Sch yen, Guido; Ma one, Mau icio; Yang, Jiaqi
A icle — Published Ve sion
Explo ing he scope o gene a i e AI in li e a u e e iew
de elopmen
Elec onic Ma ke s
P o ided in Coope a ion wi h:
Sp inge Na u e
Sugges ed Ci a ion: Sch yen, Guido; Ma one, Mau icio; Yang, Jiaqi (2025) : Explo ing he scope o
gene a i e AI in li e a u e e iew de elopmen , Elec onic Ma ke s, ISSN 1422-8890, Sp inge ,
Be lin, Heidelbe g, Vol. 35, Iss. 1,
h ps://doi.o g/10.1007/s12525-025-00754-2
This Ve sion is a ailable a :
h ps://hdl.handle.ne /10419/319041
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)
Elec onic Ma ke s (2025) 35:13
h ps://doi.o g/10.1007/s12525-025-00754-2
INVITED PAPER
Explo ing hescope o gene a i e AI inli e a u e e iew de elopmen
GuidoSch yen1 · Mau icioMa one2· JiaqiYang2,3
Recei ed: 4 Ap il 2024 / Accep ed: 7 Janua y 2025
© The Au ho (s) 2025
Abs ac
A i icial in elligence (AI) has he po en ial o ans o m he way esea ch is conduc ed, pa icula ly h ough gene a i e AI
(GenAI) ools which can enhance w i en communica ion and os e inno a ion ia knowledge de elopmen . This s udy
ocuses on he la e , examining he ole o GenAI in speci ic knowledge de elopmen ac i i ies wi hin li e a u e e iews.
Th ough an epis emological lens, we dis inguish six key knowledge de elopmen ac i i ies: esea ch syn hesis, e idence
agg ega ion, c i ique, heo y building, esea ch gap iden i ica ion, and esea ch agenda de elopmen . Ou analysis demon-
s a es bo h he capabili ies and limi a ions o GenAI in suppo ing hese ac i i ies, highligh ing how GenAI can assis in
syn hesizing p e ious wo k, disco e ing and in eg a ing concep s, and ad ancing a ious knowledge domains. We emphasize
a human-cen e ed, syne gis ic app oach whe e GenAI complemen s esea che s’ e o s, a he han eplacing hem. Addi-
ionally, ou ac i i y-cen ic analysis p o ides insigh s in o how di e en ypes o li e a u e e iews can e ec i ely bene i
om GenAI suppo , he eby con ibu ing o a b oade unde s anding o AI in eg a ion in in o ma ion sys ems esea ch.
Keywo ds Gene a i e AI· Li e a u e e iews· Knowledge de elopmen · Inno a ion goal
JEL Classi ica ion M00
In oduc ion
Gene a i e a i icial in elligence (GenAI) is a highly po en
sub-ca ego y o a i icial in elligence (AI) ha has gained
conside able p ominence, la gely due o exempla s such
as Cha GPT. GenAI ope a es by le e aging deep lea ning
models o gene a e human-like con en , such as images and
wo ds, in esponse o complex and di e se linguis ic inpu s,
ins uc ions, o inqui ies (Lim e al., 2023).
GenAI ools, such as Cha GPT and Google Gemini
( o me ly known as Ba d), ha e he po en ial o enhance
schola ly wo k. Fo example, Viscon i (2021) c ea es a
machine-gene a ed li e a u e e iew o clima e, plane a y,
and e olu iona y sciences. The capabili ies o AI ools a e
apidly e ol ing, o en su passing ou p edic ions. In e ms
o academic esea ch, hey could achie e p ima y goals:
he imp o emen o w i ing (communica ion goal) and he
gene a ion o new ideas (inno a ion goal) (Dwi edi e al.,
2023). Focusing on he communica ion goal, GenAI ools
like Cha GPT can aid in p oo eading, edi ing, and e ining
he w i ing o he pape s. They complemen exis ing w i ing
ools, such as G amma ly and Spellcheck, which a e pa icu-
la ly bene icial o non-na i e English-speaking esea che s.
GenAI can imp o e language quali y and cla i y, ensu ing
ha complex ideas a e communica ed e ec i ely. Many
schola s p o ide p elimina y eedback ega ding he use o
GenAI ools o scien i ic communica ion, o en making ec-
ommenda ions and expounding bes p ac ices. Fo example,
eade s can e e o he wo ks o Bu iak e al. (2023), an
Dis e al. (2023), and Schlagwein and Willcocks (2023).
Responsible Edi o : Raine Schmid
* Guido Sch yen
[email p o ec ed]
Mau icio Ma one
mau icio.ma [email p o ec ed]
Jiaqi Yang
jiaqi.yang@hd .mq.edu.au
1 Depa men o Managemen In o ma ion Sys ems, Pade bo n
Uni e si y, Wa bu ge S . 100, 52062Aachen, Ge many
2 P esen Add ess: Macqua ie Business School, Macqua ie
Uni e si y, 4 Eas e n Road, Sydney2113, Aus alia
3 School o In o ma ion Sys ems andTechnology
Managemen , UNSW Sydney, E15 College Rd,
Kensing onNSW2033Sydney, Aus alia
Elec onic Ma ke s (2025) 35:13 13 Page 2 o 26
Al hough employing GenAI o enhance esea ch commu-
nica ion is ela i ely s aigh o wa d, le e aging i o achie e
esea ch inno a ion is complex and has gene a ed conside -
able deba e. The inno a ion goals highligh GenAI’s ole in
explo ing and gene a ing ideas, in eg a ing mul idisciplina y
pe spec i es, sol ing esea ch p oblems c ea i ely, and p o-
posing new heo e ical insigh s (Dwi edi e al., 2023). Unlike
communica ion goals, which highligh GenAI’s ole in schol-
a ly p esen a ion, inno a ion goals emphasize GenAI’s ole in
a ious knowledge ac i i ies. F om he inno a ion goals, we
de i e ou unde s anding ha knowledge de elopmen is one
impo an ype o inno a ion, as i in ol es he con inuous c e-
a ion, e inemen , and in eg a ion o exis ing and new concep s
o ad ance he knowledge domain. On one hand, GenAI holds
he po en ial o deeply engage in he knowledge de elopmen
p ocess by con ibu ing o selec ing heo e ical p oduc s, iden-
i ying ocal ideas, and es ablishing heo y-building appa a us
(Ja enpaa & Klein, 2024). I can be used o unco e insigh s
ha a e no immedia ely ob ious o esea che s, se ing as
s imuli o no el ideas and encou aging he explo a ion o new
knowledge (Benbya e al., 2024). On he o he hand, nume ous
challenges ela ed o GenAI ools in knowledge de elopmen ,
including hallucina ion, in e p e abili y, and ins i u ionaliza-
ion biases, a e well no ed (Susa la e al., 2023). These ools,
based on gene a i e ex ual engines, a e designed o ely on
wo ds and ph ases om hei aining da a, a he han on
logic, seman ic, o epis emic models. As a esul , hey ha e
been desc ibed as “s ochas ic pa o s” ha build sen ences
om da a aces (Bende e al., 2021). Mo eo e , GenAI’s
eliance on pas da a and inabili y o g asp subjec i e expe i-
ences o con ex may es ic i s abili y o de elop new ideas;
i equen ly pe pe ua es ou da ed p ac ices, which can lead o
misin o ma ion and s i le inno a ion in knowledge de elop-
men (Benbya e al., 2024). The e o e, pu suing he inno a ion
goal wi h GenAI equi es us o del e deepe in o i s sui abili y
o suppo ing speci ic knowledge ac i i ies.
We acknowledge ha knowledge ac i i ies encompass a
a ie y o esea ch pa adigms, me hods, and gen es—such as
knowledge c ea ion and gene a ion (Ala i & Leidne , 2001),
knowledge cap u e and disco e y (Paul, 2006), knowledge
in eg a ion and syn hesis (Majch zak e al., 2013), and
knowledge e inemen and e olu ion (Ramak ishnan e al.,
2023). Gi en he b ead h and complexi y o hese knowl-
edge ac i i ies, a comp ehensi e explo a ion o GenAI’s ole
ac oss all domains would be beyond he scope o a single
pape . Ins ead, his wo k ocuses on he dedica ed knowledge
de elopmen ac i i ies wi hin li e a u e e iews as a ypi-
cal elemen o (almos ) all esea ch endea o s and publica-
ions, including “ egula ” esea ch pape s and s andalone
li e a u e e iews. The pa icula impo ance o li e a u e
e iews in he con ex o knowledge de elopmen lies in he
unde s anding ha e e y li e a u e e iew gene a es some
knowledge h ough i s manda o y ac i i y o syn hesizing
p e ious wo k (Sch yen e al., 2020). This syn hesiz-
ing ac i i y in ol es no only he disco e y o knowledge
h ough a ypically s uc u ed li e a u e sea ch and e alua-
ion p ocess bu also he p ocess o desc ibing concep s and
using hem o in eg a e, ela e, con as , and o ganize he
disco e ed knowledge in a concep -cen ic manne (Sch yen,
2015; Webs e & Wa son, 2002). This p ocess equi es c ea-
i i y and human unde s anding o he disco e ed body o
knowledge, esul ing in a new knowledge con ibu ion o i s
own; i.e., e en hose li e a u e e iews ha me ely syn he-
size p io knowledge also de elop new knowledge h ough
he o mula ion o concep s and hei use o p esen disco -
e ed knowledge.
O e all, he ques ion o how o suppo knowledge de el-
opmen in li e a u e e iews wi h GenAI is ele an o mos
esea che s. F om his ocus, we de i e ou cen al ques-
ion, which we seek o answe “How can GenAI ools be
used e ec i ely o suppo speci ic knowledge de elopmen
ac i i ies in in o ma ion sys ems (IS) li e a u e e iews?”
Ou ocus is o explo e how he use o GenAI can p o ide
me hodological suppo and os e knowledge de elopmen
in li e a u e e iews (as s andalone e iews o pa s o o he
esea ch wo ks) in a human-AI collabo a ion.
To add ess ou esea ch ques ion, we ake an epis emolog-
ical pe spec i e on li e a u e e iews and d aw on a widely
adop ed se o knowledge de elopmen ac i i ies, including
syn hesizing (including disco e ing) p io esea ch, c i iciz-
ing p io esea ch, agg ega ing e idence, heo y building,
iden i ying esea ch gaps, and de eloping a esea ch agenda.
Ou goal is o e alua e he sui abili y o GenAI ools o con-
duc ing hese ac i i ies, and om his analysis, we hen o e
ecommenda ions o e ec i ely using GenAI ools. How-
e e , i should be no ed ha he epis emological pe spec i e
on GenAI is di e en om an analysis o how he inc eas-
ingly sophis ica ed echnical capabili ies o GenAI ools
can be used o p ocess (e.g., summa ize, ex ac , compa e,
consolida e, modi y) ex documen s, sp eadshee s, images,
ideos, audio, e c. While such ac i i ies e e o a p ede ined
se o inpu and can mos ly be pe o med wi hou human
in e en ion, knowledge de elopmen ac i i ies e e o he
as body o all aining da a o he GenAI LLM and equi e
some o m o human-AI collabo a ion. I should be u he
cla i ied ha ou s udy does no aim o examine how di e -
en GenAI ools espond o di e en que ies; ou s udy is
also nei he con i ma o y no explo a o y. Ins ead, we d aw
on he epis emological na u e o knowledge de elopmen
ac i i ies and he p inciples o GenAI ools, and we u ilize
examples o GenAI que ies o showcase ou ecommenda-
ions and implica ions. Th ough his, we aim o demons a e
he po en ial o GenAI ools in human-AI collabo a ion o
compiling IS LRs and o sugges s a egies o enhancing he
e iciency o he esea ch p ocess and imp o ing he quali y
o he esea ch esul s.
Elec onic Ma ke s (2025) 35:13 Page 3 o 26 13
Backg ound
Li e a u e e iews h oughanepis emological lens
The gen e o li e a u e e iew has a ac ed much in e es
in many scien i ic ields o decades, esul ing in a ious
classi ica ions o LRs. In he IS discipline, esea che s ha e
de eloped ypologies ha classi y LRs acco ding o hei
esea ch goals and me hods. Fo example, Rowe (2014) dis-
inguishes ou goals (desc ibing, explaining, unde s and-
ing, and heo y es ing). In line wi h hese goals, Pa é e al.
(2015) dis inguish nine ypes o LRs ha syn hesize p io
knowledge, agg ega e o in eg a e da a, cons uc explana-
ions, o assess ex an li e a u e c i ically. An epis emo-
logical pe spec i e o dis inguish LRs has been p oposed
by Sch yen e al. (2020). Table1 shows which knowledge
de elopmen 1 ac i i ies can be iden i ied in LRs and how
hey align wi h he abo e ypologies.
As can be seen om Table1, six knowledge de elop-
men ac i i ies a e dis inguished: syn hesizing esea ch
(SYN), agg ega ing e idence (AE), c i icizing (CRI), heo y
building (TB), iden i ying esea ch gaps (RG), and de el-
oping a esea ch agenda (RA). While he ac i i ies SYN,
AE, and CRI ocus on pas esea ch and can be conside ed
backwa d-o ien ed, he ac i i ies RG, RA, and TB poin o
u u e esea ch and a e hus o wa d-o ien ed. As we d aw
on hese ac i i ies o discuss whe he and how GenAI can
be used o suppo knowledge de elopmen in LRs, Table2
b ie ly explains he essence o hese ac i i ies.
P inciples o GenAI
GenAI b oadly e e s o a class o AI models ha p oduces
seemingly new and meaning ul con en in he o m o ex ,
images, o o he media. These models unc ion by lea ning
pa e ns om hei ex ensi e aining da ase s and gene a e
con en based on hose pa e ns (Su sala e al., 2023). No a-
ble GenAI ools include Gemini and Cha GPT as Gene a i e
Language Models, Dall-E 3, S agle Di usion and So a as
Gene a i e Image and Video Models, and Pe plexi y as an
AI-enhanced Sea ch Engine. The elease o ools like GPT-
4o and Dall-E 3 allows he gene a ion o human-like ou pu
in ex and isual o ma s wi h g ea sophis ica ion. A he
hea o hese ad ancemen s a e deep neu al ne wo ks and
ans o ma ional a chi ec u es as well as he a ailabili y o
as amoun s o aining da a, which enable hese models
o p edic and gene a e con en in ways ha closely mimic
human language, simila o an ad anced o m o au ocom-
ple e echnology (Feue iegel e al., 2024), isual a i ac s
de eloped by humans, including eal-wo ld pic u es, scien-
i ic illus a ions, a is ic pain ings, and o he media.
Despi e he imp essi e capabili ies o GenAI, signi i-
can limi a ions exis based on using, aining, and apply-
ing LLMs and he esul ing way in which in o ma ion
is p ocessed and ou pu is gene a ed. While GenAI can
p o ide in o ma ion, summa ize ex ensi e ma e ial, and
gene a e cohe en ex , i does no “unde s and” he ma e-
ial in he human sense. Fo ins ance, AI can summa ize
esea ch indings o explain concep s as desc ibed in he
p omp s, bu i does no inhe en ly g asp he unde lying
p inciples, con ex , o impo ance o hese concep s beyond
hei ex ual ep esen a ion. In ligh o his, schola s deba e
whe he GenAI ools genuinely “unde s and” hei ou pu s
(Mi chell e al., 2023). AI sys ems can ecognize pa e ns
and ep oduce da a based on p obabili ies, bu hey lack
he abili y o con ex ualize knowledge wi hin b oade
philosophical o heo e ical amewo ks unless explici ly
ou lined in he da a hey p ocess. GenAI does no possess
he human-like abili y o c ea i ely heo ize o specula e
in a way ha e lec s deep unde s anding and inno a i e
hough . Also, o pa icula impo ance is he inabili y o
GenAI o include o simula e “human in ui ion,” which
goes beyond accessing and ep oducing da a. I lacks he
abili y o ques ion deeply, hink c i ically, and engage
wi h ma e ial in a way ha challenges and ex ends exis -
ing knowledge bounda ies. Human in ui ion and expe ien-
ial lea ning play c ucial oles in hese p ocesses, allowing
schola s o disce n/disen angle sub le ies and implica ions
ha migh no be e iden h ough AI. As a consequence
o he desc ibed GenAI-inhe en cha ac e is ics, we s ill
need esea che s “ o know” and a collabo a i e human AI
wo king model.
One esul ing majo issue o he a o emen ioned GenAI
p inciples is he p opensi y o hese models o p oduce
inco ec o misleading esul s (o en called “AI hallucina-
ions”), whe e gene a ed con en appea s plausible bu is
ac ually inco ec o nonsensical (Hicks e al., 2024). This
p oblem is oo ed in he p obabilis ic na u e o hese models,
which gene a e he mos likely esponse o a p omp , a he
han e i ying i s u h ulness (Feue iegel e al., 2024).
Addi ionally, GenAI models equen ly exhibi biases e lec-
i e o he human-gene a ed da a hey a e ained on, pe -
pe ua ing s e eo ypes and p ejudices p esen in he aining
da a (Bail, 2024). Copy igh iola ions also pose a signi i-
can limi a ion, as GenAI models can p oduce ou pu s ha
esemble exis ing wo ks wi hou pe mission o a ibu ion
o he o iginal c ea o s (Feue iegel e al., 2024). Add ess-
ing hese limi a ions equi es ongoing esea ch o imp o e
model anspa ency, bias mi iga ion, and he de elopmen o
e hical guidelines o AI deploymen .
1 While Sch yen e al. (2020) use he e m “knowledge-building
ac i i ies,” we p e e o use he e m “knowledge de elopmen ac i i-
ies” o emain e minologically consis en wi h he common e m
“knowledge de elopmen .”
Elec onic Ma ke s (2025) 35:13 13 Page 4 o 26
Table 1 De i a ion o knowledge de elopmen ac i i ies o li e a u e e iews om me hodological pape s (Sch yen e al., 2020, p. 137)
a Goals based on Rowe (2014)
b Based on Pa é e al. (2015), who dis inguish and illus a e he e iew ypes based on nine dimensions
c SYN syn hesizing, AE agg ega ing e idence, CRI c i icizing, TB heo y building, RG iden i ying esea ch gaps, RA de eloping a esea ch agenda
GoalsaRe iew ypesbKnowledge de elopmen ac i i yc
Backwa d-o ien ed Fo wa d-o ien ed
Desc ibing Na a i e Re iew
Le y and Ellis (2006), Ha (2009)
• Na a i e summa y o p io indings on a opic (SYN) • Iden i ica ion o esea ch gaps (RG)
• De elopmen o an agenda o esea ch and p ac ice
(RA)
Desc ip i e Re iew
King and He (2005)
• Quan i a i e and na a i e summa y o wha we know
abou a opic (SYN)
• Iden i ica ion o ends o e ime (SYN)
• De elopmen o ecommenda ions o in luence he
de elopmen o a opic, domain, o me hod (RA)
Scoping Re iew
A ksey and O’Malley (2005), Le ac e al. (2010)
• Na a i e summa y o he size and na u e o ex an
li e a u e (SYN)
• Iden i ica ion o esea ch gaps (RG)
• De elopmen o a esea ch agenda wi h po en ial impli-
ca ions o esea ch and p ac ice (RA)
Unde s anding C i ical Re iew
Rowe (2014), Al esson and Sandbe g (2011)
• Summa ize pas knowledge on a domain o in e es
(SYN)
• C i ical accoun o ex an li e a u e, e ealing weak-
nesses, o inconsis encies (CRI)
• P o iding a ocus o a new di ec ion o s udies (RA)
Explaining Theo e ical Re iew
Ri a d (2014), Rowe (2014), To aco (2005), Walke
and A an (2011), Webs e and Wa son (2002)
• Syn hesis o p io li e a u e (SYN) • Theo y de i a ion: De elopmen o a heo y om he
explana ions in ano he ield (TB)
• Theo y syn hesis: De elopmen o a heo y om pulling
oge he p io e idence abou a phenomenon (TB)
• Theo y analysis: Examina ion o a heo y and iden i ica-
ion o he need o addi ional e inemen (TB)
• De eloping a esea ch agenda (RA)
Realis Re iew
Pawson e al. (2005)
• Syn hesis o e idence and dissemina ion o indings
(SYN)
• De elopmen o a heo y aimed a explaining wha i
is abou an in e en ion ha wo ks, o whom, in wha
ci cums ances and why (TB)
Theo y es ing Me a-analysis
King and He (2005), Rosen hal and DiMa eo (2001),
Ca d (2011)
• In eg a ion o knowledge gained in empi ical s udies
(SYN)
• S a is ical agg ega ion o empi ical indings (AE)
• Explo a ion o mode a o s can p o ide o wa d-looking
ideas o u u e esea ch (RA)
Quali a i e Sys ema ic Re iew
Gough e al. (2012), Pe ic ew and Robe s (2008)
• Syn hesis o e idence (SYN)
• Na a i e agg ega ion o possibly he e ogeneous
empi ical indings (AE)
• De elopmen o implica ions o policy, p ac ice, and
u he esea ch (RG)
Umb ella Re iew
Thomson e al. (2010)
• Syn hesis o he indings om p io e iews (SYN)
• Na a i e and/o s a is ical agg ega ion o p io e iew
indings (AE)
• Iden i ica ion o a eas whe e mo e esea ch is needed
(RG)
Elec onic Ma ke s (2025) 35:13 Page 5 o 26 13
GenAI in he esea ch p ocess
The po en ial o GenAI o e olu ionize academic esea ch
ex ends beyond i s capaci y o enhance academic w i ing,
such as wi h ools like G amma ly. Despi e i s inhe en limi-
a ions in de eloping knowledge, i p esen s a complex se o
oppo uni ies and challenges, al e ing he esea ch landscape
ega ding how knowledge is c ea ed, sha ed, and consumed
(Benbya e al., 2024), oge he wi h he e ol ing ole o
au ho s, e iewe s, and edi o s (Yoo, 2024). On one hand,
GenAI p omises o enhance he e iciency o knowledge syn-
hesis, democ a ize access o expe ise, and s eamline he
pee e iew p ocess, hus po en ially expedi ing he knowl-
edge disco e y p ocess and mi iga ing he ep oducibili y
c isis (Ala i e al., 2024). I also o e s he p ospec o aug-
men ing human capabili ies in gene a ing explici knowledge
om aci unde s anding and p o iding ailo ed coaching,
hus acili a ing “long jumps” in he knowledge explo a-
ion p ocess (Schwa z & Te’eni, 2024; Ala i e al., 2024;
Yoo, 2024). On he o he hand, his echnological leap is no
wi hou i s pi alls. GenAI in oduces isks such as biases,
e hical conce ns, and he po en ial o hallucina ion, which
could comp omise he quali y, anspa ency, and explain-
abili y o esea ch ou comes (e.g., Else, 2023; Kankanhalli,
2024; Lund e al., 2023; Ngwenyama & Rowe, 2024). Mo e-
o e , he e is a looming ea o homogeniza ion in esea ch,
unde mining inno a ion, and impac ing he no ms o scien-
i ic discou se (Ngwenyama & Rowe, 2024; Webe , 2024).
The limi a ions o GenAI in e ec i ely iden i ying gaps in
(in e )disciplina y knowledge and con o ming o scien i ic
no ms u he unde sco e he need o mo e esea ch o mi i-
ga e hese challenges. Thus, while GenAI b ings o h new
Table 2 Knowledge de elopmen ac i i ies in LRs (based on Sch yen e al. (2020), pp. 138 )
Knowledge de elopmen ac i i y Key cha ac e is ics
Syn hesizing esea ch (SYN) • Manda o y ac i i y in any LR
• Summa izes and o ganizes published knowledge, es ablishes o de in p e ious esea ch, and makes ans-
pa en how esea ch con ibu ions ela e o each o he
• Follows a sys ema ic app oach and p o ides anspa ency ega ding he s a e and p og ess o domain
knowledge
• May ake se e al o ms and in ol e a ying deg ees o in e p e a ion
• May begin by cla i ying undamen al aspec s such as de ini ions, domain- ele an a iables, ela ionships
be ween concep s, and domain ocabula y
• May e eal cen al hemes and esea ch s eams
Agg ega ing e idence (AE) • Takes heo e ical models as a ame, ga he s empi ical s udies, ex ac s he e idence, and pe o ms
s a is ical agg ega ion (e.g., me a-analysis o o e coun ing) o e alua e he deg ee o which he e idence
suppo s exis ing heo e ical models
• Focuses on agg ega ing e ec sizes in ela i ely homogeneous models and migh include quali ica ions in
he o m o mode a o analyses
• Applicable when enough empi ical esea ch has accumula ed
• Me a-analyses a e he mos common ype o e iew ha agg ega es empi ical e idence
C i icizing (CRI) • Shows ha knowledge ela ed o a p oblem is in some ways inadequa e and p e en s a domain om
p og essing
• May occu in di e en o ms by p oblema izing assump ions o iden i ying me hodological, logical, o
concep ual p oblems
• In con as o cumula i e ex ensions o exis ing knowledge, c i icism sugges s a e olu iona y pa h ha is
likely o be i econcilable wi h exis ing knowledge
Theo y building (TB) • P o ides p o isional, possibly conjec u al knowledge in he o m o new hypo heses and heo e ical mod-
els ha need o be es ed by subsequen esea ch
• Encompasses de eloping new heo ies, and e ining o syn hesizing heo ies
Iden i ying esea ch gaps (RG) • Desc ibes a misma ch be ween knowledge ha is p o ided by ex an esea ch and knowledge ha is
equi ed o expec ed
• Is expec ed o s imula e o he au ho s by subs an ia ing a need o esea ch and mo i a ing esea che s o
close he gaps
• Co esponds o he p ocess o spo ing gaps in he exis ing body o knowledge
De eloping a esea ch agenda (RA) • Elabo a es on how u u e esea ch should be conduc ed o achie e meaning ul p og ess, possibly sugges -
ing speci ic esea ch designs, empi ical se ings, o o e ing s a egic ecommenda ions
• Ac i i y is con ingen on he iden i ica ion o esea ch gaps o a c i ique o p io esea ch
• De elops a ision on behal o he au ho s ha is o ien ed owa ds a p omising esea ch goal and a co -
esponding cha o u he esea ch
• Should make speci ic and ac ionable ecommenda ions ha can e en ake he o m o a de ailed deploy-
men plan, which could include speci ic esea ch p oposi ions, sugges ions on esea ch designs, and
empi ical me hods
Elec onic Ma ke s (2025) 35:13 13 Page 6 o 26
a enues o ad ancing academic esea ch, i also necessi a es
a cau ious app oach o add ess i s inhe en isks and ensu e
ha i suppo s, a he han unde mines, epis emic alues.
Among hose eme ging issues, he use o GenAI o
conduc ing LRs has ga ne ed pa icula a en ion among
schola s (e.g., Dasbo ough, 2023; Dwi edi e al., 2023; Pan
e al., 2023). GenAI helps wi h li e a u e e iews by p o-
cessing di e se uns uc u ed and s uc u ed da a o unco e
hidden pa e ns, ela ionships, and insigh s wi hin scien i ic
li e a u e (Ala i, 2024). I can e ie e mains eam o domi-
nan iews om exis ing li e a u e, allowing esea che s
o e iew, c i icize, alida e, and ex end he baseline wi h
hei hough expe imen s (Ngwenyama & Rowe, 2024).
Resea che s ha e e ec i ely used GenAI o suppo a ange
o LR ac i i ies, such as gene a ing e e ences, analyzing
li e a u e, d a ing pape s, unde s anding di e en pe spec-
i es, and p o iding a ounda ion o heo izing (Ja enpaa
& Klein, 2024). The cu en applica ion o GenAI in LRs
shows some ini ial insigh s. Fo example, GenAI has p o en
i s po en ial o gene a e e ec i e Boolean que ies o sys em-
a ic li e a u e sea ches, whe e i is able o ollow complex
ins uc ions and gene a e que ies wi h high p ecision (Wang
e al., 2023). Ne e heless, in ano he esea ch con ex , using
GenAI o li e a u e sea ches did no gene a e ideal esul s.
Gwon e al. (2024) compa ed he pe o mance o Cha GPT
and Mic oso Bing AI in conduc ing li e a u e sea ches on
Pey onie’s disease. Thei indings showed ha ou o 1287
s udies iden i ied by Cha GPT, only 7 (0.5%) we e di ec ly
ele an . In con as , Bing AI iden i ied 48 s udies, o which
19 (40%) we e ele an , app oaching he human benchma k
o 24 ele an s udies. The inconsis ency in indings high-
ligh s he a ying pe o mance o GenAI ools in execu ing
li e a u e sea ches.
Beyond li e a u e sea ch, GenAI can also assis in elemen
mapping and coding o ele an publica ions; he gene a i e
aspec o GenAI allows he esea che o econcep ualize he
elemen maps based on hei expe ise and insigh s ga he ed
du ing he p ocess, ollowed by he o mula ion o discus-
sion and conclusion (Pan e al., 2023). Fo hese challenging
ac i i ies, he eliabili y and consis ency o GenAI ha e been
ound o be s ill on pa wi h hose o human esea che s
(Jenko e al., 2024; Maniaci e al., 2024). Rega ding e-
quen conce n o e ab ica ion (in o ma ion ha is plausible-
sounding bu no ac ually accu a e) and e o s gene a ed by
GenAI in LRs, esea ch shows a g ea imp o emen om
GPT-3.5 o GPT-4; ab ica ion was ound in 55% o GPT-
3.5 ci a ions bu jus 18% o hose in GPT-4 (Wal e s &
Wilde , 2023). Emb acing ecen ad ancemen s in GenAI,
s udies e alua e he capabili y o bo h human esea che s
and GenAI o delinea e he socio- echnical equi emen s o
using GenAI in LRs. These equi emen s include a oiding
he backwa d na u e o da a collec ion, ensu ing anspa -
ency o pa ame e s and model weigh s, acili a ing i e a i e
dialogue be ween GenAI and esea che s, selec ing GenAI
ools ha allow o c i ical in e oga ion o da a, and main-
aining awa eness o GenAI’s in luence on he esea ch p o-
cess (Ngwenyama & Rowe, 2024).
Some ea ly a emp s ha e been made o empi ically e al-
ua e he p ac ical u ili y o GenAI ools in li e a u e e iews:
Fo ins ance, Si e al. (2024) conduc ed a la ge-scale human
s udy wi h o e 100 NLP esea che s o assess whe he la ge
language models can gene a e no el esea ch ideas, inding
ha LLM-gene a ed ideas we e judged as mo e no el han
hose om human expe s, albei sligh ly weake in easibil-
i y. Simila ly, de la To e-López e al. (2024) p esen ed a
su ey o AI echniques p oposed o e he pas 15yea s o
assis esea che s in conduc ing sys ema ic analyses o scien-
i ic li e a u e, p o iding a his o ical pe spec i e on he e o-
lu ion o AI in li e a u e e iews. Addi ionally, Gwon e al.
(2024) e alua ed he pe o mance o Cha GPT and Mic o-
so Bing AI in conduc ing li e a u e sea ches o sys ema ic
e iews, sugges ing ha while hese gene a i e AI ools hold
p omise, hey a e no ye su icien ly accu a e o easible o
eal- ime e idence gene a ion in medical esea ch. These
ea ly a emp s ha e p o ided ini ial empi ical e idence on
he e ec i eness o using GenAI in li e a u e e iews. How-
e e , and in con as o esea ch on how non-GenAI can sup-
po LRs (e.g., an Din e e al., 2021; Wagne e al., 2022),
he li e a u e is s ill silen on how GenAI can suppo epis e-
mological ac i i ies when compiling a li e a u e e iew. To
add ess his gap, ou s udy p oposes an insigh ul pe spec-
i e guiding he epis emological use o GenAI in LRs while
emphasizing bes p ac ices o human-AI collabo a ion.
Suppo ing knowledge de elopmen
ac i i ies wi hgene a i e AI ools
App oaching ou esea ch goal o e ealing he po en ial o
GenAI ools o compiling IS LRs and o de elop ecom-
menda ions, we p oceed by using (a) he sample domain
o “IS business alue”; (b) se e al GenAI ools, including
Cha GPT (model GPT-4), Pe plexi y (model GPT-3), Bing
AI (now Mic oso Copilo ) (model GPT-4), and Google
Gemini (model 1.0 p o); and (c) sample que ies o illus-
a e ou ecommenda ions. Howe e , ou ecommenda ions
a e no speci ic o his pa icula domain o he selec ed AI
ools, models, o que ies used. Ra he , hey a e based on
and d i en by he undamen al na u e o di e en knowledge
de elopmen ac i i ies and he gene al pa adigms unde lying
la ge language model-based GenAI ools.
We de ail and p o ide examples o how GenAI ools can
suppo esea che s in each o he men ioned knowledge
de elopmen ac i i ies in he sense o human-AI collabo a-
ion. We en e ed nume ous p omp s in o he GenAI ools and
p esen selec ed examples o he ou pu s p oduced o show
Elec onic Ma ke s (2025) 35:13 Page 7 o 26 13
hei po en ial. Howe e , no all p omp s e u ned p omis-
ing esul s: we show an example o he limi ed abili y o
cu en GenAI ools o suppo a ious ypes o e iews,
as discussed in he “Implica ions o e iew ypes” sec ion.
Syn hesizing
Syn hesizing esea ch (SYN) in ol es iden i ying schola ly
wo k and summa izing, compa ing, and con as ing i , ide-
ally, in a concep -cen ic way (Webs e & Wa son, 2002).
The iden i ica ion o li e a u e is usually done by que ying
li e a u e da abases, scanning ables o con en s, e c.; o
a me hodology, see, o example, he u o ial by Sch yen
(2015). Complemen ing and going beyond hese asks,
GenAI may be used o iden i y li e a u e ha adop s a spe-
ci ic pe spec i e on a opic o in es iga ion, be i om a
pa icula heo e ical o epis emological pe spec i e. This
app oach allows he iden i ica ion o li e a u e om a spe-
ci ic pe spec i e and he o ganiza ion o i s p esen a ion.
I also os e s he adop ion o a mul i- iew pe spec i e on
a opic. Figu e1 shows a Cha GPT que y and answe wi h
which IS business alue is iewed om he pe spec i e o
he IS success model o DeLone and McLean (1992).
I should be no ed ha Cha GPT pa ially hallucina es
because “se ice quali y” and “ne bene i s” do no belong
o he ca ego ies o IS success men ioned by DeLone and
McLean (1992) and should be eplaced by he ca ego ies
“indi idual impac ” and “o ganiza ional impac .” While,
unsu p isingly, a ho ough unde s anding o he o iginal
model equi es eading he a icle o DeLone and Mclean,
he answe is use ul o syn hesizing IS business alue in
se e al o he ways. Fi s , i quickly p o ides a ough unde -
s anding o he na u e o he sugges ed IS success model.
Second, he inad e en ly included ca ego ies o “se ice
quali y” and “ne bene i s” p o ide s a ing poin s o a li -
e a u e sea ch on he ole o hese concep s in IS business
alue. Fo example, “se ice quali y” (as a ed by cus ome s)
was ound o ha e a posi i e e ec on he “in ended use”
o ope a ional CRM echnology (Hsieh e al., 2011), which
highligh s a ela ionship be ween he concep o “se ice
quali y” and he concep o “in ended use,” which is ela ed
o he ca ego y o “use” included in he DeLone and McLean
model. Thi d, esea che s can now di e deepe in o he li -
e a u e o u he elabo a e how IS has con ibu ed o a i-
ous elemen s o he IS success model. Succeeding esea ch
s eps should in ol e in es iga ing he p o ided e e ences
and de eloping app op ia e ollow-up que ies.
In o de o adop a complemen a y pe spec i e on IS
business alue, a que y may look a his opic om he pe -
spec i e o he p ocess model sugges ed by Soh and Ma kus
(1995); Fig.2 shows an example o such a que y and he
Cha GPT answe .
Beyond p o iding e e ences, he answe p o ides an ini-
ial o e iew o he key di e ences be ween he wo pe spec-
i es and includes ideas o how he wo pe spec i es may
complemen each o he and wha a e he sha ed insigh s.
Fo example, bo h pe spec i es ecognize he impo ance o
conside ing he impac o IS on o ganiza ional pe o mance,
which is a mul idimensional cons uc . Thus, u he li e a-
u e analysis can di e deepe in o his concep by analyz-
ing wha he IS business alue li e a u e, including he wo
models men ioned abo e, has ound on he impac o IS on
di e en dimensions o o ganiza ional pe o mance.
As a syn hesis migh also include cla i ying undamen al
aspec s, such as de ini ions and ela ionships be ween con-
cep s, a que y may be an en y poin o a body o concep -
de ining li e a u e (see Fig.3). Howe e , i should be no ed
ha one o he sho comings o GenAI is ha i s aining
da a may no be up o da e, bu , i i was able o connec o an
up- o-da e da abase o schola a icles, esul s may imp o e.
Agg ega ing e idence
The agg ega ion o e idence (AE) in ol es analyzing
quan i a i e da a by means o quan i a i e o quali a i e
app oaches. A he quali a i e le el, e idence agg ega ion
in ol es a na a i e in e p e a ion o quan i a i e da a.
Then, GenAI ools may be used in simila ways as when
syn hesizing esea ch. In con as , a he quan i a i e le el,
agg ega ing e idence usually includes he s a is ical agg e-
ga ion o empi ical s udies, such as me a-analysis o o e
coun ing, and in ol es ga he ing exis ing s udies, app ais-
ing he quali y o e idence, de e mining agg ega ed e ec
sizes, and es ing hei signi icance (Sch yen e al., 2020).
The na u e o hese asks equi es any suppo ing GenAI
ools o include s a is ical me hodologies. In con as o non-
gene a i e AI ools (see, o example, Wagne e al. (2022)),
pu ely ex -gene a i e AI ools a e no capable o os e ing
s udies ha agg ega e e idence. Howe e , we en ision he
de elopmen o GenAI ools ha gene a e ex based on he
s a is ical analysis o a se o empi ical s udies.
C i icizing
C i icizing (CRI) e eals ha knowledge ela ed o a p oblem
p e en s a domain om p og essing. I can be implemen ed
by, o example, p oblema izing assump ions o iden i ying
me hodological, logical, o concep ual p oblems. Con a y
o wo k ha cumula i ely ex ends exis ing knowledge, c i i-
cism sugges s a e olu iona y pa h ha is no likely o be
econciled wi h exis ing knowledge (Sch yen, 2015). The
dis up i e cha ac e o c i icizing p io esea ch makes i
challenging o exploi GenAI ools o suppo his ype o
knowledge c ea ion, as hey ely on his o ical aining da a
and, hus, can be expec ed o p o ide esul s ha os e
Elec onic Ma ke s (2025) 35:13 13 Page 8 o 26
Fig. 1 Cha GPT que y: he
pe spec i e o he model o
DeLone and McLean (1992)
Elec onic Ma ke s (2025) 35:13 Page 15 o 26 13
use hem e ec i ely in he cou se o knowledge de elopmen
human-AI collabo a ion.
Using GenAI ools does no elease esea che s om he
need o hink c i ically and show c ea i i y. Wi h ega d
o he o me equi emen , i holds ha , o all ac i i ies,
i should be aken o g an ed ha he use o GenAI ools
equi es cau ion on he pa o esea che s, as esul s may
include laws and misleading in o ma ion, and possibly non-
exis en e e ences. As wi h any o he ool ha suppo s aca-
demic esea ch, esul s mus no be conside ed “p oduc s”
ha a e eady o use in scien i ic wo k. The la e equi e-
men includes he challenge o esea che s o de elop a
se ies o consecu i e que ies o GenAI ools and o adop
an i e a i e app oach in o de o de i e p omising esul s.
I should be no ed ha he knowledge de elopmen ac i i-
ies conside ed, which may bene i om using GenAI ools,
include bo h backwa d-o ien ed (syn hesizing, agg ega ing
e idence) and o wa d-o ien ed knowledge de elopmen
ac i i ies (c i icizing, heo y building, iden i ying esea ch
gaps, de eloping a esea ch agenda) (Sch yen e al., 2020).
While i seems ha dly su p ising ha GenAI ools can sup-
po backwa d-o ien ed knowledge de elopmen ac i i ies,
i may ha e been conside ed less ob ious ha hey can also
os e o wa d-o ien ed knowledge de elopmen ac i i ies.
Implica ions o e iew ypes
Ha ing analyzed he po en ial uses o GenAI ools o s an-
dalone LRs a he le el o knowledge de elopmen ac i i-
ies, we p oceed wi h de i ing implica ions o a ious
ypes o LRs in he IS ield (Pa é e al., 2015), which can
be pe cei ed, om an epis emological pe spec i e, as bun-
dles o knowledge de elopmen ac i i ies (Sch yen e al.,
2020); see Table1 o an o e iew o he e iew ypes. We
wish o emphasize ha ou ocus a he le el o e iew ypes
is examining ways o de i ing insigh s in o he ex en o
which key ac i i ies o speci ic e iews can be acili a ed
h ough human-AI collabo a ion. We do no aim o make
p esc ip i e sugges ions such as “c ea e a e iew o ype X
on opic Y,” as ou expe imen s, simila o hose o Susa la
e al. (2023), showed discou aging esul s. We p o ide wo
nega i e examples below when discussing di e en ypes o
li e a u e e iews.
Na a i e, desc ip i e, and scoping e iews aim o
desc ibe phenomena and belong o he e iew g oup ha
p ima ily summa izes p io knowledge and adop s a b oad
scope o ques ions. Na a i e e iews a e selec i e, as hey
do no in ol e a sys ema ic and comp ehensi e li e a u e
sea ch. These e iews p o ide a na a i e summa y o he
Fig. 7 Bing AI que y no. 1:
(missing) knowledge o IS busi-
ness alue c ea ion
Elec onic Ma ke s (2025) 35:13 13 Page 16 o 26
li e a u e and o en con ibu e o iden i ying esea ch gaps
and de eloping a esea ch agenda. Figu e12 shows a que y
ha asks Cha GPT 4.0 o p o ide a comple e (na a i e)
li e a u e e iew on selec ed IS business alue opics. Spe-
ci ically, he GenAI ool is ins uc ed o p oduce a li e a u e
e iew ha iden i ies he ypes o alue mos ex ensi ely
discussed in he li e a u e.
While he alue ypes in his example ha e been dis-
cussed in he IS business alue li e a u e and indeed ep-
esen impo an a eas o esea ch, he e iew is silen on
a la ge body o esea ch on one o he mos ex ensi ely
s udied alue ypes: i m pe o mance (ma ke ing pe o -
mance, accoun ing pe o mance) (Sch yen, 2013). O e all,
he e iew ails o p o ide a leas a b ie o e iew o he
selec ed subse o he mos ex ensi ely s udied alue ypes as
que ied. This exempli ies ha que ies o GenAI ools should
no be expec ed o gene a e a comp ehensi e e iew.
In con as o na a i e e iews, desc ip i e e iews pu -
sue a ep esen a i e sea ch s a egy. They analyze he ex en
o which a body o empi ical s udies in a speci ic esea ch
a ea suppo s o e eals in e p e able pa e ns o ends.
Beyond summa izing wha is known abou a opic, hey usu-
ally also de elop ecommenda ions o in luence he de elop-
men o a opic, domain, o me hod. Scoping e iews adop
a comp ehensi e sea ch s a egy and examine he ex en ,
ange, and na u e o esea ch ac i i ies. They usually also
iden i y esea ch gaps in he ex an li e a u e and de elop a
esea ch agenda. All hese ypes o e iews can bene i om
GenAI ools in o ganizing li e a u e syn hesis, such as cla -
i ying de ini ions and ela ionships be ween concep s and
adop ing a mul i- iew pe spec i e. Addi ionally, GenAI can
aid in iden i ying esea ch gaps and o mula ing a esea ch
agenda h ough a se ies o que ies, making i sui able o
aiding na a i e e iews wi h na owe ocuses.
C i ical e iews pu sue he o e a ching goal o unde -
s anding phenomena and aim o summa ize pas knowl-
edge and c i ically analyze he ex an li e a u e on a b oad
opic o e eal weaknesses, con adic ions, con o e sies,
o inconsis encies. They o en p o ide a new di ec ion o
s udies. Due o hei ocus on c i icizing p io esea ch and
hei dis up i e na u e, he bene i o GenAI ools o such
e iews la gely depends on he “c ea i i y” o esea che s o
use hese ools o c i icize p io conclusions (see he “C i i-
cizing” sec ion).
Theo e ical e iews and ealis e iews ocus on explain-
ing phenomena. Theo y building can occu in di e -
en o ms, including heo y de i a ion, heo y syn hesis,
and heo y analysis. Realis e iews a e heo y-d i en
Fig. 8 Bing AI que y no. 2:
(missing) knowledge o IS busi-
ness alue c ea ion
Elec onic Ma ke s (2025) 35:13 Page 17 o 26 13
Fig. 9 Google Gemini que y no.
1: iden i ying ypes o o ganiza-
ions’ success in using OSN
Fig. 10 Google Gemini que y
no. 2: measu ing he ROI o
inc eased b and awa eness and
isibili y
Elec onic Ma ke s (2025) 35:13 13 Page 18 o 26
in e p e a i e e iews; hey syn hesize e idence and dis-
semina ion o indings. GenAI has he po en ial o assis
esea che s in heo e ical and ealis e iews by enhancing
he p ocess o heo y e inemen and de elopmen . Th ough
i s language gene a ion and analysis capabili ies, GenAI can
explo e and syn hesize di e se sou ces o e idence, acili-
a ing he iden i ica ion o key pa e ns, ela ionships, and
explana o y ac o s. By le e aging GenAI, esea che s can
e icien ly analyze la ge olumes o li e a u e and ex ac
insigh s ha con ibu e o he o mula ion and e inemen o
heo ies, pa icula ly in ealis e iews whe e he goal is o
unco e wha in e en ions a e e ec i e, o whom, unde
wha condi ions, and why.
The inal g oup o e iews, which sha e he o e all goal o
da a agg ega ion and in eg a ion, consis s o me a-analysis,
quali a i e sys ema ic e iews, and umb ella e iews. They
ocus on a na ow se o ques ions. Me a-analysis ocuses
on he s a is ical agg ega ion o e idence. The cu en
Fig. 11 Google Gemini que y
no. 3: measu ing he ROI o
e ec i e ma ke esea ch and
insigh s
Elec onic Ma ke s (2025) 35:13 Page 19 o 26 13
Table 3 Suppo ing knowledge de elopmen ac i i ies in LRs wi h GenAI ools
Knowledge de elopmen ac i i y GenAI capabili ies GenAI limi a ions Recommenda ions o esea che s
Syn hesizing esea ch • Iden i y di e se li e a u e sou ces o encou age a
mul i ace ed iew on a opic, including he adop-
ion o a ying heo e ical, epis emological, o
me hodological pe spec i es
• Sea ch li e a u e, complemen ing adi ional
manual li e a u e sea ch p ocedu es h ough
da abases
• Concep -cen ic syn hesis h ough summa iza-
ion, compa ison, and con as a ion o iden i ied
li e a u e in a concep -cen ic way, p o iding
a ough unde s anding o complex models and
amewo ks
• Hallucina ions, in pa icula inco ec concep s
ha do no belong o es ablished models
• Ou da ed aining da a limi he accu acy and
ele ance o in o ma ion and e e ences p o ided
• O e emphasize mo e known li e a u e, o e look-
ing less popula bu equally impo an sou ces,
po en ially leading o a biased unde s anding o
he opic
• Resea che s need o c oss- e i y GenAI-gene a ed
in o ma ion wi h o iginal schola ly sou ces o
ensu e accu acy
• Use GenAI in conjunc ion wi h up- o-da e da a-
bases o schola ly a icles
• C i ically analyze and e ine he insigh s p o ided
by GenAI, le e aging hei own knowledge, expe-
ience, and, some imes, in ui ion
• De elop app op ia e ollow-up que ies o u he
in es iga e he p o ided e e ences and deepen he
unde s anding o he iden i ied li e a u e
Agg ega ing e idence • Pe o m quali a i e analysis, such as na a i e
in e p e a ion o empi ical s udies based on
s a is ical analysis esul s
• In quan i a i e analysis, GenAI lacks s a is i-
cal me hodologies o app op ia e quan i a i e
agg ega ion o empi ical e idence, such as
equi ed o me a-analysis
• Resea che s may in eg a e GenAI wi h s a is ical
ools and consul wi h GenAI ools o unde s and
and complemen he esul s ob ained h ough
quan i a i e analysis
• Resea che s need o c i ically e alua e and
alida e he na a i e in e p e a ions gene a ed by
GenAI ools agains he ac ual s a is ical da a and
analyses
C i icizing • Iden i y and summa ize c i icisms p e iously
o mula ed by o he esea che s
• Iden i y con adic ions by analyzing a bunch o
li e a u e and highligh ing he con adic ions in
he indings and claims
• GenAI does no ha e he capabili y o o mula e
“i s own” c i ique on exis ing li e a u e, which
in ol es ques ioning whe he “ hings we e done
igh ”
• Resea che s should p o ide ele an li e a u e
sou ces and o mula e que ies o GenAI o u he
elabo a e on al eady iden i ied esea ch issues
• A e GenAI iden i ies he con adic ions in
li e a u e, esea che s need o conduc high-le el
p oblema iza ion o see he oo causes o hem,
such as challenging he unde lying assump ions o
exis ing s udies
Theo y building • Suppo heo y de elopmen by p o iding ope a-
ionaliza ions and ex ensions o b oad ca ego ies,
helping esea che s de elop axonomies
• Iden i y causal chains p o ided in mul idiscipli-
na y li e a u e, suppo ing he de elopmen o
explana o y heo ies by in eg a ing ideas om
a ious disciplines
• P o ide ini ial insigh s and e e ences ha
esea che s can use as s a ing poin s o deepe
in es iga ion in o speci ic i ems o concep s
• The causal ela ionship p oposed by GenAI may
be supe icial and lack su icien easoning
• GenAI is no capable o gene a ing en i ely new
heo e ical models o hypo heses wi hou human
in e en ion and c ea i i y
• Wi h he ini ial insigh s p o ided by GenAI,
esea che s should le e age hei own abs ac
hinking, c ea i i y, and in ui ion o heo ize he
phenomenon o in e es
• Resea che s should complemen he ini ial
insigh s p o ided by GenAI wi h ho ough e iews
o ela ed empi ical s udies o de elop well-
ounded and inno a i e hypo heses
• Resea che s should employ an i e a i e p ocess o
que ying GenAI ools and analyzing he esponses
o e ine and deepen he unde s anding o speci ic
heo e ical cons uc s and ela ionships
Elec onic Ma ke s (2025) 35:13 13 Page 20 o 26
gene a ion o GenAI ools is unable o suppo such asks.
Quali a i e sys ema ic e iews a emp o sea ch, iden i y,
selec , app aise, and abs ac da a om quan i a i e empi i-
cal s udies. While employing he ypical sys ema ic e iew
p ocess, hey use na a i e and mo e subjec i e ( a he han
s a is ical) me hods (Pa é e al., 2015). Due o i s na ow se
o esea ch ques ions, his ype o e iew may pa icula ly
bene i om ac i i ies ha suppo syn hesizing esea ch and
iden i ying esea ch gaps. Umb ella e iews, also e e ed
o as o e iew o sys ema ic e iews, sys ema ic e iew
o sys ema ic e iews, and me a- e iews, in ol e a ious
ac i i ies o syn hesizing p io esea ch, agg ega ing e i-
dence, and iden i ying esea ch gaps unde a na owe se
o esea ch ques ions. The bene i s o using GenAI ools o
umb ella e iews a e simila o hose o quali a i e sys em-
a ic e iews.
As s a ed abo e, que ies o GenAI ools a e no in ended
o gene a e comple e li e a u e e iews. A u he example o
an unsuccess ul eques o a e iew can be seen in Fig.13,
which shows a eques o an “umb ella e iew” o how he
use o a i icial in elligence in o ganiza ions has posi i ely
o nega i ely a ec ed he o ganiza ion’s business pe o -
mance. While acknowledging he di icul y o his ask and
i s lack o abili y o p o ide such an o e iew, he GenAI
ool (Cha GPT 4.0) p o ides some ecommenda ions o
a eas o u u e esea ch and e e ences o explo e. Howe e ,
e en hese sugges ions a e qui e misleading, as hey sugges
esea ch in many a eas ha a e no di ec ly ela ed o busi-
ness pe o mance (e.g., ma ke pe o mance, accoun ing pe -
o mance), such as e hical issues and wo k o ce dis up ion.
O e all, GenAI ools p esen a ascina ing mosaic o
po en ials wi hin he b oad spec um o li e a u e e iew
ypes. The ealiza ion o hese po en ials, howe e , hinges
on he c i ical examina ion o he capabili y o GenAI in
LR ac i i ies and he ca e ul de elopmen o he GenAI-
esea che collabo a ion model. Echoing ex an li e a u e on
using GenAI o LRs (Dwi edi e al., 2023; Ja enpaa &
Klein, 2024; Ngwenyama & Rowe, 2024; Pan e al., 2023),
we a gue o he impe a i e o esea che s o ecognize
hei unique s eng hs, such as in ui ion, nuances disce n-
ing, applying deep c i ical hinking in iden i ying knowl-
edge gaps, and inno a i e heo iza ion, as well as GenAI’s
ela i e ad an ages, such as me a-knowledge base, apidly
scanning as da abases, iden i ying pa e ns, and coding
hemes om exce p s wi h consis ency (Dasbo ough, 2023;
Pan e al., 2023). In addi ion, esea che s mus be awa e o
he limi a ions o GenAI in conduc ing LR ac i i ies. As we
highligh ed in ou indings, GenAI ends o o e emphasize
well-known li e a u e while po en ially o e looking less
popula bu equally impo an sou ces, leading o a biased
unde s anding o he opic. S uggling wi h he complexi y
o c i ically e alua ing p io esea ch, GenAI is no ye sui -
able o c i icizing and unco e ing new, unexplo ed esea ch
Table 3 (con inued)
Knowledge de elopmen ac i i y GenAI capabili ies GenAI limi a ions Recommenda ions o esea che s
Iden i ying esea ch gaps • Highligh and summa ize esea ch gaps and
limi a ions p e iously iden i ied in he li e a u e,
po en ially unco e ing pa e ns and hemes o
u u e esea ch di ec ions
• Map exis ing knowledge on pa icula com-
ponen s o models and heo ies, aiding in he
iden i ica ion o esea ch gaps
• GenAI is inhe en ly designed o ep oduce
exis ing knowledge (o esea ch gaps) a he han
iden i y new esea ch gaps, which in ol es ques-
ioning whe he “ he igh hings we e done”
• Resea che s should complemen he ini ial gaps
p o ided by GenAI wi h a ho ough li e a u e
e iew h ough he da abase o con i m ha he
iden i ied gaps a e indeed unde explo ed
• A e GenAI p o ided a summa y o knowledge
om exis ing li e a u e, esea che s should apply
c i ical hinking o spo he a eas ha a e unde -
explo ed
De eloping a esea ch agenda • Explo e insigh s om o he academic disciplines
o guide esea che s owa ds po en ial esea ch
pa hs and inspi e u he explo a ion, such as
inding heo ies, models, and me hodologies
used in simila con ex s
• Di ec ly que ying GenAI o esea ch p oposi-
ions, esea ch designs, and empi ical me hods is
unlikely o yield ac ionable esea ch agendas
• Based on he insigh s and gaps iden i ied wi h he
help o GenAI, esea che s should consul wi h
li e a u e and apply hei own expe ise o de elop
speci ic and ac ionable ecommenda ions, includ-
ing esea ch p oposi ions, designs, and empi ical
me hods, ensu ing hese a e well- ounded and
con ex ually ele an
Elec onic Ma ke s (2025) 35:13 Page 21 o 26 13
Fig. 12 Cha GPT ( e sion 4)
que y: a na a i e e iew on he
opic o IT business alue
Elec onic Ma ke s (2025) 35:13 13 Page 22 o 26
Fig. 13 Cha GPT ( e sion 4)
que y: a na a i e e iew on he
opic o IT business alue
Elec onic Ma ke s (2025) 35:13 Page 23 o 26 13
gaps; i also alls sho in suppo ing s udies ha equi e
e idence agg ega ion, especially in quan i a i e esea ch;
GenAI’s ole in heo y building and de eloping esea ch
agendas is also limi ed, as i canno di ec ly c ea e inno a i e
esea ch p oposi ions wi hou u he human in es iga ion
and in e p e a ion.
Consequen ly, we a gue o a human-cen ic syne gis ic
app oach whe e GenAI complemen s human esea che s
in LRs. We ecommend ha he c i ical esponsibili ies o
e iewing, c i iquing, alida ing heo ies, iden i ying gaps,
and ex ending knowledge es wi h human esea che s. They
a e poised o make he inal decisions on heo e ical appa-
a us selec ion, ensu ing alignmen wi h he esea ch ques-
ion and con ibu ion o he ield. GenAI ools se e no as
s andalone solu ions bu as ins umen al aides in he hands
o adep esea che s. Ou indings ha e shown ha GenAI
ools can au oma e he e ie al and ini ial analysis o li -
e a u e, en iching he LR p ocess by p o iding a b oad, ye
su ace-le el, o e iew o he exis ing knowledge landscape.
Howe e , hei limi a ions in dep h-o ien ed asks such as
c i ical e alua ion, heo y inno a ion, and knowledge gap
a e e iden . Human esea che s a e, he e o e, en isioned
i s as he di ec o s and hen as o e see s o p o ide guid-
ance, engage wi h, and e ine GenAI ou pu s o o e come
hose limi a ions. This human-cen ic syne gis ic app oach
in ol es a s a egic balance whe e he e iciency and b ead h
o GenAI’s li e a u e scanning and hema ic analysis capa-
bili ies a e le e aged o se he s age o deepe esea che -
led inqui ies. Resea che s’ c i ical hinking, c ea i e syn he-
sis, and e alua i e judgmen become he d i ing o ces ha
na iga e and in e p e GenAI-gene a ed insigh s, ans o m-
ing hem in o p o ound con ibu ions o knowledge. By os-
e ing a pa ne ship ha capi alizes on he s eng hs o bo h
GenAI and human esea che s, his model aims o ad ance
LR ac i i ies in a way ha is inclusi e, comp ehensi e, and
c i ical.
Conclusions
In his a icle, we explo e, om an epis emological pe spec-
i e, how GenAI ools may suppo IS esea che s in de el-
oping s andalone li e a u e e iews. Ou ocus is p ima ily
on he inno a ion goal o GenAI ools in scien i ic esea ch,
as opposed o he communica ion goal. We a gue and illus-
a e ha he e ec i eness o GenAI ools la gely depends
on, and a ies based on, speci ic knowledge de elopmen
ac i i ies. This di e si y leads o a mo e de ailed unde s and-
ing o how GenAI ools can assis in he de elopmen o
a ious ypes o li e a u e e iews, add essing he esea ch
ques ion posed in ou in oduc ion. While we ad ise agains
gene a ing li e a u e e iews in a single s ep wi h a single
que y, ou expe imen s wi h di e en GenAI ools lead o
posi i e esul s ega ding hei u ili y wi hin a human-AI
collabo a i e p ocess. These indings in i e u he esea ch
on how such ools may assis , o possibly hinde , schola s in
pu suing he inno a ion goal o hei esea ch.
Recen s udies ha e highligh ed signi ican e olu ion in
he de elopmen and usage o hese ools wi hin a ela i ely
b ie pe iod. This p og ession includes claims o educing
he capabili ies o Cha GPT—o en colloquially e e ed o
as “ne ing”—and he con inual in oduc ion o inno a i e
ea u es, such as plugins and web b owsing capabili ies.
The e o e, i is c ucial o esea che s o be well-in o med
abou ad ancemen s in his e ol ing ield.
The capabili ies and limi a ions o GenAI, as ou lined in
ou pape , ca y signi ican e hical implica ions o conduc -
ing LR. I is essen ial ha esea che s, a he han AI, bea
ul ima e esponsibili y o he in eg i y o hei wo k (Schlag-
wein & Willcocks, 2023). As GenAI inc easingly in eg a es
in o LR p ocesses, i is c ucial o adhe e o he co e alues
o accoun abili y, anspa ency, eplicabili y, and esponsi-
bili y (Blau e al., 2024). To main ain he in eg i y o scien-
i ic inqui y, all uses o GenAI in LR mus be anspa en ly
documen ed, wi h explici a ibu ion dis inguishing be ween
human e o s and AI-gene a ed con en . This documen a ion
suppo s he eplicabili y o s udies and upholds igo ous sci-
en i ic s anda ds. Mo eo e , a p ima y e hical conce n is he
managemen o biases inhe en in GenAI ools (Feue iegel
e al., 2024; S ahl & Eke, 2024). Ou a icle highligh s ha
esea che s should adop p oac i e measu es o ensu e ha
biases do no dis o esea ch ou comes. These measu es
include c oss- e i ying and alida ing in o ma ion, supple-
men ing GenAI ou pu s wi h up- o-da e da abases, employ-
ing i e a i e que ying, and c i ically e alua ing he ou pu s o
GenAI. Addi ionally, ensu ing he selec ion o GenAI ools
ha use di e se aining da a can help minimize inhe en
biases. Fu he mo e, emphasizing human o e sigh and igo -
ous c i ical e alua ion emains i al o sc u inize AI-gene a ed
ou pu s e ec i ely. Concludingly, ou s udy enhances he aca-
demic communi y’s unde s anding o GenAI ools’ po en ials,
limi a ions, and pe spec i es in suppo ing esea che s in hei
pu sui o knowledge de elopmen h ough li e a u e e iews.
Despi e he p omising insigh s p esen ed in his a icle,
se e al limi a ions wa an conside a ion. Fi s , we acknowl-
edge ha ou pape elies on hypo he ical examples and ou -
pu s gene a ed by GenAI ools like Cha GPT, and we ec-
ognize ha he absence o empi ical s udies o case s udies
limi s he obus ness o ou indings. While we ha e epo ed
on some ea ly a emp s in his di ec ion in he backg ound
sec ion, ou unde s anding emains incomple e. We encou -
age u he empi ical s udies o demons a e he p ac ical
u ili y and limi a ions o GenAI ools in eal-wo ld li e a u e
e iews. Such esea ch will enhance he obus ness o ind-
ings in his ield and guide he de elopmen o mo e e ec-
i e AI-assis ed me hodologies.
Elec onic Ma ke s (2025) 35:13 13 Page 24 o 26
Second, u u e esea ch should explo e he in eg a ion
o mul imodal GenAI ools ha go beyond ex -based in e -
ac ions. As GenAI echnologies e ol e, now inco po a -
ing capabili ies such as oice ecogni ion and isual da a
p ocessing, in es iga ing how hese ools can be e ec i ely
combined wi h adi ional esea ch me hodologies would
o e aluable insigh s in o how esea che s can le e age AI
in academic esea ch. In con as o and complemen a y o
ou epis emological ocus, his mo e ope a ional and echni-
cal pe spec i e can shed ligh on how mul imodal inpu can
be exploi ed by GenAI.
Thi d, he apid e olu ion o GenAI ools poses a signi i-
can challenge o he gene alizabili y o ou conclusions. Find-
ings pe inen oday may no hold in he nea u u e as hese
echnologies ad ance. Ano he limi a ion is he inconsis ency
in ool pe o mance, which e e s o he a ia ion in how
GenAI ools espond o p omp s, bo h wi hin he same ool by
using he same o sligh ly di e en que ies and using he same
que ies ac oss di e en GenAI ools. These di e ences a e
due o a ia ions in he unde lying algo i hms, model a chi-
ec u es, and he aining da a used o each ool. As a esul ,
he ou pu gene a ed by a GenAI ool can a y signi ican ly
depending on he speci ic p omp o use case o due o he
buil -in andomness, making i di icul o consis en ly assess
hei eliabili y. Consequen ly, while he a icle ad oca es o
a human-cen e ed app oach, i should be acknowledged ha
using GenAI ools in he li e a u e e iew p ocess mus be
app oached wi h an unde s anding o hei limi a ions.
Funding Open Access unding enabled and o ganized by P ojek
DEAL.
Open Access This a icle is licensed unde a C ea i e Commons A i-
bu 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 Commons 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 ecommons.o g/licenses/by/4.0/.
Re e ences
Ala i, M., & Leidne , D. E. (2001). Re iew: Knowledge managemen
and knowledge managemen sys ems: Concep ual ounda ions and
esea ch issues. MIS Qua e ly, 25(1), 107–136. h ps:// doi. o g/
10. 2307/ 32509 61
Ala i, M., Leidne , D. E., & Mousa i, R. (2024). A knowledge manage-
men pe spec i e o gene a i e a i icial in elligence. Jou nal o
he Associa ion o In o ma ion Sys ems, 25(1), 1–12. h ps:// doi.
o g/ 10. 17705/ 1jais. 00859. A ailable a : h ps:// aisel. aisne . o g/
jais/ ol25/ iss1/ 15. Accessed24 Oc 2024.
Al esson, M., & Sandbe g, J. (2011). Gene a ing esea ch ques ions
h ough p oblema iza ion. Academy o Managemen Re iew,
36(2), 247–271. h ps:// doi. o g/ 10. 5465/ am . 2009. 0188
A ksey, H., & O’Malley, L. (2005). Scoping s udies: Towa ds a me h-
odological amewo k. In e na ional Jou nal o Social Resea ch
Me hodology, 8(1), 19–32. h ps:// doi. o g/ 10. 1080/ 13645 57032
00011 9616
Bail, C. A. (2024). Can gene a i e AI imp o e social science? P oceed-
ings o he Na ional Academy o Sciences, 121(21), e2314021121.
h ps:// doi. o g/ 10. 1073/ pnas. 23140 21121
Benbya, H., S ich, F., & Tamm, T. (2024). Na iga ing gene a i e a i-
icial in elligence p omises and pe ils o knowledge and c ea i e
wo k. Jou nal o he Associa ion o In o ma ion Sys ems, 25(1),
23–36. h ps:// doi. o g/ 10. 17705/ 1jais. 00861. A ailable a : h ps://
aisel. aisne . o g/ jais/ ol25/ iss1/ 13. Accessed 24 Oc 2024.
Bende , E. M., Geb u, T., McMillan-Majo , A., & Mi chell, M. (2021).
On he dange s o s ochas ic pa o s: Can language models be oo
big?In P oceedings o he ACM con e ence on ai ness, accoun -
abili y, and anspa ency(FAccT '21)(pp. 610–623). Associa ion
o Compu ing Machine y.h ps:// doi. o g/ 10. 1145/ 34421 88. 34459
22
Be ge , K., Klie , J., Klie , M., & P obs , F. (2014). A e iew o in o -
ma ion sys ems esea ch on online social ne wo ks. Communica-
ions o he Associa ion o In o ma ion Sys ems, 35(1), 8. h ps://
doi. o g/ 10. 17705/ 1CAIS. 03508
Blau, W., Ce , V. G., En iquez, J., F ancisco, J. S., Gasse , U., G ay,
M. L., Jamieson, K. H., Haug, G. H., Hennessy, J. L., Ho i z, E.,
Kaise , D. I., London, A. J., Lo ell-Badge, R., McNu , M. K.,
Minow, M., Mi chell, T. M., Ness, S., Pa hasa a hy, S., Pe lmu -
e , S., P ess, W. H., Wing, J. M., & Wi he ell, M. (2024). P o ec -
ing scien i ic in eg i y in an age o gene a i e AI. P oceedings o
he Na ional Academy o Sciences o he Uni ed S a es o Ame ica,
121(22), e2407886121. h ps:// doi. o g/ 10. 1073/ pnas. 24078 86121
Bu iak, J. M., Akinwande, D., A zi, N., B inke , C. J., Bu ows, C.,
Chan, W. C. W., Chen, C., Chen, X., Chhowalla, M., Chi, L.,
Chueh, W., C udden, C. M., Di, C. D., Glo ze , S. C., He sam,
M. C., Ho, D., Hu, T. Y., Huang, J., Ja ey, A., … Ye, J. (2023).
Bes p ac ices o using AI when w i ing scien i ic manusc ip s:
Cau ion, ca e, and conside a ion: C ea i e science depends on
i . ACS Nano, 17(5), 4091–4093. h ps:// doi. o g/ 10. 1021/ acsna
no. 3c025 84
Ca d, N. A. (2011). Applied me a-analysis o social science esea ch.
The Guil o d P ess. h ps:// doi. o g/ 10. 1111/ ins . 12011_ 17
Dasbo ough, M. T. (2023). Awe-inspi ing ad ancemen s in AI: The
impac o Cha GPT on he ield o o ganiza ional beha io . Jou -
nal o O ganiza ional Beha io , 44(2), 177–179. h ps:// doi. o g/
10. 1002/ job. 2695
Da ison, R. M., Laume , S., Ta a da , M., & Wong, L. H. (2023).
Pickled eggs: Gene a i e AI as esea ch assis an o co-au ho ?
In o ma ion Sys ems Jou nal, 33(5), 989–994. h ps:// doi. o g/ 10.
1111/ isj. 12491
Da ison, R. M., Chugh ai, H., Nielsen, P., Ma abelli, M., Iannacci,
F., an O enbeek, M., Ta a da , M., T enz, M., Techa assanaso-
on o n, A. A., And ade, A. D., & Pan eli, N. (2024). The e hics
o using gene a i e AI o quali a i e da a analysis. In o ma ion
Sys ems Jou nal. h ps:// doi. o g/ 10. 1111/ isj. 12504
de la To e-López, J., Ramí ez, A., & Rome o, J. R. (2023). A i i-
cial in elligence o au oma e he sys ema ic e iew o scien i ic
li e a u e. Compu ing, 105, 2171–2194. h ps:// doi. o g/ 10. 1007/
s00607- 023- 01181-x
DeLone, W. H., & McLean, E. R. (1992). In o ma ion sys ems suc-
cess: The ques o he dependen a iable. In o ma ion Sys ems
Resea ch, 3(1), 60–95. h ps:// doi. o g/ 10. 1287/ is e.3. 1. 60
Dwi edi, Y. K., Kshe i, N., Hughes, L., Slade, E. L., Jeya aj, A., Ka ,
A. K., Baabdullah, A. M., Koohang, A., Ragha an, V., Ahuja, M.,
Albanna, H., Albash awi, M. A., Al-Busaidi, A. S., Balak ishnan,
