sel can in luence pe sonally impo an ou -
comes and inc ease he chances o achie e one’s goals (F iesen e al. 2014; Kay
e al. 2008; Landau e al. 2015). Schola s showed ha when people eel a lack o
con ol, hey ely s onge on o he en i ies which p o ide clea ules and s uc u es
and hus sa is y hei desi e o o de and con ol (F iesen e al. 2014; Kay e al.
2008). Fo ins ance, indi iduals we e mo e suppo i e o hie a chies in he wo k-
place (and a o ed hie a chy-enhancing jobs) when hei sense o con ol was h ea -
ened (F iesen e al. 2014). Simila ly, a s udy wi h people om 67 na ions showed
ha lowe pe cei ed con ol is s ongly co ela ed wi h highe suppo o go e n-
men al con ol (Kay e al. 2008). We adop ed his concep o seconda y con ol o
ou esea ch design as con ol is exe ed by he ex e nal agen — he human au ho .
2.4 Human‑AI collabo a ion aspossible escape onega i e consume esponses
oAI
One possible, bu unde - esea ched, solu ion o mi iga e he nega i e consequences
o AI use a o ganiza ional on lines lies in he collabo a ion o humans and AI,
meaning ha “AI sys ems wo k join ly wi h humans like eamma es o pa ne s o
sol e p oblems” (Lai e al. 2021, p.390). Fo a ious managemen and ma ke ing
asks, schola s consen ha AI and humans could collabo a e in mani old ways o
use he espec i e s eng hs o humans and AI (Huang and Rus 2022; Ra opoulos
e al. 2023; Zhou e al. 2021). Fo ins ance, human-AI collabo a ion can suppo
heal hca e p o essionals (Lai e al. 2021), gene al managemen (Sowa e al. 2021),
o da a scien is s (Wang e al. 2022). Humans could collabo a e wi h AI in ad e -
ising (Vak a sas and Wang 2021), ma ke ing planning and s a egy (Ameen e al.
2022), o join ly deli e cus ome se ice (Wi z e al. 2018). AI could also aug-
men salespe sons’ capabili ies in e e y s age o he sales p ocess (Da enpo e al.
2020; Paschen e al. 2020). Fo example, AI could de ec unmen ioned complain s
wi h he help o au oma ed cus ome ’s oice analysis and a human salespe son could
382
1 3
Consume esponses ohuman‑AI collabo a ion a o ganiza ional…
ollow up on his (Da enpo e al. 2020); AI could p edic leads and pe sonalize
con en , whe eas he human could e i y leads and link hem o he business con-
ex (Paschen e al. 2020); and AI could suppo use expe ience e alua ions (e.g.,
by iden i ying issues in usabili y es ideos) o enhance use engagemen and sales
(Fan e al. 2022). Table1 o e s an o e iew o ele an concep ual and empi ical
esea ch ega ding human-AI collabo a ion.
Human-AI collabo a ion can be designed in di e en ways on he con inuum
be ween he end-poin s o a sole human ac o and sole AI. Following Huang and
Rus ’s (2022) amewo k o collabo a i e AI, human-AI collabo a ion egula ly ol-
lows a s epwise pa e n: Due o he pe manen de elopmen o AI, AI s a s as aug-
men a ion and suppo o humans, and la e could eplace humans and ul ill he
ask au onomously. Howe e , in be ween suppo and eplacemen , se e al schola s
acknowledge ha AI migh pe o m he ask unde he su eillance and con ol o a
human (Longoni e al. 2019; Nyholm 2022; Osbu g e al. 2022). Rela ed o con en
c ea ion, Baile e al. (2022) dis inguish be ween AI-suppo ed human au ho ship
(i.e., labeled as “AI in he loop o human in elligence “) e sus AI ask ake-o e
wi h human con ol (i.e., labeled as “human in he loop o AI”).
Al hough schola s ha e acknowledged hese di e en collabo a ion o ma s,
esea ch cu en ly lacks empi ical e idence ega ding he impac o hese di e en
o ms o collabo a ion be ween humans and AI on consume esponses. In gene al,
schola s sugges ha AI is mo e e ec i e when i augmen s ( s. eplaces) human
ma ke ing manage s (Da enpo e al. 2020), as he coope a ion will lead o highe
alue and compe i i e ad an age compa ed o human eplacemen , e.g., in educa-
ion, medicine, business, science and o he s (Paschen e al. 2020; Zhou e al. 2021).
Se e al s udies om di e se ields showed ha in eg a ing humans in o AI asks is
educing hei ini ial algo i hm a e sion (Bu on e al. 2020; Die o s e al. 2016;
Tobia e al. 2021). Mo eo e , empi ical e idence showed ha engaging in collabo a-
i e asks wi h an AI-d i en obo inc eased consume s’ appo , coope a ion, and
engagemen le els (Seo e al. 2018).
An analysis o human-AI collabo a ion e ec s in con en ma ke ing is missing.
Howe e , schola s in he ela ed ield o jou nalism and news p oduc ion ha e s a ed
o e alua e his and labeled i “hyb id” o “ andem” au ho ship. Se e al au ho s d aw
op imis ic scena ios whe e AI could be in eg a ed in o jou nalis ic wo k, and AI and
jou nalis s could each a s a e o coope a ion ins ead o cannibaliza ion (G ae e and
Bohlken 2020; G ae e e al. 2016; Wölke and Powell 2018). Suppo ing ha , Wad-
dell (2019) asse s p agma ically ha many cu en AI sys ems in jou nalism s ill
need some human inpu anyhow, he e o e men ioning bo h human and AI as coop-
e a i e au ho s is ecommended. Empi ically, Wölke and Powell (2018) show ha
a human-AI collabo a ion o la gely s anda dized spo s and inance epo s is pe -
cei ed as an equally c edible sou ce as a human au ho , and he collabo a ion did no
lead o lowe news selec ion. These schola s assume ha his migh be oo ed ei he
in he pe cep ion o an AI as a mo e objec i e au ho o in ini ially low expec a-
ions owa d AI au ho ship. In sum, empi ical e idence gene ally suppo s posi i e
impac s o human-AI collabo a ion, bu speci ic insigh s abou he e ec s o di e -
en collabo a ion o ms a e missing. The e o e, he ques ion o how algo i hm a e -
sion can bes be escaped when decla ing AI use emains unanswe ed.
383
M.Haup e al.
1 3
Table 1 Selec ed li e a u e ela ed o Human-AI collabo a ion
Resea ch s eams Sou ce Topic Li e a u e ield(s) Media o (s), Mode a o (s) Main indings
Concep ual Da enpo e al. (2020) Mul idimensional ame-
wo k which in eg a es AI
in elligence le els, ask
ypes and appea ance
Ma ke ing – AI will in luence ma ke ing
s a egies and consume
beha io in mani old
ways. Au ho s sugges ha
human-AI collabo a ion is
mo e e ec i e han human
eplacemen , and e hical
issues need o be consid-
e ed cau iously
Huang and Rus (2022) Concep ual amewo k
o collabo a i e AI in
ma ke ing
Ma ke ing – AI ad ances om mechani-
cal, o hinking, o eeling
in elligence. Human-AI
collabo a ion can be
achie ed h ough (1) using
he espec i e s eng hs
o human o AI, (2) using
lowe -le el AI o aug-
men highe -le el human
in elligence, o (3) using
AI o au oma e lowe
in elligence p ocesses and
humans ocus on highe
in elligence asks. Possible
bounda y condi ions such
as ask (un-)desi abili y a e
discussed
384
1 3
Consume esponses ohuman‑AI collabo a ion a o ganiza ional…
2.4.1 The ela ionship be weendi e en o ms o human‑AI collabo a ion, message
c edibili y, anda i ude owa ds hecompany
Schola s acknowledge ha human-AI collabo a ion can be e alua ed based on di -
e en schemes. As cues o he message c edibili y e alua ion, people could ei he
ocus on whe he human o AI p o ided he majo pa o inpu , o he le el o pe -
cei ed human au ho i y and con ol o e AI (Bu on e al. 2020; Die o s e al.
2016). A adi ional c i e ion o e alua e con en o wo au ho s is o base he deci-
sion on he pa icula wo kload o inpu each au ho p o ided. Fo ins ance, in aca-
demic con en wi h coope a i e au ho ship, he au ho ship o de e lec s he le el o
con ibu ion and inpu sha e (Newman and Jones 2006). Gi en he endency o peo-
ple’s algo i hm a e sion (Longoni e al. 2019; Luo e al. 2019), highe inpu sha e o
a human (AI) au ho is expec ed o be pe cei ed as mo e posi i e (nega i e). Thus, a
highe le el o AI inpu sha e is expec ed o educe message c edibili y e alua ions
because people gene ally a e AI as a less c edible sou ce (Luo e al. 2019).
Nex o his, people could also e alua e a human-AI collabo a ion based on he
pe cei ed le el o human au ho i y and con ol o e AI in he con en c ea ion p o-
cess. As men ioned abo e (see 2.2), he use o AI as an au onomous sys em dep i es
people’s sense o con ol o e p ocesses and ou comes (Huang and Rus 2022;
Osbu g e al. 2022). To coun e ac his, humans ac as supe iso s in many p ocesses
whe e AI is used. Fo ins ance, humans supe ise AI’s (semi-)au onomous s ee -
ing o a ca , o a human doc o con ols AI’s medical ad ice (Longoni e al. 2019;
Osbu g e al. 2022).
Rela ed o AI au ho ship in con en c ea ion, i is p ac ically impossible o he
eade s o in luence who w i es he ex o o e i y he con en ’s u h ulness (i.e.,
objec i i y and hones y) (Waddell 2019). Ins ead, he eade has o ely on second-
a y con ol whene e possible— o ins ance o us a human co-au ho o edi o and
o hand o e he con ol o e i ica ion o he con en o hem.
In gene al, he desi e o ha e o es o e con ol o e one’s en i onmen was ound
o be an inna e human need and a qui e s ong mo i a o . Fo ins ance, when peo-
ple’s eeling o con ol is impai ed, hey eac wi h s ongly nega i e a ec including
ange , mo al ou age, o eac ance (Pun oni e al. 2021). Longoni e al. (2019) ind
ha people’s esis ance o use medical AI could be alle ia ed when AI suppo ed a
human who makes he inal decision (i.e., is in con ol) ins ead o a sole AI se ice
p o ision. These esul s suppo he e ec i eness o he o m “human-con olled AI
au ho ship”.
In con as , a high human sha e o inpu (as indica ed in he o m “AI-suppo ed
human au ho ship”) is expec ed o be a less clea and powe ul cue o he e alua-
ion o message c edibili y. Pa icula ly when he human inpu is no clea ly isible
and dis inguishable om AI inpu (e.g., as mainly gi en in human-AI collabo a-
i e cases), people pe cei e a highe le el o machine agency compa ed o human
agency, and hus a lack o au ho i y (Sunda 2020). Mo eo e , wi hou human con-
ol, indi iduals migh pe cei e an inc eased isk o inco ec in o ma ion (o ac ion)
om AI’s inpu as no hie a chies and con ol unc ions a e sough o be in place
391
M.Haup e al.
1 3
(Osbu g e al. 2022). Thus, human con ol o e AI is hough o ha e a s onge posi-
i e in luence on message c edibili y pe cep ions han human inpu sha e. In pa icu-
la , when human con ol is no speci ied, a high le el o human inpu is no expec ed
o educe he nega i e impac o AI au ho ship ( s. a sole human au ho ed mes-
sage). Howe e , when human con ol is s a ed, he pe cep ion o seconda y con ol
can mi iga e he nega i e impac o AI au ho ship e en wi h less human inpu . We
hypo hesize:
H1a AI-suppo ed human au ho ship ( s. human au ho ship) leads o lowe mes-
sage c edibili y e alua ions.
H1b Human-con olled AI au ho ship ( s. human au ho ship) does no lead o di -
e en message c edibili y e alua ions.
Following pe suasion esea ch, message c edibili y a ec s how people make
subsequen judgmen s abou he message-sending ins i u ion, such as compa-
nies o news agencies (Ho land e al. 1953). In pa icula , c edible messages we e
ound o inc ease consume s’ us and a i udes owa ds he message sende , and
a o able beha io al in en ions, including in o ma ion adop ion o pu chase in en-
ions (Ismagilo a e al. 2020; Wölke and Powell 2018). The eby, a posi i e a i ude
owa ds he company e e s o a eade s’ posi i e imp ession o he company, i s
epu a ion, o image (Da ke e al. 2008). We posi :
H2 S onge pe cep ions o message c edibili y lead o mo e posi i e a i udes
owa ds he company.
2.5 The mode a ing ole o mo ali y o AI use
Due o he inc easing popula i y o AI echnologies, AI has gained a subs an ial
impac on humans and socie y (Hagendo 2020). Despi e undoub ed imp o e-
men s o se ice quali y and cus ome expe ience, AI echnologies also pose
mo al h ea s, such as issues o ai ness, e hical misconduc , o consume p i-
acy (Pun oni e al. 2021). As a esponse, he new ield o AI e hics as pa o
applied e hics gains ele ance and momen um (Hagendo 2020; Siau and Wang
2020). As o e a ching goals, AI e hics should p omo e bene i s o humans,
os e mo al beha io o enhance social good (“bene icence”), and p e en any
ha m ul consequences (“non-male icence”) (He mann 2022; Jobin e al. 2019).
As many consume s we e ound o ha e mo al conce ns and ese a ions owa d
AI, discussions abou he mo ali y o companies’ AI use a e ongoing in di e -
en domains and conside mul iple ace s (Siau and Wang 2020). Popula mo al
conce ns a e he lack o AI con ol, non- anspa en AI p ocesses (“black box”),
disc imina ion, o low eliabili y o AI-c ea ed in o ma ion (Jobin e al. 2019;
Pun oni e al. 2021; Rai 2020). Fu he mo e, schola s acknowledged possible
mo ali y issues when AI is in eg a ed in consume - acing applica ions because
i could educe consume au onomy (Libai e al. 2020) and migh be a highly
392
1 3
Consume esponses ohuman‑AI collabo a ion a o ganiza ional…
manipula i e sys em ha could cause o suppo addic i e use beha io (Daza
and Ilozumba 2022; He mann 2022). Fo example, AI could os e exhaus i e
social media usage h ough hype -pe sonaliza ion and op imiza ion o p e e ed
con en and ads, which inc eases ma ke ing e ec i eness bu is also de imen al
o public heal h (e.g., causing dep ession o anxie y) (Daza and Ilozumba 2022).
Finally, a ecen s udy wa ned ha he inc eased use o Cha GPT o ela ed AI-
d i en echnologies is supposed o c ea e immense e hical issues, including a
ising le el o disin o ma ion due o au oma ed ake news, massi e low-quali y
con en c ea ion, and a mo e indi ec communica ion be ween s akeholde s in he
socie y (Illia e al. 2023).
Ne e heless, he s eng h o hese mo al conce ns ela ed o AI echnologies
a ies om pe son o pe son. In pa icula , some people we e ound o ha e a
high echnological a ini y and a e less wo ied abou mo ali y issues o pos-
sible downsides o AI use (Pa asu aman and Colby 2015; Pun oni e al. 2021).
These indi iduals migh mainly ocus on he inno a i eness o AI and ha e li le
conce ns abou mo al iola ions ela ed o hei p i acy o eedom in decision-
making. In con as , o he consume s pe cei e a high isk and a he dis us AI.
This g oup is mo e likely o belie e ha AI is employed o decei e hem o ake
o e con ol (Bu on e al. 2020; Pa asu aman and Colby 2015). In gene al, mo al
judgemen s we e ound o in luence consume s’ pe cep ions and beha io (Finkel
and K äme 2022; Sche me ho n 2002; Siau and Wang 2020). Resea ch showed
ha pe cep ion o (non-) e hical beha io o a company is an impo an ac o
du ing he pu chase decision p ocess. Indi iduals ewa ded a company’s e hical
beha io by showing a highe willingness o pu chase and by paying highe p ices
o p oduc s (C eye and Ross 1997). Mo eo e , a ecen s udy in he ela ed ield
o humanoid obo s e ealed ha consume s’ mo ali y pe cep ions posi i ely
in luenced obo c edibili y a ibu ions (Finkel and K äme 2022). Simila ly,
ela ed o ideo news, posi i e mo ali y judgmen s we e ound o lead o highe
message c edibili y (Nelson and Pa k 2015).
Building on hese esul s, we expec ha mo al judgemen s will in luence mes-
sage c edibili y pe cep ions and downs eam a i udes and beha io s. In pa icula ,
we ocus on pe cei ed mo ali y o AI use, which ela es o consume s’ e alua ion
o how mo ally accep able a company’s AI use is o hem. When people pe cei e
companies’ AI use as immo al (i.e. low mo ali y), he use and decla a ion o au ho -
ship o ms wi h AI in ol emen (i.e., AI o human-AI collabo a i e au ho ships) is
sough o ha m message c edibili y pe cep ions. In con as , when people pe cei e
companies’ AI use as mo ally accep able (i.e., high mo ali y), he ac ual use o
AI as sole au ho o co-au ho should no be an e hical issue. As hese consume s
exhibi lowe mo al objec ions o his kind o AI use, AI should also be pe cei ed as
a c edible (co-)au ho , simila o a adi ional human au ho (C eye and Ross 1997).
The e o e, high mo ali y pe cep ions a e expec ed o dele e he nega i e e ec s o
au ho ships on message c edibili y whe e AI is in ol ed. Thus, we hypo hesize:
H3 Pe cei ed mo ali y o AI use mode a es he ela ionship be ween au ho ship ype
and message c edibili y: In case o low pe cei ed mo ali y o AI use, message c ed-
ibili y is lowe o au ho ships whe e AI is in ol ed han o human au ho ship, and
393
M.Haup e al.
1 3
he e is no di e ence in message c edibili y ac oss au ho ship ypes when pe cei ed
mo ali y o AI use is high.
Figu e1 depic s he concep ual model.
3 S udy 1
3.1 Pa icipan s andp ocedu e
To examine he p oposed causal ela ionships, we c ea ed an expe imen and embedded
i in o an online su ey (Hulland e al. 2018). In exchange o a small compensa ion ($
0.75), pa icipan s (wi h a 95% app o al a e in o me asks) we e ec ui ed om he
pla o m P oli ic. P oli ic is one o he la ges online pla o ms wi h o e 130,000 pa -
icipan s and widely used in managemen esea ch o conduc su eys o expe imen s.
These pla o ms gene ally each a mo e di e se popula ion han adi ional sampling
me hods and allow a qui e apid and inexpensi e da a collec ion (Gosling and Mason
2015). In a la ge compa a i e s udy wi h six majo esea ch pla o ms and panels, Pee
e al. (2022) con i med he da a quali y o P oli ic o academic esea ch. To con ol
o possible e ec s om a esponden ’s coun y o o igin, we ec ui ed pa icipan s
wi h English as na i e language om he U.S. and UK. These coun ies we e chosen
as many AI- ela ed s udies a e based on one o hese Wes e n coun ies and he pool o
esponden s was la ge enough o ensu e a a ie y o pa icipan s (Fig.1).
A e excluding pa icipan s who ailed he a en ion check (i.e., “I you ead
his, please p ess bu on 4”), he inal sample consis ed o 243 pa icipan s (54.3%
emale, Mage = 35yea s, SDage = 18.29). As scena io, esponden s we e exposed o a
p oduc in o ma ion websi e (i.e., depic ing in o ma ion abou a jeans) om a ic i-
ious clo hing company (see Fig.4 in he appendix). We used a simula ed company
name and websi e o exclude possibly con ounding e ec s due o p io consume
expe iences o a achmen s wi h a eal b and. Mo eo e , he jeans scena io was cho-
sen as i ep esen s a common p oduc in he ield o consume goods and does no
end o be a gende -speci ic p oduc .
Fig. 1. Concep ual model
394
1 3
Consume esponses ohuman‑AI collabo a ion a o ganiza ional…
To design he scena io con en , we ha e e iewed he design o leading online
clo hing companies (based on he anking o he op e-comme ce s o es in he ash-
ion indus y based on e enue in 2022; ECDB 2023). We ha e included he mos -
common ea u es o hese websi es o c ea e a ealis ic appea ance. Mo eo e , we
conduc ed a p e- es wi h en consume s who a e expe ienced in ashion online
shopping. They con i med ha he websi e c ea ed esembles hose o common
clo hing companies.
The websi e was equal ac oss all condi ions, excep o he au ho label. Respond-
en s we e andomly assigned o one o ou expe imen al condi ions, and ead one
o he ollowing au ho desc ip ions: The ex was c ea ed by (1) a human au ho
(label: “W i en by Ma y Smi h”), (2) AI-suppo ed human au ho ship (label: “W i -
en by Ma y Smi h suppo ed by A i icial In elligence”), (3) a human-con olled AI
au ho ship (label: “Gene a ed by A i icial In elligence con olled by Ma y Smi h”,
(4) an AI au ho (label: “Gene a ed by A i icial In elligence”). The au ho labels
we e delibe a ely p esen ed wi hou u he de ails abou he o m o suppo o con-
ol. A p e- es wi h se en quali a i e in e iews wi h business manage s con i med
ha manage s would label he human-AI collabo a ion o m wi hou any u he
in o ma ion. The e o e, he labels used could ep esen a likely business p ac ice.
Mo eo e , he manage s acknowledged ha human con ol e e s o a inal check o
con en e aci y and indica es human esponsibili y. In con as , AI suppo ( o a
human) indica es ha AI helps wi h asks such as ex e inemen , co ec g amma ,
and spelling. In sum, hese esul s suppo he heo e ical ope a ionaliza ion o he
wo labels (see chap e 2.2).
A e seeing he espec i e scena io, pa icipan s we e asked o a e hei
pe cei ed message c edibili y (Appelman and Sunda 2016; Obe mille e al.
2005) wi h ou i ems on a 7-poin Like scale ( om 1 = “s ongly disag ee” o
7 = “s ongly ag ee”). Fu he mo e, h ee i ems we e used o assess esponden s’
a i ude owa ds he company (Da ke e al. 2008). Nex , we in eg a ed an a en ion
check i em and e alua ed he case ealism wi h wo i ems om Wagne e al. (2009),
namely, “I belie e ha he desc ibed si ua ion could happen in eal li e” and “I could
imagine eading a ex like he one p esen ed ea lie in eal li e” (α = 0.86; M: 5.26,
SD: 1.46). Finally, we asked o pa icipan s’ age, gende , and educa ion. No signi i-
can di e ences we e ound be ween he au ho g oups ega ding hese h ee con ol
a iables (each p > 0.1), sugges ing a success ul andomiza ion. All psychome ic
measu es we e abo e he ecommended le els (see Table2), indica ing cons uc
eliabili y and alidi y (Hulland e al. 2018).
As manipula ion check, esponden s we e asked o es ima e he sha e o human
e sus AI inpu . Figu e2 illus a es he means, e lec ing he expec ed o de . Resul s
o an ANOVA compa ing he ou au ho ypes showed ha people pe cei ed ha
w i ing sha es di e be ween he au ho ypes (F(3,239) = 89.24, p < 0.001). Pos -
hoc es s (Bon e oni) showed ha all au ho g oups we e pe cei ed signi ican ly
di e en om each o he (each p < 0.001)—excep o one. The di e ence be ween
sole AI au ho ship and AI con olled by human we e no di e en (p = 0.13).
Mo eo e , o e alua e he e ec o au ho ship ypes on pe cep ions o human
con ol o e AI, esponden s had o indica e “who had he inal esponsibili y o
he ex ”, anging om 1 = AI o 9 = Human) (see Fig. 2). Fo he ANOVA, he
395
M.Haup e al.
1 3
homogenei y o a iances was no gi en (Le ene’s F = 11.14, p < 0.001). To ade-
qua ely con ol o his, we used he ecommended Welch es and Games-Howell
pos -hoc es s (Toma ken and Se lin 1986). Resul s e ealed signi ican di e ences
Table 2 Scale i ems and s a is ics
Cons uc name and i ems S anda dized
loadings
S udy 1 S udy 2
Message C edibili y (S udy 1/S udy 2: α = .88/.91; CR = .88/.86; AVE = .65/.61)
This ex …
… is gene ally u h ul 0.75 0.75
… lea es one eeling accu a ely in o med 0.79 0.74
… is belie able 0.84 0.85
… is au hen ic 0.83 0.77
A i ude owa ds he company (S udy 1/S udy 2: α = .95/.89; CR = .92/.85; AVE = .79/.66)
This company is a good company 0.88 0.85
This company is a nice company 0.90 0.84
I like he company 0.89 0.74
Mo ali y o AI use (S udy 2: α = .93; CR = .93; AVE = .77)
Companies using a i icial in elligence (AI) in ma ke ing ex s a e…
C uel (1) e sus Kind-hea ed (7) 0.88
Immo al (1) e sus Mo al (7) 0.90
Unca ing (1) e sus Ca ing (7) 0.83
Une hical (1) e sus E hical (7) 0.89
Fig. 2 S udy 1 Consume s’ pe cep ions o Sha e o Inpu and Le el o Con ol. Scale anging om 1 = AI
o 9 = Human
396
1 3
Consume esponses ohuman‑AI collabo a ion a o ganiza ional…
be ween he g oups (FWelch (3,131.41) = 13.80, p < 0.001). Human con ol was high-
es in he case o sole human au ho ship as no AI was in ol ed, ollowed by he
human-con olled AI au ho ship and he AI-suppo ed human au ho ship. Ob i-
ously, he lowes le el o human con ol was assigned o sole AI au ho ship. Pos -
hoc es s (Games-Howell) showed ha human con ol o e AI was signi ican ly
highe o human au ho ship e sus AI-suppo ed human au ho ship o AI (each
p < 0.001), bu no signi ican ly di e en om he human-con olled AI au ho ship
(p = 0.25).
3.2 Resul s
To es H1 and H2 in one comp ehensi e model, we an a media ion model (PRO-
CESS model 4 wi h 5,000 boo s apped samples and 95% CI’s (Hayes 2018)). The
au ho ypes we e he mul ica ego ical independen a iable, message c edibili y
was he media o , a i ude owa ds he company was he ou come a iable, and age,
and gende , educa ion, and coun y o o igin we e co a ia es. Rela ed o he au ho
ypes, he human au ho was selec ed as base case o mee he pe cep ions and a i-
udes ha we e gi en be o e AI in eg a ion. Compa ed o a human-au ho ed mes-
sage, esponden s pe cei ed an AI au ho (b = − 0.45, p < 0.05) and an AI-suppo ed
human au ho (b = − 0.71, p < 0.005) as signi ican ly less c edible. In con as , a
human-con olled AI au ho was no pe cei ed signi ican ly di e en (p = 0.21). All
co a ia es had no signi ican impac on message c edibili y (each p > 0.1). Thus,
H1a and H1b could be suppo ed.
In u n, message c edibili y had a signi ican impac on a i ude owa ds he com-
pany (b = 0.70, p < 0.001)—suppo ing H2.
The o al e ec s o au ho ship ypes on a i ude owa ds he company we e sig-
ni ican ly nega i e o AI au ho ship (b = − 0.54, p < 0.05) and o he human au ho
suppo ed by AI (b = − 0.57, p < 0.05), bu no signi ican o a human-con olled AI
au ho (p = 0.15). No ably, no di ec e ec s o au ho ship ype on a i ude owa ds
he company we e signi ican (each p > 0.1), indica ing a ull media ion o he o -
me wo au ho ypes. Rega ding he co a ia es, no co a ia e had a o al e ec on
a i ude owa ds he company (each p > 0.1).
In sum, bo h an AI au ho ship and an AI-suppo ed human au ho ship ha e nega-
i e e ec s on eade s’ a i ude owa ds he company, media ed by lowe message
c edibili y pe cep ions—whe eas a human-con olled AI au ho ship had no such a
nega i e e ec ( s. a human au ho ).
4 S udy 2
S udy 2 aimed o alida e he esul s o S udy 1 in ano he business- ela ed con ex .
In pa icula , a company’s ision s a emen was chosen as a highly ele an message
exp essing company alues and a ge s. Fu he mo e, S udy 2 assessed he mode a -
ing e ec s o mo ali y o AI use (H3) on message and company e alua ions.
397
M.Haup e al.
1 3
4.1 Pa icipan s andp ocedu e
In exchange o a mone a y compensa ion ($ 0.75), pa icipan s om he U.S. we e
ec ui ed ia Amazon mTu k, and andomly assigned o one o he condi ions in he
4 (au ho : human s. human suppo ed by AI s. AI con olled by human s. AI) × 2
(indus y: ki chen s. clo hing) be ween-subjec s design. We chose mTu k as one o
he mos p ominen online pla o ms o social science and managemen esea ch
o al e he pla o m used in s udy 1 and he e o e con ol o possible con ounding
e ec s. Responden s had o su pass 95% comple ion a e o o me asks and iden-
i y English as hei na i e language.
A e excluding esponden s who ailed he a en ion check o he co ec ec-
ogni ion o he au ho (s), he inal da ase consis ed o n = 217 esponden s (46.5%
emales, Mage = 38yea s, SD = 11.37, wi h an equal o highe han 95% o me asks
app o al a io). We al e ed he indus y o con ol o possible e ec s due o a mo e
echnical o emo ional business. Resul s o wo independen samples - es s showed
ha he indus y ype did no in luence message c edibili y (p = 0.31), bu he mes-
sage om he ashion indus y was a ed ma ginally mo e posi i e han om he
ki chen indus y (MFashion: 5.59, SD: 1.66, MKi chen: 5.26, SD: 1.38, (215) = − 1.86,
p < 0.1).
A e accessing he su ey, esponden s we e asked o ead a ic i ious scena io
ega ding a company’s ision s a emen ha was p esen ed on a websi e (see Fig.5
in he appendix). Again, we simula ed he s imuli o exclude possible con ounding
e ec s (as in s udy 1). To design his scena io, we compa ed elemen s om se -
e al la ge e-comme ce companies om he u ni u e and ashion indus y (ECDB
Fu ni u e 2023, ECDB Fashion 2023). As in s udy 1, a p e- es wi h en espond-
en s con i med ha he design o he ic i ious websi e is likely o be ealis ic o a
ki chen o ashion company. We used a ision s a emen as con ex as i ep esen s a
ele an business message and a common online con en o many companies. While
holding he ex equal ac oss he g oups, we al e ed he au ho ypes and he indus-
y o he espec i e company. As measu es, pa icipan s’ pe cep ions abou mes-
sage c edibili y, and a i ude owa ds he company we e assessed using he same
i ems as in S udy 1. Addi ionally, pe cei ed mo ali y o companies’ AI use o c ea e
ma ke ing con en was e alua ed wi h a 4-i em 7-poin seman ic di e en ial (Olson
e al. 2016). Finally, esponden s en e ed hei age, gende , and educa ion. All i ems
and ac o loadings a e shown in Table2. All psychome ic measu es we e abo e
he ecommended le els (see Table2), sugges ing cons uc eliabili y and alid-
i y (Hulland e al. 2018). Mo eo e , he expe imen g oups p esen ed no signi ican
di e ences ega ding he con ol a iables (each p > 0.1), sugges ing a success ul
andomiza ion.
As manipula ion check, eade s o he di e en au ho g oups had o e alua e he
human ( s. AI) sha e o inpu . We used he Welch es and Games-Howell pos -hoc
es s because he assump ion o homogenei y o a iances was iola ed. The pe -
cei ed sha e o human o AI-inpu di e ed signi ican ly ac oss he g oups (FWelch
(3,114.63) = 180.67, p < 0.001). Pos -hoc es s (Games-Howell) showed ha all
g oups a e signi ican ly di e en om each o he (p < 0.05). As expec ed, people in
he human au ho scena io pe cei ed he highes sha e o human-inpu (M: 8.29,
398
1 3
Consume esponses ohuman‑AI collabo a ion a o ganiza ional…
SD: 1.32), ollowed by he AI-suppo ed human au ho ship (M: 4.42; SD: 2.06) and
he human-con olled AI au ho (M: 3.39, SD: 1.88), and pe cei ed he lowes sha e
o human au ho ship in he AI au ho ship scena io (M: 2.11; SD: 1.39). Rega ding
human con ol o e AI (i.e., “who had he inal esponsibili y o he ex ”, ang-
ing om 1 = AI o 9 = Human), esul s we e again signi ican ly di e en be ween he
g oups (FWelch(3,105.34) = 23.92, p < 0.001). Human con ol was highes in he case
o sole human au ho ship as no AI was in ol ed, ollowed by human-con olled AI
au ho ship, he AI-suppo ed human au ho ship, and was leas o sole AI au ho -
ship. Pos -hoc es s (Games-Howell) showed ha human con ol o e AI was sig-
ni ican ly highe o human au ho ship s. AI-suppo ed human au ho ship o s. AI
(each p < 0.001), bu no signi ican ly di e en om a human-con olled AI au ho -
ship (p = 0.62).
Scena io ealism was assessed wi h wo i ems om S udy 1. Again, all scena ios
we e pe cei ed as ealis ic (α = 0.81; M: 5.97, SD: 1.00), and ealism sco es did no
di e be ween he au ho g oups (p > 0.1). Responden s con i med ha hey “wan
o know abou he use o AI” (M: 5.34, SD: 1.48 on a 7-poin scale). Fu he mo e,
he call o anspa ency (Eu opean Pa liamen 2023; Jobin e al. 2019) was also
e lec ed, as esponden s ag eed ha “companies should be obliged o disclose he
use o AI” (M: 5.18, SD: 1.59). On a e age, people seem o pe cei e companies’ AI
usage as mo ally a he accep able (M: 5.00, SD: 1.36), and his pe cep ion did no
di e among he au ho ship g oups (p > 0.1).
4.2 Resul s
To assess he hypo hesized e ec s o he au ho s on message c edibili y (H1) and
subsequen ly on a i ude owa ds he company (H2), and he mode a ing e ec o
mo ali y (H3) in one comp ehensi e model, we used a mode a ed media ion analy-
sis wi h PROCESS (model 8 wi h 5,000 boo s apped samples and 95% CIs (Hayes
2018)) based on he same se up as in S udy 1. As mode a o , we included mo ali y
o AI use, and we con olled o age, gende , and indus y ype. Table3 illus a es
he esul s.
Responden s a ed he ex o sole AI au ho ship as signi ican ly less c edible
han a (sole) human-au ho ed ex (b = − 2.95, p < 0.005). Again, he collabo a i e
au ho ships we e pe cei ed di e en ly: A ex om human-con olled AI au ho ship
was no signi ican ly di e en om a human au ho ship (p = 0.65), bu a ex om an
AI-suppo ed human au ho ship was a ed signi ican ly wo se (b = − 1.73, p < 0.05).
Thus, al hough consume s acknowledged ha he la e o m con ains a highe sha e
o human inpu , his e sion was a ed less c edible han a collabo a ion o ma wi h
less human inpu (bu human con ol). The co a ia es age, gende , indus y ype, and
educa ion had no impac on message c edibili y (p > 0.1).
In u n, message c edibili y had a signi ican impac on a i ude owa ds he com-
pany (b = 0.70, p < 0.001). None o he au ho ypes had a di ec impac on a i-
ude owa ds he company (each p > 0.1, see Table 3), indica ing a ull media ion
ia message c edibili y. A i udes owa ds he company we e no in luenced by age,
399
M.Haup e al.
1 3
gende , o educa ion (each p > 0.1), while he ashion indus y ( s. ki chen) ma gin-
ally inc eased he a i udinal e alua ions (b = 0.21, p < 0.1).
In sum, hese esul s suppo H1 (a and b) and H2 again. Pe cep ions o human
con ol o e AI we e ound o be mo e ele an han sha e o human inpu when
e alua ing message c edibili y. In pa icula , human-AI collabo a ion including
explici human con ol was ound o be equally c edible as a sole human au ho -
ship, whe eas he collabo a ion wi h highe human inpu bu wi hou such a human
con ol (i.e., AI-suppo ed human au ho ) was a ed as less c edible. Thus, in a col-
labo a i e se ing, people we e ound o be a he insensi i e o human inpu , bu
sensi i e o human con ol o e AI (H1). In u n, s onge message c edibili y led o
mo e a o able a i udes owa ds he company (H2).
Table 3 S udy 2. Condi ional p ocess model o message c edibili y as media o , mo ali y o AI use as
mode a o , and a i ude owa ds he company as ou come
Condi ional indi ec e ec (s) o X (au ho ypes) on Y (a i ude owa ds he company) a alues o he
mode a o (M−1SD, M, M+1SD). Boo s ap 95 pe cen con idence in e als o condi ional indi ec
e ec s. †p < .1, *p < .05, **p < .01, ***p < .001, M mean, SD S anda d de ia ion, n.s. no signi ican
Media o Ou come
Message c edibili y A i ude owa ds he company
b b
X1: Human suppo ed by AI e sus Human − 1.72 − 2.07* − 0.48 − 0.71n.s
X2: AI con olled by Human e sus Human 0.39 0.47n.s − 0.71 − 1.07n.s
X3: AI e sus Human − 2.93 − 3.33** − 0.01 − 0.01n.s
W: Mo ali y o AI use 0.40 3.22** − 0.03 − 0.26n.s
M: Message c edibili y − − 0.70 12.48***
X1*W 0.22 1.35n.s 0.06 0.44n.s
X2*W − 0.12 − 0.75n.s 0.12 0.90n.s
X3*W 0.42 2.38* − 0.01 − 0.07n.s
COV: Age 0.01 0.80n.s − 0.00 − 0.50n.s
COV: Gende − 0.13 − 0.91n.s 0.00 0.02n.s
COV: Indus y ype 0.12 0.79n.s 0.20 1.70
Mo ali y bLowe Uppe
X1: Human suppo ed by AI e sus Human 3.67 − 0.64 − 1.19 − 0.09
5.02 − 0.43 − 0.73 − 0.15
6.37 − 0.23 − 0.53 0.06
X2: AI con olled by Human e sus Human 3.67 − 0.03 − 0.59 0.43
5.02 − 0.15 − 0.42 0.11
6.37 − 0.26 − 0.54 0.03
X3: AI e sus Human 3.67 − 0.96 − 1.47 − 0.47
5.02 − 0.56 − 0.86 − 0.27
6.37 − 0.16 − 0.47 0.17†
400
1 3
Consume esponses ohuman‑AI collabo a ion a o ganiza ional…
Fou h, in ou s udy, he disclosu e o human con ol o e AI in he scena ios
does delibe a ely no include he o m o con ol implemen a ion o de ails o i s
execu ion. Howe e , acco ding o Nyholm (2022), di e en o ms o con ol exis
and migh hus be e alua ed di e en ly. Fu u e s udies could e alua e he impac o
di e en con ol o ms o con ol amings on consume s’ pe cep ions and company
assessmen s.
Finally, his s udy uses a c oss-sec ional design and ep esen s a cu en snap-
sho on his dynamic opic. As AI is con inuously and apidly e ol ing, u u e
esea ch migh in es iga e long- e m e ec s, o ins ance whe he amilia iza ion
wi h AI-gene a ed con en leads o mo e a o able AI e alua ions. Pa allel o he
g ow h o AI ools, esea ch om di e en disciplines should o ches a e e o s
o explo e u he e ec s o human-AI collabo a ions and he human con ol unc-
ion o e AI, o achie e an e hical and bene icial use o AI.
Appendix
See Figs. 4 and 5.
Fig. 4 S udy 1. Exempla y scena io
407
M.Haup e al.
1 3
Funding Open Access unding enabled and o ganized by P ojek DEAL.
Da a a ailabili y The da a ha suppo he indings o his s udy a e a ailable om he co esponding
au ho upon eques .
Decla a ions
Con lic o in e es The au ho s did no ecei e suppo om any o ganiza ion o he submi ed wo k. The
au ho s ha e no compe ing in e es s o decla e ha a e ele an o he con en o his a icle.
Open Access This a icle is licensed unde a C ea i e Commons A ibu ion 4.0 In e na ional License,
which pe mi s use, sha ing, adap a ion, dis ibu ion and ep oduc ion in any medium o o ma , as long as
you gi e app op ia e c edi o he o iginal au ho (s) and he sou ce, p o ide a link o he C ea i e Com-
mons licence, and indica e i changes we e made. The images o o he hi d pa y ma e ial in his a icle
a e included in he a icle’s C ea i e Commons licence, unless indica ed o he wise in a c edi line o he
ma e ial. I ma e ial is no included in he a icle’s C ea i e Commons licence and you in ended use is
no pe mi ed by s a u o y egula ion o exceeds he pe mi ed use, you will need o ob ain pe mission
di ec ly om he copy igh holde . To iew a copy o his licence, isi h p:// c ea i eco mmons. o g/ licen
ses/ by/4. 0/.
Re e ences
Al onseca M, Ceb ian M, Fe nandez An a A, Co iello L, Abeliuk A, Rahwan I (2021) Supe in elligence
canno be con ained: lessons om compu abili y heo y. J A i In ell Res 70:65–76. h ps:// doi. o g/
10. 1613/ jai .1. 12202
Ameen N, Sha ma GD, Ta ba S, Rao A, Chop a R (2022) Towa d ad ancing heo y on c ea i i y in ma -
ke ing and a i icial in elligence. Psychol Ma k 39(9):1802–1825. h ps:// doi. o g/ 10. 1002/ ma . 21699
Appelman A, Sunda SS (2016) Measu ing message c edibili y. Jou nal Mass Commun Q 93(1):59–79.
h ps:// doi. o g/ 10. 1177/ 10776 99015 606057
Fig. 5 S udy 2. Exempla y scena io
408
1 3
Consume esponses ohuman‑AI collabo a ion a o ganiza ional…
Baile W, Thallinge G, K awa ik V, Schell K, E el halne V (2022) AI o he media indus y applica-
ion po en ial and au oma ion le els. In: Hue B, Þó Jónsson B, Gu in C, T an MT, Dang-Nguyen
DT, Hu AMC, Huynh Thi Thanh B (eds) Lec u e no es in compu e science. mul imedia modeling.
Sp inge In e na ional Publishing, Cham, pp 109–118
Billo e S, Anisimo a T (2021) Panic buying esea ch: a sys ema ic li e a u e e iew and u u e esea ch
agenda. In J Consum S ud 45(4):777–804. h ps:// doi. o g/ 10. 1111/ ijcs. 12669
Bu on JW, S ein M-K, Jensen TB (2020) A sys ema ic e iew o algo i hm a e sion in augmen ed deci-
sion making. J Beha Decis Mak 33(2):220–239. h ps:// doi. o g/ 10. 1002/ bdm. 2155
Cas elo N, Bos MW, Lehmann DR (2019) Task-dependen algo i hm a e sion. J Ma k Res 56(5):809–
825. h ps:// doi. o g/ 10. 1177/ 00222 43719 851788
Chan-Olms ed SM (2019) A e iew o a i icial in elligence adop ions in he media indus y. In J Media
Manag 21(3–4):193–215. h ps:// doi. o g/ 10. 1080/ 14241 277. 2019. 16956 19
Chen CY, Lee L, Yap AJ (2017) Con ol dep i a ion mo i a es acquisi ion o u ili a ian p oduc s. J Con-
sum Res 43:1031–1047. h ps:// doi. o g/ 10. 1093/ jc / ucw068
Chui M, Robe s R, Yee L (2022) Gene a i e AI is he e: how ools like Cha Gp could change you busi-
ness. www. mckin sey. com/ capab ili i es/ quan umbla ck/ ou - insig h s/ gene a i e- ai- is- he e- how- ools-
like- cha g p - could- change- you - busin ess/
C eye EH, Ross WT (1997) The in luence o i m beha io on pu chase in en ion: do consume s eally
ca e abou business e hics? J Consum Ma k 14(6):421–432. h ps:// doi. o g/ 10. 1108/ 07363 76971
01859 99
Cu igh KM, Wu EC (2023) In and ou o con ol: pe sonal con ol and consume beha io . Consum
Psychol Re 6(1):33–51. h ps:// doi. o g/ 10. 1002/ a cp. 1083
Da ke PR, Ashwo h L, Ri chie RJ (2008) Damage om co ec i e ad e ising: causes and cu es. J Ma k
72(6):81–97
Da enpo T, Guha A, G ewal D, B essgo T (2020) How a i icial in elligence will change he u u e o
ma ke ing. J Acad Ma k Sci 48(1):24–42. h ps:// doi. o g/ 10. 1007/ s11747- 019- 00696-0
Daza MT, Ilozumba UJ (2022) A su ey o AI e hics in business li e a u e: maps and ends be ween
2000 and 2021. F on Psychol 13:1042661. h ps:// doi. o g/ 10. 3389/ psyg. 2022. 10426 61
de C eme D, Kaspa o G (2021) The e hical AI—pa adox: why be e echnology needs mo e and no
less human esponsibili y. AI E hics. h ps:// doi. o g/ 10. 1007/ s43681- 021- 00075-y
Die o s BJ, Simmons JP, Massey C (2015) Algo i hm a e sion: people e oneously a oid algo i hms
a e seeing hem e . J Exp Psychol Gen 144(1):114–126. h ps:// doi. o g/ 10. 1037/ xge00 00033
Die o s BJ, Simmons JP, Massey C (2016) O e coming algo i hm a e sion: people will use impe ec
algo i hms i hey can (E en sligh ly) modi y hem. Manage Sci 64(3):1155–1170. h ps:// doi. o g/
10. 1287/ mnsc. 2016. 2643
Dwi edi YK, Kshe i N, Hughes L, Slade EL, Jeya aj A, Ka AK, Baabdullah AM, Koohang A,
Ragha an V, Ahuja M, Albanna H, Albash awi MA, Al-Busaidi AS, Balak ishnan J, Ba le e Y,
Basu S, Bose I, B ooks L, Buhalis D, W igh R (2023) “So wha i Cha GPT w o e i ?” Mul idisci-
plina y pe spec i es on oppo uni ies, challenges and implica ions o gene a i e con e sa ional AI
o esea ch, p ac ice and policy. In J In Manage 71:102642. h ps:// doi. o g/ 10. 1016/j. ijin omg .
2023. 102642
Fan M, Yang X, Yu T, Liao QV, Zhao J (2022) Human-AI collabo a ion o UX e alua ion: e ec s o
explana ion and synch oniza ion. P oc ACM Hum-Compu In e ac 6(CSCW1):1–32. h ps:// doi.
o g/ 10. 1145/ 35129 43
ECDB Fashion (2023) eComme ceDB. S o e Ranking. Top eComme ce s o es in he Fashion ma ke ,
Global e enue 2022 in USD.h ps:// ecomm e cedb. com/ anki ng/ s o es/ ww/ ashi on
Finkel M, K äme NC (2022) Humanoid obo s – a i icial. Human-like. C edible? empi ical compa i-
sons o sou ce c edibili y a ibu ions be ween humans, humanoid obo s, and non-human-like
de ices. In J Soc Robo 14(6):1397–1411. h ps:// doi. o g/ 10. 1007/ s12369- 022- 00879-w
F ey CB, Osbo ne MA (2017) The u u e o employmen : how suscep ible a e jobs o compu e isa ion?
Technol Fo ecas Soc Chang 114:254–280. h ps:// doi. o g/ 10. 1016/j. ech o e. 2016. 08. 019
F iesen JP, Kay AC, Eibach RP, Galinsky AD (2014) Seeking s uc u e in social o ganiza ion: compen-
sa o y con ol and he psychological ad an ages o hie a chy. J Pe s Soc Psychol 106(4):590–609.
h ps:// doi. o g/ 10. 1037/ a0035 620
Fügene A, G ahl J, Gup a A, Ke e W (2022) Cogni i e challenges in human-a i icial in elligence col-
labo a ion: in es iga ing he pa h owa d p oduc i e delega ion. In Sys Res 33(2):678–696. h ps://
doi. o g/ 10. 1287/ is e. 2021. 1079
409
M.Haup e al.
1 3
ECDB Fu ni u e (2023) eComme ceDB. S o e Ranking. Top eComme ce s o es in he Fu ni u e ma ke ,
Global e enue 2022 in USD. h ps:// ecomm e cedb. com/ anki ng/ s o es/ ww/ u ni u e
Gosling SD, Mason W (2015) In e ne esea ch in psychology. Annu Re Psychol 66:877–902. h ps://
doi. o g/ 10. 1146/ annu e - psych- 010814- 015321
G ae e A, Bohlken N (2020) Au oma ed jou nalism: a me a-analysis o eade s’ pe cep ions o human-
w i en in compa ison o au oma ed news. Media Commun 8(3):50–59. h ps:// doi. o g/ 10. 17645/
mac. 8i3. 3019
G ae e A, Haim M, Haa mann B, B osius H-B (2016) Reade s’ pe cep ion o compu e -gene a ed news:
c edibili y, expe ise, and eadabili y. Jou nalism 19(5):595–610. h ps:// doi. o g/ 10. 1177/ 14648
84916 641269
Hagendo T (2020) The e hics o AI e hics: an e alua ion o guidelines. Mind Mach 30(1):99–120.
h ps:// doi. o g/ 10. 1007/ s11023- 020- 09517-8
Ha mon RR, Coney KA (1982) The pe suasi e e ec s o sou ce c edibili y in buy and lease si ua ions. J
Ma k Res 19(2):255–260. h ps:// doi. o g/ 10. 1177/ 00222 43782 01900 209
Hassani H, Sil a ES, Unge S, Taj Mazinani M, Mac Feely S (2020) A i icial in elligence (AI) o in el-
ligence augmen a ion (IA): wha is he u u e? AI 1(2):143–155. h ps:// doi. o g/ 10. 3390/ ai102 0008
Hayes AF (2018) In oduc ion o media ion, mode a ion and condi ional p ocess analysis. A eg ession-
based app oach, 2nd edn. The Guil o d P ess, New Yo k
He mann E (2022) Le e aging a i icial in elligence in ma ke ing o social good-an e hical pe spec i e. J
Bus E hics JBE 179(1):43–61. h ps:// doi. o g/ 10. 1007/ s10551- 021- 04843-y
Ho land CI, Janis IL, Kelley HL (1953) Communica ion and pe suasion; psychological s udies o opinion
change. Yale Uni e si y P ess, New Ha en
Huang M-H, Rus RT (2022) A amewo k o collabo a i e a i icial in elligence in ma ke ing. J Re ail
98(2):209–223. h ps:// doi. o g/ 10. 1016/j. j e ai. 2021. 03. 001
Hulland J, Baumga ne H, Smi h KM (2018) Ma ke ing su ey esea ch bes p ac ices: e idence and
ecommenda ions om a e iew o JAMS a icles. J Acad Ma k Sci 46(1):92–108. h ps:// doi. o g/
10. 1007/ s11747- 017- 0532-y
Illia L, Colleoni E, Zyglidopoulos S (2023) E hical implica ions o ex gene a ion in he age o a i icial
in elligence. Bus E hics En i on Responsib 32(1):201–210. h ps:// doi. o g/ 10. 1111/ bee . 12479
Ismagilo a E, Slade E, Rana NP, Dwi edi YK (2020) The e ec o cha ac e is ics o sou ce c edibili y on
consume beha iou : a me a-analysis. J Re ail Consum Se 53:101736. h ps:// doi. o g/ 10. 1016/j.
j e c onse . 2019. 01. 005
Jobin A, Ienca M, Vayena E (2019) The global landscape o AI e hics guidelines. Na Mach In ell
1(9):389–399. h ps:// doi. o g/ 10. 1038/ s42256- 019- 0088-2
Kahn I (2019) Küns liche in elligenz im con en ma ke ing. In: Wesselmann M (ed) Con en Gekonn :
s a egie, o ganisa ion, umse zung, ROI-messung und allbeispiele aus de p axis. Sp inge Gable ,
Cham, pp 211–225
Kanbach DK, Heiduk L, Bluehe G, Sch ei e M, Lahmann A (2023) The GenAI is ou o he bo le:
gene a i e a i icial in elligence om a business model inno a ion pe spec i e. Re Manage ial Sci.
h ps:// doi. o g/ 10. 1007/ s11846- 023- 00696-z
Kay AC, Gauche D, Napie JL, Callan MJ, Lau in K (2008) God and he go e nmen : es ing a com-
pensa o y con ol mechanism o he suppo o ex e nal sys ems. J Pe s Soc Psychol 95(1):18–35.
h ps:// doi. o g/ 10. 1037/ 0022- 3514. 95.1. 18
Köbis N, Mossink LD (2021) A i icial in elligence e sus Maya Angelou: expe imen al e idence
ha people canno di e en ia e AI-gene a ed om human-w i en poe y. Compu Hum Beha
114:106553. h ps:// doi. o g/ 10. 1016/j. chb. 2020. 106553
Kozine s RV, G e zel U (2021) Commen a y: a i icial in elligence: The ma ke e ’s dilemma. J Ma k
85(1):156–159. h ps:// doi. o g/ 10. 1177/ 00222 42920 972933
Lai Y, Kankanhalli A, Ong D (2021) Human-AI collabo a ion in heal hca e: a e iew and esea ch
Agenda. In: P oceedings o he annual Hawaii in e na ional con e ence on sys em sciences. Hawaii
In e na ional Con e ence on Sys em Sciences, pp 390–399. h ps:// doi. o g/ 10. 24251/ hicss. 2021. 046
Landau MJ, Kay AC, Whi son JA (2015) Compensa o y con ol and he appeal o a s uc u ed wo ld.
Psychol Bull 141(3):694–722. h ps:// doi. o g/ 10. 1037/ a0038 703
Lange EJ (1975) The illusion o con ol. J Pe s Soc Psychol 32(2):311–328. h ps:// doi. o g/ 10. 1037/
0022- 3514. 32.2. 311
Lange M, Lande s RN (2021) The u u e o a i icial in elligence a wo k: a e iew on e ec s o decision
au oma ion and augmen a ion on wo ke s a ge ed by algo i hms and hi d-pa y obse e s. Compu
Hum Beha 123:106878. h ps:// doi. o g/ 10. 1016/j. chb. 2021. 106878
410
1 3
Consume esponses ohuman‑AI collabo a ion a o ganiza ional…
Libai B, Ba Y, Gensle S, Ho acke CF, Kaplan A, Kö e hein ich K, K oll EB (2020) B a e new wo ld?
On AI and he managemen o cus ome ela ionships. J In e ac Ma k 51:44–56. h ps:// doi. o g/ 10.
1016/j. in ma . 2020. 04. 002
Longoni C, Cian L (2022) A i icial in elligence in u ili a ian s. hedonic con ex s: he “Wo d-o -
machine” e ec . J Ma k 86(1):91–108. h ps:// doi. o g/ 10. 1177/ 00222 42920 957347
Longoni C, Bonezzi A, Mo ewedge CK (2019) Resis ance o medical a i icial in elligence. J Consum
Res 46(4):629–650. h ps:// doi. o g/ 10. 1093/ jc / ucz013
Luo X, Tong S, Fang Z, Qu Z (2019) F on ie s: machines s. humans: he impac o a i icial in elligence
Cha bo disclosu e on cus ome pu chases. Ma k Sci. h ps:// doi. o g/ 10. 1287/ mksc. 2019. 1192
Nelson MR, Pa k J (2015) Publici y as co e ma ke ing? The ole o pe suasion knowledge and e hical
pe cep ions on belie s and c edibili y in a ideo news elease s o y. J Bus E hics 130(2):327–341.
h ps:// doi. o g/ 10. 1007/ s10551- 014- 2227-3
Newman A, Jones R (2006) Au ho ship o esea ch pape s: e hical and p o essional issues o sho - e m
esea che s. J Med E hics 32(7):420–423. h ps:// doi. o g/ 10. 1136/ jme. 2005. 012757
Nyholm S (2022) A new con ol p oblem? Humanoid obo s, a i icial in elligence, and he alue o con-
ol. AI E hics. h ps:// doi. o g/ 10. 1007/ s43681- 022- 00231-y
Obe mille C, Spangenbe g E, MacLachlan DL (2005) Ad skep icism: he consequences o disbelie . J
Ad e ising 34(3):7–17
Ohanian R (1990) Cons uc ion and alida ion o a scale o measu e celeb i y endo se s’ pe cei ed expe -
ise, us wo hiness, and a ac i eness. J Ad e 19(3):39–52. h ps:// doi. o g/ 10. 1080/ 00913 367.
1990. 10673 191
Olson JG, McFe an B, Mo ales AC, Dahl DW (2016) Weal h and wel a e: di e gen mo al eac ions o
e hical consume choices. J Consum Res 42(6):879–896. h ps:// doi. o g/ 10. 1093/ jc / uc 096
Olson P (2022) Google aces a se ious h ea om Cha GPT. h ps:// www. washi ng on pos . com/ busin ess/
ene gy/ google- aces-a- se io us- h ea - om- cha g p / 2022/ 12/ 07/ 363d2 440- 75 5- 11ed- a199- 927b3
34b93 9 _ s o y. h ml
Önkal D, Goodwin P, Thomson M, Gönül S, Pollock A (2009) The ela i e in luence o ad ice om
human expe s and s a is ical me hods on o ecas adjus men s. J Beha Decis Mak 22(4):390–409.
h ps:// doi. o g/ 10. 1002/ bdm. 637
Osbu g VS, Yogana han V, Kunz WH, Ta ba S (2022) Can (A)I gi e you a ide? De elopmen and ali-
da ion o he CRUISE amewo k o au onomous ehicle se ices. J Se Res. h ps:// doi. o g/ 10.
1177/ 10946 70522 11182 33
Pa asu aman A, Colby CL (2015) An upda ed and s eamlined echnology eadiness index. J Se Res
18(1):59–74. h ps:// doi. o g/ 10. 1177/ 10946 70514 539730
Eu opean Pa liamen (2023) EU AI Ac : i s egula ion on a i icial in elligence. Eu opean Pa liamen .
h ps:// www. eu op a l. eu opa. eu/ news/ en/ headl ines/ socie y/ 20230 601ST O93804/ eu- ai- ac - i s - egul
a ion- on- a i i cial- in el ligen ce
Paschen J, Wilson M, Fe ei a JJ (2020) Collabo a i e in elligence: how human and a i icial in elligence
c ea e alue along he B2B sales unnel. Bus Ho izons 63(3):403–414. h ps:// doi. o g/ 10. 1016/j.
busho . 2020. 01. 003
Pee E, Ro hschild D, Go don A, E e nden Z, Dame E (2022) Da a quali y o pla o ms and panels
o online beha io al esea ch. Beha Res Me hods 54(4):1643–1662. h ps:// doi. o g/ 10. 3758/ 2Fs13
428- 021- 01694-3
Pee E, Ro hschild D, Go don A, E e nden Z, Dame E (2022) Da a quali y o pla o ms and panels
o online beha io al esea ch. Beha Res Me hods 54:1643–1662. h ps:// doi. o g/ 10. 3758/
s13428- 021- 01694-3
Pun oni S, Reczek RW, Giesle M, Bo i S (2021) Consume s and a i icial in elligence: an expe ien ial
pe spec i e. J Ma k 85(1):131–151. h ps:// doi. o g/ 10. 1177/ 00222 42920 953847
Ra opoulos M, Hama iJ (2023) Human-AI collabo a ion in o ganisa ions: A li e a u e e iew on ena-
bling alue c ea ion. In: Eu opean con e ence on in o ma ion sys ems, p 31
Rai A (2020) Explainable AI: om black box o glass box. J Acad Ma k Sci 48(1):137–141. h ps:// doi.
o g/ 10. 1007/ s11747- 019- 00710-5
Rus RT (2020) The u u e o ma ke ing. In J Res Ma k 37(1):15–26. h ps:// doi. o g/ 10. 1016/j. ij es ma .
2019. 08. 002
Sales o ce (2023) A i icial in elligence. Businesses Adop ing AI Risk a ‘T us Gap’ wi h Cus ome s
– Sales o ce Repo . 28.08.2023. h ps:// www. sales o ce. com/ news/ s o i es/ cus o me - engag emen -
esea ch- 2023/
411
M.Haup e al.
1 3
San ana M, Díaz-Fe nández M (2023) Compe encies o he a i icial in elligence age: isualisa ion
o he s a e o he a and u u e pe spec i es. RMS 17(6):1971–2004. h ps:// doi. o g/ 10. 1007/
s11846- 022- 00613-w
San oni de Sio F, Van den Ho en J (2018) Meaning ul human con ol o e au onomous sys ems: a philo-
sophical accoun . F on Robo AI 5:15. h ps:// doi. o g/ 10. 3389/ ob . 2018. 00015
Sche me ho n JR (2002) Managemen , 7 h edn. Wiley, Hoboken
Schlei hJ, Ho mannH, No ku eM, CechmanekB (2022) Human in he loop in o ma ion ex ac ion
inc eases e iciency and us . h ps:// doi. o g/ 10. 18420/ muc20 22- mci- ws12- 249
Seo SH, G i in K, Young JE, Bun A, P en ice S, Lou ei o-Rod íguez V (2018) In es iga ing people’s
appo building and hinde ing beha io s when wo king wi h a collabo a i e obo . In J Soc Robo
10(1):147–161. h ps:// doi. o g/ 10. 1007/ s12369- 017- 0441-8
Shneide man B (2020) Human-cen e ed a i icial in elligence: eliable, sa e & us wo hy. In J Hum-
Compu In e ac 36(6):495–504. h ps:// doi. o g/ 10. 1080/ 10447 318. 2020. 17411 18
Siau K, Wang W (2020) A i icial in elligence (AI) e hics. J Da abase Manage 31(2):74–87. h ps:// doi.
o g/ 10. 4018/ JDM. 20200 40105
Skinne EA (1996) A guide o cons uc s o con ol. J Pe s Soc Psychol 71(3):549–570. h ps:// doi. o g/
10. 1037/ 0022- 3514. 71.3. 549
Sowa K, P zegalinska A, Ciechanowski L (2021) Cobo s in knowledge wo k. J Bus Res 125:135–142.
h ps:// doi. o g/ 10. 1016/j. jbus es. 2020. 11. 038
Sunda SS (2020) Rise o machine agency: a amewo k o s udying he psychology o human–AI in e -
ac ion (HAII). J Compu -Media Commun 25(1):74–88. h ps:// doi. o g/ 10. 1093/ jcmc/ zmz026
Tobia K, Nielsen A, S emi ze A (2021) When does physician use o AI inc ease liabili y? J Nucl Med
O Publ Soc Nucl Med 62(1):17–21. h ps:// doi. o g/ 10. 2967/ jnumed. 120. 256032
Toma ken AJ, Se lin RC (1986) Compa ison o ANOVA al e na i es unde a iance he e ogenei y and
speci ic noncen ali y s uc u es. Psychol Bull 99(1):90–99. h ps:// doi. o g/ 10. 1037/ 0033- 2909. 99.1.
90
Vak a sas D, Wang X (2021) A i icial in elligence in ad e ising c ea i i y. J Ad e 50(1):39–51. h ps://
doi. o g/ 10. 1080/ 00913 367. 2020. 18430 90
Visen in M, Pizzi G, Pichie i M (2019) Fake news, eal p oblems o b ands: he impac o con en u h-
ulness and sou ce c edibili y on consume s’ beha io al in en ions owa d he ad e ised b ands. J
In e ac Ma k 45:99–112. h ps:// doi. o g/ 10. 1016/j. in ma . 2018. 09. 001
Waddell TF (2018) A obo w o e his? Digi Jou nal 6(2):236–255. h ps:// doi. o g/ 10. 1080/ 21670 811.
2017. 13843 19
Waddell TF (2019) Can an algo i hm educe he pe cei ed bias o news? Tes ing he e ec o machine
a ibu ion on news eade s’ e alua ions o bias, an h opomo phism, and c edibili y. Jou nal Mass
Commun Q 96(1):82–100. h ps:// doi. o g/ 10. 1177/ 10776 99018 815891
Wagne T, Hennig-Thu au T, Rudolph T (2009) Does cus ome demo ion jeopa dize loyal y? J Ma k
73(3):69–85. h ps:// doi. o g/ 10. 1509/ jmkg. 73.3. 69
Wang AY, Wang D, D ozdal J, Mulle M, Pa k S, Weisz JD, Liu X, Wu L, Dugan C (2022) Documen a-
ion ma e s: human-cen e ed AI sys em o assis da a science code documen a ion in compu a ional
no ebooks. ACM T ans Compu -Hum In e ac 29(2):1–33. h ps:// doi. o g/ 10. 1145/ 34894 65
Wes MD (1994) Valida ing a scale o he measu emen o c edibili y: a co a iance s uc u e modeling
app oach. Jou nal Q 71(1):159–168. h ps:// doi. o g/ 10. 1177/ 10776 99094 07100 115
Wien AH, Peluso AM (2021) In luence o human e sus AI ecommende s: he oles o p oduc ype and
cogni i e p ocesses. J Bus Res 137:13–27. h ps:// doi. o g/ 10. 1016/j. jbus es. 2021. 08. 016
Wi z J, Pa e son PG, Kunz WH, G ube T, Lu VN, Paluch S, Ma ins A (2018) B a e new wo ld: se ice
obo s in he on line. J Se Manag 29(5):907–931. h ps:// doi. o g/ 10. 1108/ JOSM- 04- 2018- 0119
Wi z J, Kunz WH, Ha ley N, Ta bi J (2022) Co po a e digi al esponsibili y in se ice i ms and hei
ecosys ems. J Se Res. h ps:// doi. o g/ 10. 1177/ 10946 70522 11304 67
Wölke A, Powell TE (2018) Algo i hms in he news oom? News eade s’ pe cei ed c edibili y and selec-
ion o au oma ed jou nalism. Jou nalism. h ps:// doi. o g/ 10. 1177/ 14648 84918 757072
Yeomans M, Shah A, Mullaina han S, Kleinbe g J (2019) Making sense o ecommenda ions. J Beha
Decis Mak 32(4):403–414. h ps:// doi. o g/ 10. 1002/ bdm. 2118
Ye ushalmy J (2023) Ge man publishe Axel Sp inge says Jou nalis s could be eplaced by AI. h ps://
www. hegu a dian. com/ echn ology/ 2023/ ma / 01/ ge man- publi she - axel- sp in ge - says- jou n alis s-
could- be- epla ced- by- ai
412
1 3
Consume esponses ohuman‑AI collabo a ion a o ganiza ional…
Au ho s and A ilia ions
Ma inHaup 1,2 · JanF eidank2· Alexande Haas1
* Ma in Haup
ma in.haup @w. hm.de
Jan F eidank
jan. eidank@w. hm.de
Alexande Haas
alexande .haas@wi scha .uni-giessen.de
1 Jus us-Liebig Uni e si y, Ma ke ing andSales Managemen , Liche S asse 66, 35394Giessen,
Ge many
2 Technische Hochschule Mi elhessen, THM Business School, Wiesens asse 14, 35390Giessen,
Ge many
Zanzo o FM (2019) Viewpoin : human-in- he-loop a i icial in elligence. J A i In ell Res 64:243–252.
h ps:// doi. o g/ 10. 1613/ jai .1. 11345
Zhou L, Paul S, Demi kan H, Yuan L, Spoh e J, Zhou M, Basu J (2021) In elligence augmen a ion:
owa ds building human- machine symbio ic ela ionship. AIS T ans Hum-Compu In e ac
13(2):243–264. h ps:// doi. o g/ 10. 17705/ 1 hci. 00149
Publishe ’s No e Sp inge Na u e emains neu al wi h ega d o ju isdic ional claims in published maps
and ins i u ional a ilia ions.
413
