Abbad-Andaloussi, Amine; F ancesche i, Ma co; López, Hugo A.; Sch eibe ,
Clemens; Webe , Ba ba a
A icle — Published Ve sion
When a model gi es you mixed signals: cogni i e e ec s
and isual beha io
P ocess Science
P o ided in Coope a ion wi h:
Sp inge Na u e
Sugges ed Ci a ion: Abbad-Andaloussi, Amine; F ancesche i, Ma co; López, Hugo A.; Sch eibe ,
Clemens; Webe , Ba ba a (2025) : When a model gi es you mixed signals: cogni i e e ec s and
isual beha io , P ocess Science, ISSN 2948-2178, Sp inge In e na ional Publishing, Cham, Vol. 2,
Iss. 1,
h ps://doi.o g/10.1007/s44311-025-00022-8
This Ve sion is a ailable a :
h ps://hdl.handle.ne /10419/330732
S anda d-Nu zungsbedingungen:
Die Dokumen e au EconS o dü en zu eigenen wissenscha lichen
Zwecken und zum P i a geb auch gespeiche und kopie we den.
Sie dü en die Dokumen e nich ü ö en liche ode komme zielle
Zwecke e iel äl igen, ö en lich auss ellen, ö en lich zugänglich
machen, e eiben ode ande wei ig nu zen.
So e n die Ve asse die Dokumen e un e Open-Con en -Lizenzen
(insbesonde e CC-Lizenzen) zu Ve ügung ges ell haben soll en,
gel en abweichend on diesen Nu zungsbedingungen die in de do
genann en Lizenz gewäh en Nu zungs ech e.
Te ms o use:
Documen s in EconS o may be sa ed and copied o you pe sonal
and schola ly pu poses.
You a e no o copy documen s o public o comme cial pu poses, o
exhibi he documen s publicly, o make hem publicly a ailable on he
in e ne , o o dis ibu e o o he wise use he documen s in public.
I he documen s ha e been made a ailable unde an Open Con en
Licence (especially C ea i e Commons Licences), you may exe cise
u he usage igh s as speci ied in he indica ed licence.
h ps://c ea i ecommons.o g/licenses/by/4.0/
RESEARCH Open Access
© The Au ho (s) 2025. 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 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 e a i e c o m m o n s . o g / l i c e n s e s / b y / 4 . 0 / .
Abbad-Andaloussi e al. P ocess Science (2025) 2:19
h ps://doi.o g/10.1007/s44311-025-00022-8
† A m i n e Abbad-Andaloussi
and Ma co F ancesche i
con ibu ed equally o his
wo k.
*Co espondence:
Amine Abbad-Andaloussi
[email protected]
Ma co F ancesche i
ma co[email p o ec ed]
1Uni e si y o S .Gallen, S . Gallen,
Swi ze land
2Technical Uni e si y o Denma k,
Kgs. Lyngby, Denma k
3Ka ls uhe Ins i u e o Technology,
Ka ls uhe, Ge many
When a model gi es you mixed signals:
cogni i e e ec s and isual beha io
AmineAbbad-Andaloussi1*†, Ma coF ancesche i1*†, Hugo A.López2, ClemensSch eibe 3 and Ba ba aWebe 1
In oduc ion
Business p ocess models a e widely used o imp o e he unde s anding o business p o-
cesses by combining in o ma ion isualiza ion p inciples wi h he seman ics o o mal
modeling languages. Se e al p ocess modeling languages, such as Business P ocess
Model and No a ion (BPMN)(OMG 2014) and Pe i ne s(Reisig 2012), o e o mally
de ined seman ics and a e commonly adop ed o his pu pose. Howe e , he e is g ow-
ing ecogni ion ha business p ocess models, e en wi h o mal seman ics, may be p one
o ambigui y(Dijkman e al. 2008; Fan e al. 2016; Pi ke e al. 2015), aising he need o
u he in es iga ion in o i s impac . Ambigui y in a p ocess model is a phenomenon
ha can lead o mul iple alid in e p e a ions o he p ocess: o an example o lexical
ambigui y a ec ing ac i i y labels, an ac i i y labeled Schedule e iew can be in e p e ed
as an ac i i y o scheduling a e iew o as an ac i i y o e iewing a schedule. A ecen
s udy inF ancesche i e al. (2023) examined ambigui y ac oss di e en a i ac s wi hin
he Business P ocess Managemen (BPM) li ecycle, including in o mal speci ica ions,
Abs ac
Ambigui y in business p ocess models can esul in mul iple in e p e a ions by model
eade s. This leads o undesi able ou comes such as misunde s andings, unclea
alloca ion o esponsibili ies, and unexpec ed beha io s. Despi e hese po en ial
consequences, he impac o ambigui y on model eade s has ecei ed limi ed
a en ion so a . This a icle p esen s an eye- acking s udy designed o in es iga e
he e ec s o a ious ypes o ambigui y (i.e., layou , seman ic, syn ac ic, and lexical)
on eade s’ cogni i e load, comp ehension, and isual associa ions while in e p e ing
p ocess models. In addi ion, he s udy del es in o he beha io s o model eade s
when esol ing ambigui y in p ocess models. These beha io s a e in es iga ed
ollowing a quali a i e app oach combining bo h eye- acking and hink-aloud
da a. The esul s demons a e ha ambigui ies signi ican ly in luence cogni i e
load, comp ehension, and isual associa ions, emphasizing he nega i e e ec s o
ambigui y. Mo eo e , he quali a i e insigh s sugges ha pa icipan s exhibi speci ic
beha io s when ying o esol e ambigui ies. These indings unde sco e he need o
ad anced mechanisms o de ec and mi iga e ambigui y in p ocess models.
Keywo ds Ambigui y, P ocess models, Eye- acking, Cogni i e load, Visual beha io
P ocess Science
Page 2 o 38Abbad-Andaloussi e al. P ocess Science (2025) 2:19
models, and e en logs. The s udy highligh ed he pe asi e na u e o ambigui y in
p ocess models h ough examples d awn om li e a u e and public da ase s. As p o-
cess models a e equen ly adop ed as a means o communica ion be ween s akehold-
e s, he exis ence o mul iple alid in e p e a ions can unde mine hei communica ion
e ec i eness due o he po en ial misma ch be ween a eade ’s in e p e a ion and he
modele ’s in ended message. Indeed, as de ailed inF ancesche i e al. (2023), ambigu-
i y in p ocess models en ails he isk o misunde s andings, unclea esponsibili ies,
unexpec ed beha io s, as well as cascading e ec s ac oss o he p ocess- ela ed a i ac s
managed in he BPM li ecycle. Consequen ly, ambigui y in p ocess models can comp o-
mise ecognized bene i s o adop ing P ocess-awa e In o ma ion Sys ems such as clea
p ocess-based communica ion be ween s akeholde s, inc eased e iciency, educed
edundancies, and moni o ing suppo (Dumas e al. 2018), by diminishing hese ben-
e i s(Kindle 2009; Oppl 2016). The e o e, co ec ly comp ehending p ocess models is
impe a i e o he success ul enginee ing and ope a ion o a P ocess-awa e In o ma ion
Sys em(Oppl 2016).
P e ious esea ch has examined a ious ypes o ambigui ies and s a egies o educe
hem in business p ocess models a design ime (c .Fan e al. (2016); Mendling e al.
(2010b); Pi ke e al. (2015)). Howe e , o he bes o ou knowledge, no p io s udy has
alida ed he impac o ambigui y on he cogni i e and beha io al aspec s o p ocess
model eade s. On he one hand, p io s udies ha e explo ed he impac on hese aspec s
in ela ion o model complexi y(Pe usel e al. 2017), model quali y(Heggse e al. 2015),
no a ional de iciencies in modeling languages(Figl e al. 2013), o eade -speci ic a i-
bu es like modeling expe ience and p ocess knowledge(Mendling e al. 2012). On he
o he hand, he speci ic impac o ambigui y on he cogni i e and beha io al aspec s
emains, o da e, unexamined. Unlike aspec s such as model co ec ness, ambigui y is
cha ac e ized by he po en ial o mul iple equally alid in e p e a ions. While in ce -
ain con ex s, such as no ma i e p ocesses, his ea u e is in en ionally inco po a ed
o p o ide lexibili y in model in e p e a ion, in o he con ex s ambigui y can ha e an
unin ended nega i e impac as i gene a es con usion(F ancesche i e al. 2023). This
s udy ocuses on unde s anding how his unique cha ac e is ic, which can lead o mul-
iple in e p e a ions, in luences he cogni i e and beha io al aspec s o p ocess model
eade s.
Pa icula ly, in his s udy, we add ess he challenge o measu ing he cogni i e impac
o ambigui y in p ocess models on eade s’ cogni i e load (i.e., men al e o ), comp e-
hension, and isual associa ions (i.e., shi s o a en ion be ween model elemen s, which
sugges s an inc eased men al demand o in eg a ing in o ma ion, c .Be a e al. (2019)).
Addi ionally, we explo e he isual beha io o model eade s when con on ed wi h
ambigui y. To achie e his, we use eye- acking, a me hod ha has p o en e ec i e in
p e ious esea ch on p ocess model comp ehension asks, o e ing aluable insigh s in o
he cogni i e and beha io al aspec s o model eade s (c .Be a e al. (2019); Pe usel and
Mendling (2013); Pe usel e al. (2016); Wang e al. (2022); Win e e al. (2023)).
O e all, we add ess he ollowing esea ch ques ions:
•RQ1. How do di e en ambigui ies a ec model eade s’ cogni i e load,
comp ehension and isual associa ions?
•RQ2. Wha pa e ns o isual beha io do model eade s exhibi when esol ing
ambigui ies in p ocess models?
Page 3 o 38Abbad-Andaloussi e al. P ocess Science (2025) 2:19
This pape ex ends ou p io wo k p esen ed inF ancesche i e al. (2024), which in es-
iga ed RQ1, by addi ionally in es iga ing RQ2. To answe RQ1, we in es iga ed he eye
mo emen s o model eade s while hey pe o med di e en comp ehension asks on
p ocess models wi h and wi hou ambigui ies. To answe RQ2, we pe o med a quali a-
i e s udy in which we obse ed he eye mo emen s o model eade s when con on ed
wi h ambigui y, and iangula ed hese obse a ions wi h he e bal in o ma ion p o-
ided by he model eade s du ing e ospec i e hink-aloud sessions. Ou esul s dem-
ons a e he use ulness o eye- acking o de ec ambigui y and ha ambiguous p ocess
models lead o highe cogni i e load, challenged comp ehension, and inc eased isual
associa ions (RQ1) when compa ed o p ocess models wi hou ambigui ies. Mo eo e ,
ou quali a i e s udy e ealed i e dis inc isual beha io pa e ns ha eme ge when
model eade s a e con on ed wi h ambigui y, each associa ed wi h a ying le els o
isual a en ion, anging om andom o highly ocused (RQ2). Ha ing demons a ed
he nega i e e ec s o ambigui y and how i in luences model eade s’ cogni i e aspec s
and isual beha io , u u e wo k could ocus on de eloping me hods and echniques
o suppo model eade s by au oma ically de ec ing ambigui ies in p ocess models, o
assis ing hem in ocusing hei a en ion on model elemen s ha p o ide disambigua-
ion cues. Ou esul s also s imula e u he s udies on he impac o ambigui y in ela-
ion o expe ise and on disambigua ion s a egies.
This pape is s uc u ed as ollows: in Backg ound and ela ed wo ksec ion, we ecall
backg ound concep s and o mally de ine ambigui ies in p ocess models. In Resea ch
me hodsec ion, we epo on ou s udy design. In Findingssec ion, we epo on he
indings. In Discussionsec ion, we elabo a e on a discussion o he indings, he implica-
ions o he s udy esul s, and h ea s o alidi y. In Conclusionsec ion, we conclude he
pape .
Backg ound and ela ed wo k
In his sec ion, we i s es ablish he heo e ical ounda ions o ambigui y in BPM, o -
malizing ou ambigui y ypes ound in p ocess models (c . Ambigui y in p ocess mod-
elssec ion). We hen se he heo e ical ounda ions on cogni i e heo ies ele an o
ou empi ical s udy, se ing he unde pinnings o and mo i a ing ou s udy design (c .
Cogni i e load, comp ehension, and isual associa ionssec ion). Finally, we p o ide
an o e iew o quali a i e app oaches in es iga ing modele s’ isual beha io when
engaging wi h p ocess models (c . Quali a i e analysis o isual beha io du ing p ocess
comp ehensionsec ion).
Ambigui y in p ocess models
In gene al, ambigui y in BPM e e s o he po en ial o a business p ocess ep esen a-
ion o yield mul iple admissible in e p e a ions. Acco ding o he cha ac e iza ion p o-
ided inF ancesche i e al. (2023), speci ically ambigui y in p ocess models may a ise
om in insic ac o s (i.e., speci ic o he modeling language, such as he inabili y o ep-
esen he esou ces pe spec i e in s anda d Pe i ne s) o ex insic ac o s (i.e., ela ed
o he modeling ask, such as an unde speci ied ga eway condi ion by an inexpe ienced
modele no alida ed by he modeling ool).
As p ocess models a e o malized using modeling languages ha combine bo h ex-
ual and g aphical elemen s, hey encompass he ollowing aspec s: syn ac ic aspec s in
Page 4 o 38Abbad-Andaloussi e al. P ocess Science (2025) 2:19
ela ion o he use o he modeling language g amma , seman ic aspec s in ela ion o he
modeled beha io , lexical aspec s in ela ion o ex ual elemen s, and layou aspec s in
ela ion o he g aphical p esen a ion. Since all hese aspec s in luence a model eade ’s
in e p e a ion o he model, in his s udy we aim o ex end he a o emen ioned ambigu-
i y cha ac e iza ion(F ancesche i e al. 2023) by de ining layou 1, seman ic, syn ac ic,
and lexical ambigui ies in p ocess models. To he bes o ou knowledge, hese ypes o
ambigui ies a e s ill no clea ly de ined in he BPM li e a u e. To de ine hese ambigui-
ies in he BPM con ex , we conside ed wo main op ions.
On he one hand, we could adop co esponding de ini ions om he ield o linguis-
ics(Be y And Kams ies 2004; Fo uny And Pay a ó 2024; Senne 2011). Howe e , such
de ini ions apply speci ically o na u al language, which is ypically p esen ed in o al
o ex ual o m, and do no ully apply o he diag amma ic na u e o p ocess models,
which combine bo h ex ual and g aphical elemen s. Thus, de ini ions om linguis ics
do no ully cap u e he ways in which ambigui y a ises in business p ocess models, since
p ocess model in e p e a ion depends no only on na u al language bu also on diag am-
ma ic con en ions, domain mappings, and layou (Figl 2017; Haisjackl e al. 2015). Mo e-
o e , he dis inc ion be ween he di e en ypes o ambigui y in linguis ics does no ully
align wi h he one om ou BPM ield. Fo example, he BPM li e a u e includes layou
cha ac e is ics as pa o p agma ic aspec s(Haisjackl e al. 2015), whe eas in linguis-
ics p agma ics conce ns con ex ual and in e en ial aspec s o communica ion ela ed o
speake in en (Fo uny And Pay a ó 2024).
On he o he hand, we could also ely on p io BPM li e a u e ha has deal wi h some
o hese ambigui ies (c .Amna and Poels (2022); Fan e al. (2016); Haisjackl e al. (2015);
Leopold e al. (2010); Mendling e al. (2010b); Pi ke e al. (2015)), bu has no achie ed
consolida ed BPM-speci ic de ini ions. Mo eo e , p io li e a u e ha has s udied model
quali y aspec s in BPM could also be ele an (c .K ogs ie (2012, 2016); Lindland e al.
(1994)) as he iola ion o quali y guidelines o en leads o ambigui y. Howe e , ambi-
gui y and quali y a e di e en concep s. Indeed, while ambigui ies a e o en associa ed
wi h quali y issues, in gene al ambigui y is no educible o quali y: e en a o mally co -
ec and high-quali y model may gi e ise o ambigui y. This is obse ed, o example, in
models o legal p ocesses, which a e o high-quali y bu can in en ionally be designed
ambiguous o allow o a lexible applica ion o he law(Hildeb and 2018; Slaa s e al.
2013). As a esul , we decided o de elop ou own BPM-speci ic de ini ions o ambigu-
i y, d awing om he non-consolida ed de ini ions in exis ing BPM li e a u e, while also
being in o med by p ocess model quali y amewo ks.
De ini ion 1 (Layou Ambigui y) (Adap ed omAmna and Poels (2022); Haisjackl e al.
(2015)) Layou ambigui y is a phenomenon ha occu s a he layou le el causing a p o-
cess model o lack cla i y in one o mo e p ocess pe spec i es, allowing o mul iple in e -
p e a ions wi hou a ec ing he beha io o he execu able p ocess model.
In ou BPM-speci ic iew, layou ambigui y e e s o ambigui y ha a ises no om he
s uc u e o seman ics o a p ocess model pe se, bu om aspec s ela ed o he p e-
sen a ion and layou o he model. This includes, o example, o e lapping edges, poo ly
aligned lows, o ambiguous isual g oupings–which may allow o mul iple plausible
1 In ou p e ious wo k(F ancesche i e al. 2024), his ambigui y is called p agma ic. He e, we use he mo e speci ic
e m layou o be e ma k he dis inc ion om p agma ic aspec s in linguis ics, which a e no conce ned wi h layou .
Page 5 o 38Abbad-Andaloussi e al. P ocess Science (2025) 2:19
in e p e a ions o he p ocess elemen s. Figu e1 illus a es an example o layou ambi-
gui y. In he example, o e lapping con ol low edges be ween he ac i i ies can be
obse ed. This o e lap allows o mul iple in e p e a ions o he p ecedence cons ain s
be ween he ac i i ies. In he speci ic model agmen , i is unclea whe he ac i i y
“Fo ge piece LDF” is ollowed by “D ill piece LBG” o by “Ex ude piece CTG”. The po en-
ial o ambigui y a ising om he p esen a ion layou was discussed inPe e (1995),
howe e wi hou ouching upon he impac on a eade ’s cogni i e and beha io al
aspec s. P ocess model comp ehension in ela ion o layou aspec s was u he s udied
in subsequen wo ks such asFigl (2017); Haisjackl e al. (2015); Pe usel e al. (2016).
S ill, he e ec on comp ehension speci ically a ising om ambigui y in he layou , i.e.,
he mul iple possible in e p e a ions yielded by he model layou , was no conside ed by
hese s udies. The e o e, he ques ion abou how (layou ) ambigui ies a ec model ead-
e s’ cogni ion emained, so a , unadd essed.
De ini ion 2 (Seman ic Ambigui y) (Adap ed omFan e al. (2016); Haisjackl e al.
(2015); K ogs ie (2012)) Seman ic ambigui y is a linguis ic phenomenon ela ed o he
usage o a modeling language ha occu s as a consequence o a p ocess model lacking
alidi y (i.e., all s a emen s in he model a e co ec and ela ed o he p ocess) o com-
ple eness (i.e., he e is a one- o-one mapping be ween model cons uc s and domain con-
cep s), o o di e ences in domain and con ex knowledge be ween model c ea o s and
eade s. I allows o mul iple in e p e a ions by model eade s due o he eade s no
being able o es ablish clea mappings be ween model cons uc s and domain concep s.
In ou BPM-speci ic iew, seman ic ambigui y e e s o si ua ions whe e elemen s o a
p ocess model can be in e p e ed in mo e han one way wi h espec o he ( o mal) p o-
cess model beha io , in ela ion o he unde lying business domain. This ambigui y may
esul om de iciencies in seman ic quali y such as lack o alidi y o comple eness, bu
can also a ise in o mally co ec models due o di e ences in domain knowledge o con-
ex among eade s. Unlike seman ic ambigui y in linguis ics, which ocuses on mul iple
meanings o wo ds o ph ases, seman ic ambigui y in BPM encompasses he b oade
challenge o aligning model elemen s wi h a sha ed domain unde s anding. Ou ocus
lies on he in e p e i e consequence: ha di e en eade s may assign di e en mean-
ings o he same model agmen , ega dless o i s o mal co ec ness. An example o
seman ic ambigui y in ela ion o alidi y is illus a ed in Fig.2 (le ). He e, he sequence
“App o e piece LDF” – “Rejec piece LDF” is p esen ed. Simply applying common sense,
a model eade would na u ally assume ha a piece is ei he app o ed o ejec ed, since
hese e bs exp ess opposi e ac ions. The e o e, in he example i is unclea whe he he
Fig. 1 BPMN agmen o a p ocess wi h layou ambigui y
Page 6 o 38Abbad-Andaloussi e al. P ocess Science (2025) 2:19
wo ac i i ies e e o di e en pieces o he ac i i ies a e no supposed o be bo h exe-
cu ed in he same ace. An example o seman ic ambigui y in ela ion o comple eness
bo owed omFan e al. (2016) is ha o a single ac i i y “Send eques ” used o model
wo dis inc ypes o eques s in an online auc ion– om he selle o ini ia e he auc ion
and a buye o join he auc ion. The lack o a one- o-one mapping be ween domain con-
cep s ( wo dis inc ac i i ies) and model elemen s (one ac i i y) esul s in a loss o he
ac i i y seman ics which makes i unclea which eques he modeled ac i i y e e s o.
Seman ic ambigui y in p ocess models was s udied inDijkman e al. (2008), speci ically
ocusing on BPMN models. Seman ic issues de i ing om p ocess model quali y issues
we e in es iga ed inHaisjackl e al. (2015); K ogs ie (2012). To educe seman ic ambi-
gui y in p ocess models, a possible app oach is o le e age seman ic in o ma ion abou
he p ocess domain ep esen ed h ough on ologies, as p oposed by he au ho s inFan
e al. (2016). In he examples, he app oach helps o (i) alida e he ac i i y a angemen ,
de ec ing ha “App o e piece LDF” and “Rejec piece LDF” in he sequence co espond
o mu ually exclusi e on ology classes, and (ii) iden i y cons uc excess, de ec ing he
misma ch be ween one modeled ac i i y and wo sepa a e on ology classes (c .Fan e al.
(2016)).
De ini ion 3 (Syn ac ic Ambigui y) (Adap ed omAmna and Poels (2022)) Syn ac ic
ambigui y is a g amma ical phenomenon ha occu s when a agmen o a p ocess model
M o malized in a modeling language
L
can be pa sed using mo e han one g amma ical
s uc u e o
L
, allowing o mul iple possible in e p e a ions o M.
In ou BPM-speci ic iew, syn ac ic ambigui y is conce ned wi h he pa sing o he mod-
eling language cons uc s. Figu e2 ( igh ) illus a es an example o syn ac ic ambigu-
i y. In he igu e, he con ol low spli s a an XOR-ga eway, which has no condi ions
a ached. The e o e, he ga eway can be in e p e ed ei he as an unde speci ied (i.e.,
wi h unknown condi ion) XOR-ga eway, which makes he subsequen ac i i ies mu u-
ally exclusi e, o as an inco ec ly assigned AND-ga eway, which en o ces ha he
subsequen ac i i ies a e bo h execu ed in all aces. Syn ac ic ambigui y in p ocess
desc ip ions was in es iga ed inAmna and Poels (2022), while he po en ial eme gence
o ambigui y due o syn ac ic quali y de iciencies in BPMN models was in es iga ed
inHaisjackl e al. (2015). The ela ion be ween syn ac ic aspec o p ocess models and
model unde s andabili y was discussed inCo adini e al. (2018), along wi h he p oposal
o a se o modeling guidelines o a oid he eme gence o syn ac ic ambigui y. Despi e
hese s udies, o he bes o ou knowledge, he ela ion be ween syn ac ic ambigui y and
he cogni i e and beha io al aspec s o model eade s emains, o da e, unadd essed.
Fig. 2 BPMN agmen s o p ocesses wi h seman ic (le ) and syn ac ic ( igh ) ambigui ies
Page 7 o 38Abbad-Andaloussi e al. P ocess Science (2025) 2:19
De ini ion 4 (Lexical Ambigui y) (Adap ed omPi ke e al. (2015)) Lexical ambigui y is
a linguis ic phenomenon ela ed o he usage o he na u al language ha occu s when a
ex ual label in a p ocess model can be in e p e ed in mul iple ways.
Acco ding o he s udies p esen ed inCo adini e al. (2018); Pus ejo sky (1998), lexical
ambigui y in BPM can be caused by he use o abb e ia ions, homonyms, synonyms, and
polysemic wo ds (i.e, wo ds wi h mul iple meanings). Fo example, conside an ac i -
i y labeled “Recei e epo ” ollowed by an ac i i y labeled “E alua e summa y”. He e,
i is unclea whe he he e ms “ epo ” and “summa y” a e used as synonyms and
e e o he same da a objec , o hey e e o di e en da a objec s. The use o speci ic
labeling s yles, such as passi e oice o noun o m o e bs, migh as well lead o lexi-
cal ambigui y(Mendling e al. 2010b). Conside he example bo owed omPi ke e al.
(2015) o an ac i i y labeled “Plan in eg a ion”: he ac i i y could be in e p e ed ei he
as he planning o some in eg a ion, o as he in eg a ion o some plan. P io wo k om
Mendling e al. in es iga ed he usage o ac i i y labels and he esul ing lexical ambi-
gui y(Mendling e al. 2010a, b). These s udies esul ed in guidelines(Mendling e al.
2010c), e ac o ing ecommenda ions(Leopold e al. 2010), and au oma ic app oaches
o de ec and esol e lexical ambigui y in p ocess models(Pi ke e al. 2015). To he bes
o ou knowledge, howe e , hese s udies did no explici ly in es iga e he impac on he
speci ic cogni i e and beha io al aspec s ha we se o measu e in his pape .
Cogni i e load, comp ehension, and isual associa ions
In his sec ion, we p o ide backg ound and se he heo e ical unde pinnings o in es-
iga e he impac o di e en ambigui ies in p ocess models on eade s’ cogni i e load,
comp ehension, and isual associa ions. This in u n will allow us o add ess RQ1 (c .
In oduc ionsec ion).
Cogni i e Load. Cogni i e load deno es he wo kload imposed on he human wo king
memo y du ing asks equi ing men al p ocessing(Swelle 2011). The Cogni i e Load
Theo y (CLT) disce ns h ee ypes o cogni i e load: In insic, Ex aneous, and Ge -
mane(Swelle 2011; Chen e al. 2016). In insic load eme ges om he essen ial com-
plexi y o he p ocess model, which is inhe en o he encoded p ocess speci ica ions.
Ex aneous load, in u n, eme ges om he acciden al complexi y o he p ocess model,
which is ypically associa ed wi h he way he model is p esen ed o he use . Finally,
Ge mane load a ises om he di icul y o in eg a e he in o ma ion ex ac ed om he
model wi h ones’ men al schema in o de o de elop an o e a ching unde s anding o
he p ocess.
Typically, eade s a emp o iden i y, in he ma e ial a hand, ea u es ha signal asso-
cia ion wi h hei p e-exis ing men al schemas (An 2013). This iden i ica ion acili a es
schema ac i a ion, de ined as he p ocess h ough which indi iduals e ie e and apply
p e iously acqui ed knowledge s uc u es om memo y o in e p e new in o ma ion
(An 2013). Fo ins ance, when eading a p ocess model, hey may associa e an XOR
ga eway wi h hei p e-exis ing schema ela ed o mu ual exclusion. Howe e , when
ambigui ies a e p esen , eade s may s uggle o iden i y and ac i a e he app op ia e
schema because mul iple in e p e a ions become plausible. As depic ed in Fig. 2 ( igh ),
an XOR ga eway wi h unlabeled ou going edges c ea es ambigui y. Reade s migh ei he
ely on con ex ual in o ma ion o selec he mos sui able b anch, hence associa ing he
unde speci ied XOR ga eway wi h hei mu ual-exclusion schema. Al e na i ely, hey
Page 8 o 38Abbad-Andaloussi e al. P ocess Science (2025) 2:19
can in e p e he XOR ga eway as an AND ga eway, whe e unlabeled ou going edges
make sense, hence associa ing he ga eway wi h hei p e-exis ing schema on concu -
ency. This sugges s ha ambigui ies can challenge he in eg a ion o new in o ma ion
wi h exis ing schemas, consequen ly inc easing ge mane cogni i e load. This p oposi-
ion is suppo ed by Campbell’s wo k(Campbell 1988). In his li e a u e e iew, Camp-
bell iden i ied se e al cha ac e is ics o complex asks ha impose high men al demands
(i.e., cogni i e load) on eade s. One such cha ac e is ic is he p esence o unce ain o
con lic ing in o ma ion wi hin an a i ac . Ano he is he a ailabili y o mul iple iable
solu ions o he ask. This, in u n, inc eases eade s’ cogni i e load as hey mus look o
addi ional in o ma ion, assess al e na i es, and make decisions among simila ly iable
op ions.
Exis ing li e a u e p oposed se e al measu es aimed a es ima ing eade s’ cogni i e
load(Figl 2017; Figl e al. 2024; Figl And Laue 2015; Abbad-Andaloussi e al. 2023a;
Wang e al. 2022; Sch eibe e al. 2024; Zugal 2013; Webe e al. 2021). Among hem,
sel - epo ed measu es ely on indi iduals’ own assessmen o pe cei ed di icul y(Figl
2017; Figl And Laue 2015; Abbad-Andaloussi e al. 2023a; Wang e al. 2022; Sch eibe
e al. 2024; Zugal 2013), which o example can be a ed by eade s using a 5-poin Like
scale ( anging om 0: “ e y easy” o 4: “ e y di icul ”) a e comple ing a ask(Sch eibe
e al. 2024; Zugal 2013). Beside sel -assessmen , which can be subjec i e, eye- acking
measu es can p o ide eliable insigh s in o cogni i e load(Holmq is e al. 2011; Figl
e al. 2024; Abbad-Andaloussi e al. 2023a; Sch eibe e al. 2024). Eye- acking enables
he analysis o ixa ion cha ac e is ics o eade s (i.e., he amoun o ime he eye emains
s a iona y a a speci ic posi ion o he s imulus, e.g., a p ocess model(Holmq is e al.
2011)) o es ima e hei cogni i e load(Holmq is e al. 2011). The use o ixa ion ea-
u es as indica o s o cogni i e load is g ounded in he eye-mind hypo hesis(Holmq is
e al. 2011), which pos ula es ha he mind p ocesses he con en cu en ly ixa ed by
he eyes. Glöckne and He bold (2008) ex ended his heo y by sugges ing ha ixa ions
las ing
≥250ms
signi y men al p ocessing, which can be linked o cogni i e load.
While he eye–mind hypo hesis was challenged, pa icula ly because o i s limi a-
ions in accoun ing o asynch onies be ween a en ion and eye mo emen s, as well as
i s inabili y o cap u e he in luence o pe iphe al ision on cogni i e p ocessing(Hol-
mq is e al. 2011), i con inues o se e as a cen al assump ion in esea ch explo ing
how use s engage wi h so wa e a i ac s. This is e iden in a wide ange o s udies ha
adop he hypo hesis o examine a ious cogni i e and beha io al aspec s o use s’ in e -
ac ions wi h so wa e a i ac s (o e iew inSha a i e al. (2015); Webe e al. (2021);
Ba is aDua e e al. (2021)).
To e alua e he e ec o ambigui ies on cogni i e load, we adop bo h he sel -assess-
men o pe cei ed di icul y measu e and he eye- acking measu es e lec ing cogni i e
load. Fo he o me , we use a 5-poin Like scale ques ionnai e o cap u e pe cei ed
di icul y. Fo he la e , we use he mean numbe o ixa ions las ing
≥250ms
as an
indica o o cogni i e load. We hypo hesize ha bo h measu es will exhibi signi ican
inc eases when eade s a e con on ed wi h ambiguous p ocess models, e lec ing he
addi ional men al e o equi ed o esol e he ambigui ies.
Comp ehension. In he cogni i e science li e a u e, e y high le els o cogni i e
load can impai eade s’ pe o mance, pa icula ly in e ms o ask accu acy (Vel man
And Jansen 2005) and esponse ime (Chen e al. 2016). Mo ing o he p ocess model
Page 15 o 38Abbad-Andaloussi e al. P ocess Science (2025) 2:19
ask allowed pa icipan s o ge accus omed wi h he in e ace and ask o ma wi hou
in luencing he s udy esul s, as he ou comes o his ask we e excluded om he sub-
sequen da a analysis. To add ess po en ial lea ning and a igue e ec s, he p esen a ion
o de o he emaining asks was andomized. This andomiza ion ensu ed ha lea ning
e ec s and pa icipan a igue we e dis ibu ed e enly ac oss he di e en expe imen
asks. A e comple ing each ask, pa icipan s a ed he pe cei ed di icul y o sol ing
he ask o each sub-p ocess using a Like scale. These a ings we e used o compu e
he pe cei ed di icul y measu e o cogni i e load (c . S udy designsec ion). Pa ici-
pan s we e also asked e ospec i ely o s a e whe he hey encoun e ed any issues while
sol ing he ask and o desc ibe how hey o e came hem. This was used o in o m us
whe he hey had no iced he ambigui ies in he p ocess models and how hey esol ed
hem. We used his in o ma ion in he da a analysis o conside only asks whe e ambi-
gui y was no iced and o quali a i ely analyze pa icipan s’ beha io when esol ing
ambigui ies (c . Da a collec ion and analysissec ion).
Da a collec ion and analysis
The da a was ga he ed using EyeMind(Abbad-Andaloussi e al. 2023b), which is an
eye- acking ool designed o cap u ing eye- acking da a on p ocess models displayed
wi hin an in e ac i e edi o . This ool allows use s o seamlessly na iga e a ious pa s o
a model and explo e i s sub-p ocesses. A signi ican ad an age o EyeMind is i s abili y o
suppo dynamic eye- acking s imuli, enabling use s o eely b owse h ough di e en
iews, sc oll, and zoom in a ious pa s o he s imulus(Abbad-Andaloussi e al. 2023b;
Holmq is e al. 2011). Conduc ing expe imen s wi h dynamic s imuli is ecognized as
a complex and ime-in ensi e p ocess(Abbad-Andaloussi e al. 2023b; Holmq is e al.
2011). Consequen ly, esea che s o en ely on s a ic s imuli, using small, non-in e ac i e
p ocess models p esen ed as images. While simple o implemen , his app oach does
no accu a ely ep esen he complexi y and usabili y o eal-wo ld p ocess models, lim-
i ing he ecological alidi y (i.e., abili y o gene alize indings)(Abbad-Andaloussi e al.
2023b; Holmq is e al. 2011). To o e come he limi a ions o elying on s a ic s imuli
and be e e lec he size and complexi y o eal-wo ld models, we selec ed EyeMind. To
he bes o ou knowledge, i is he only ool capable o p o iding his unc ionali y o
p ocess models.
A e da a collec ion, we selec ed he ials3 in which pa icipan s co ec ly iden i-
ied he ambigui ies embedded in he models. This selec ion was based on he assump-
ion ha non-iden i ied ambigui ies would ha e no e ec on pa icipan s’ cogni i e and
beha io al esponses. Fu he mo e, his selec ion was based on ou goal o in es iga e
he e ec s o ecognized ambigui y as he enable o he cogni i e and beha io al aspec s
ha we se o s udy. This is because we could no measu e he e ec s o wha was no
pe cei ed as ambiguous and did no esul in such e ec s. Ambigui ies we e iden i ied by
he pa icipan s in 342 ials ou o 528 ials (de i ed om 44 pa icipan s comple ing
12 asks, excluding he amilia iza ion ask). To de e mine hese ials, pa icipan s who
s a ed ha ing encoun e ed issues in sol ing a ask we e asked o na iga e h ough he
sub-p ocess models o show us he speci ic agmen esponsible o hei di icul ies and
explain us he espec i e easons. We selec ed he ials in which he indica ed agmen
3 A ial e e s o an ins ance o a pa icipan pe o ming a ask
Page 16 o 38Abbad-Andaloussi e al. P ocess Science (2025) 2:19
ma ches wi h he ambigui y and he explana ion could be associa ed wi h he ambigu-
i y. In a limi ed numbe o cases, pa icipan s no ed encoun e ing addi ional issues ha
we e no ela ed o ambiguous agmen s, such as inding an xo -ga eway condi ion no
immedia ely clea a i s sigh . In hese cases, he e we e no unin ended ambigui ies
disco e ed by he pa icipan s, bu only issues beyond ambigui y. Since hese issues we e
no ela ed o ambigui y and ou s udy ocuses on he e ec s o ambigui y, we did no
include he espec i e da a in ou analysis. To add ess RQ1 (c . In oduc ionsec ion), we
calcula ed he measu es co esponding o he cons uc s on he dependen a iables side
o he esea ch model illus a ed in Fig.3. We conduc ed hese calcula ions a he le el
o each sub-p ocess. Since pa icipan s we e assigned h ee asks pe ambigui y ype (c .
S udy designsec ion), h ee da a poin s we e collec ed pe ac o le el and pa icipan .
To mi iga e in e -dependencies among da a poin s, we compu ed he mean alue o each
measu e a each ac o le el o each pa icipan . We compu ed bo h desc ip i e and
in e en ial s a is ics, as epo ed in Table3. Desc ip i e s a is ics acili a ed pai wise
wi hin-subjec compa isons be ween he mean alues o each measu e ac oss he ac o
le els: ambigui y and no ambigui y o each ambigui y ype. To de e mine he s a is ical
signi icance o he obse ed di e ences, we used he Wilcoxon Signed-Rank in e en ial
Table 3 Desc ip i e and in e en ial s a is ics
A. H. Measu e Desc ip i e In e en ial
No Ambigui y Ambigui y p- alue
Mean Mean
Layou A. H1 Cogni i e Load
Pe cei ed Di icul y 1.038 3.331
<
.001
Fixa ions
>= 250
ms 32.754 74.508 <.001
H2 Comp ehension
Comp ehension E iciency 27198.497 51667.651
<
.001
H3 Visual Associa ions
AOI Run Coun 21.159 46.357
<
.001
Seman ic A. H1 Cogni i e Load
Pe cei ed Di icul y 0.893 2.240
<
.001
Fixa ions
>= 250
ms 28.354 41.547
<
.001
H2 Comp ehension
Comp ehension E iciency 25809.034 39514.319
<
.001
H3 Visual Associa ions
AOI Run Coun 20.067 34.573
<
.001
Syn ac ic A. H1 Cogni i e Load
Pe cei ed Di icul y 0.973 2.217
<
.001
Fixa ions
>= 250
ms 25.212 34.308
<
.001
H2 Comp ehension
Comp ehension E iciency 26414.001 32977.135
<
.001
H3 Visual Associa ions
AOI Run Coun 19.151 27.489
<
.001
Lexical A. H1 Cogni i e Load
Pe cei ed Di icul y 0.786 1.976
<
.001
Fixa ions
>= 250
ms 21.610 21.463 0.634
H2 Comp ehension
Comp ehension E iciency 16258.680 19436.565 0.019
H3 Visual Associa ions
AOI Run Coun 18.086 22.610 0.008
Comp ehension E iciency uni : milliseconds. No e:
p<
0.05 in o ms ha he pai wise di e ence o means be ween he no
ambigui y and ambigui y le els is signi ican
Page 17 o 38Abbad-Andaloussi e al. P ocess Science (2025) 2:19
es (Wilcoxon 1945), which is sui able o pai wise wi hin-subjec compa isons and
does no impose assump ions on he no mal dis ibu ion o he da a.
Besides alida ing he impac o ambigui y, we conduc ed a quali a i e explo a o y
analysis, using he pa icipan s’ eye acking and hink-aloud da a (c . o e iew in Fig.5).
He ein, he aim was o in es iga e hei beha io when esol ing ambigui y (RQ2, c .
In oduc ionsec ion). Speci ically, we used a quali a i e coding app oach om g ounded
heo y (Cha maz 2006; B yan and Cha maz 2019) o iden i y common beha io s
adop ed by he pa icipan s. The codes desc ibing isual beha io pa e ns we e de el-
oped based on p elimina y obse a ions o he ideo eco dings (o app oxima ely 40
minu es each) o a sample o 23% o he s udy pa icipan s. An ini ial coding(B yan
and Cha maz 2019) phase was conduc ed on hese ideos o iden i y eme ging isual
beha io s. He ein, we assigned desc ip i e labels o ideo segmen s, e lec ing he pa -
icipan s’ beha io al ac ions (e.g., ixa ing on a speci ic model elemen o a long pe iod
o ime o shi ing apidly ac oss he model), wi hou making assump ions abou he
meaning o hese ac ions. This phase allowed us o emain open o all possible in e p e-
a ions and o cap u e he ull ange o beha io s pa icipan s exhibi ed when engaging
wi h ambiguous p ocess models. A e wa d, we gene a ed AOIs-o de o e ime plo s o
all he ials in which ambigui ies we e no iced.
Figu e6 depic s an example o an AOIs-o de o e ime plo . This plo isualizes how
a pa icipan ’s a en ion shi s ac oss di e en elemen s ( ep esen ed as AOIs) o a p o-
cess model while pe o ming a ask. The ba s on he X-axis show he AOIs isi ed by
he pa icipan , so ed by hei o de o occu ence in ime. This empo al pe spec i e
e lec s he sequence o p ocess model elemen s ha he pa icipan isi ed du ing he
ask. The black AOIs e e o he ambiguous elemen s o he p ocess model, while he
emaining AOIs e e o he non-ambiguous elemen s o he model. This la e se o
AOIs ha e di e en colo s, allowing, in u n, o iden i y hose ha we e isi ed se e al
imes by he pa icipan . This colo coding helps isualize how he pa icipan shi ed
Fig. 5 O e iew o he quali a i e explo a o y analysis
Page 18 o 38Abbad-Andaloussi e al. P ocess Science (2025) 2:19
Fig. 6 Example o a simpli ied AOIs-o de o e ime plo
Page 19 o 38Abbad-Andaloussi e al. P ocess Science (2025) 2:19
Fig. 7 P ocess maps compa ing he isual associa ions o a pa icipan (SP7) when eading a sub-p ocess wi hou (le ) and wi h ( igh ) a seman ic ambigui y. A highe esolu ion o his igu e is a ailable in he
online appendix. The ci cle wi h a do inside deno es he p ocess s a , he double ci cle wi h a squa e inside deno es he p ocess end. Rec angles e e o isi s o he di e en p ocess model ac i i ies; edges
e e o he ansi ions o isi ing one ac i i y om ano he . The colo scale o he ec angles e e s o he absolu e isi equency o an ac i i y; he hickness and labels on he edges e e o he absolu e
ansi ion equency, esp. he numbe o ansi ions be ween each pai o ac i i ies
Page 20 o 38Abbad-Andaloussi e al. P ocess Science (2025) 2:19
be ween ambiguous and non-ambiguous elemen s o he model. The X-axis is addi ion-
ally deco a ed wi h ho izon al blocks, colo ed in blue, ed, o g een o in e espec i ely
whe he he isi ed AOI belongs o he i s , second, o hi d sub-p ocess o ou model.
As men ioned in S udy designsec ion, he second sub-p ocess was he one inco po-
a ing an ambigui y. On he Y-axis, he heigh o he ba s de ines he du a ion o each
AOI isi (in milliseconds). Mo eo e , we ha e se 250ms as a h eshold on he Y-axis,
allowing o dis inguish he isi s e lec ing men al p ocessing (Glöckne And He bold
2008). P ojec ing isi du a ion on o he Y-axis acili a es he compa ison o how long
pa icipan s engaged wi h ambiguous e sus non-ambiguous model elemen s, and helps
o iden i y isi s likely associa ed wi h men al p ocessing. All in all, he AOIs-o de o e
ime plo enables he analysis o pa icipan s’ isual beha io by e ealing he sequence
in which p ocess model elemen s a e isi ed ( ia he X-axis), he du a ion o each isi o
a speci ic elemen ( ia he Y-axis), and which isi s a e likely o in ol e men al p ocess-
ing (Figs. 7 and 8).
Following he gene a ion o he AOIs-o de o e ime plo s, we conduc ed ocused
coding(B yan and Cha maz 2019) on all hese plo s h ough he obse a ion and cod-
ing o pa icipan s’ beha io al pa e ns, which we e aligned wi h hose obse ed in he
ideos. In his phase, we sys ema ically ca ego ized ecu ing beha io s by g ouping
simila ini ial codes in o mo e abs ac and concep ually meaning ul codes. Fo ins ance,
shi ing sho isi s all o e he p ocess model elemen s wi hou ocusing on speci ic ele-
men s was coded as Sweep, while shi ing long isi s ac oss a la ge a ea o he model
co e ing se e al p ocess elemen s was coded as Explo e. When long isi s occu ed
be ween a limi ed numbe o speci ic p ocess elemen s, we assigned he code Ta ge .
Back-and- o h ansi ions be ween he ambigui y and ano he speci ic p ocess elemen
Fig. 8 Exce p s om ambiguous p ocess models showing he ambiguous model agmen s. The comple e mod-
els a e a ailable in he online appendix
Page 21 o 38Abbad-Andaloussi e al. P ocess Science (2025) 2:19
Fig. 9 AOIs-o de o e ime plo showing he Sweep isual beha io o pa icipan SP7 while ying o esol e an ambigui y, isible in he sequence o sho isi s (
<250ms
) o mul iple p ocess elemen s
shown in di e en colo s be ween he i s and he las isi s o he ambigui y (shown in black)
Page 22 o 38Abbad-Andaloussi e al. P ocess Science (2025) 2:19
we e cap u ed unde he code Bounce, whe eas mos ly unin e up ed sequences o long
isi s o he ambiguous p ocess elemen s hemsel es we e labeled Ho e . These ocused
codes allowed us o in e he common s a egies pa icipan s used when a emp ing
o esol e ambigui ies. To ensu e he obus ness o ou codes, he coding was done by
one au ho and subsequen ly e iewed by ano he au ho ; in case o disag eemen s, he
au ho s discussed he code un il con e gence. This coding s a egy (aligned wi h p io
li e a u e, e.g.,Jaya aman e al. (2024)) was in en ionally chosen o e independen cod-
ing conside ing he inhe en complexi y o he AOIs-o de o e ime plo s, whose isual
analysis equi es collabo a i e e o , i e a i e back-and- o h discussions, and ca e ul
inspec ion wi h mul iple pai s o eyes o disce n he beha io al pa e ns (c . Figs.9, 10,
11, 12, 13 and 14). Mo eo e , o ensu e he alidi y o ou pa e ns iden i ica ion, we
inco po a ed as supplemen a y e idence he e bal da a ob ained by asking e ospec-
i ely a he end o each ask which s a egies he pa icipan s employed o o e come any
di icul ies encoun e ed. To do his, we lis ened o he audio ack o he ideo eco d-
ings o he da a collec ion sessions, imes amping he pe iods whe e ele an answe s
we e p o ided, and aking no e o he answe s. This allowed us o iangula e he coded
beha io s wi h he e bal insigh s p o ided by he pa icipan s. Finally, we applied axial
coding o es ablish ela ionships be ween ou codes(B yan and Cha maz 2019). In
doing so, we o ganized he ocused codes in o a scale o a en ion ocus, a anging he
iden i ied beha io s acco ding o hei associa ed le els o a en ion, anging om an-
dom a en ion (e.g., Sweep) o ocused a en ion (e.g., Ho e ). This las coding phase was
conduc ed collabo a i ely be ween he au ho s.
Da a a ailabili y and ep oducibili y
To ensu e anspa ency, ep oducibili y, and eplicabili y, we p o ide an online appen-
dix, which includes:
• he ull se o p ocess models used in he asks o he s udy;
•a comp ehensi e epo de ailing he complexi y me ics o he p ocess models;
•lis o he expe imen asks wi h he speci ic guideline iola ions and he ambigui ies
hese iola ions in oduced;
•demog aphic in o ma ion o he pa icipan s;
•a highe - esolu ion e sion o Figs.4, 7 and 8;
•p ocess maps illus a ing he pa icipan s’ isual associa ions;
•all AOIs-o de o e ime plo s de i ed om he eye- acking da a;
• he quali a i e codes wi h he beha io s obse ed;
• he Py hon no ebook used o da a analysis, along wi h he co esponding esul s.
The online appendix can be accessed a h ps://doi.o g/10.5281/zenodo.16738478.
Findings
In his sec ion, we p esen ou indings o ganized by esea ch ques ion.
RQ1. How do di e en ambigui ies a ec model eade s’ cogni i e load, comp e-
hension and isual associa ions? Cogni i e load. Cogni i e load was measu ed using
pe cei ed di icul y and mean numbe o ixa ions wi h du a ion
≥250ms
, collec ed a
Page 23 o 38Abbad-Andaloussi e al. P ocess Science (2025) 2:19
Fig. 10 AOIs-o de o e ime plo showing he Explo e isual beha io o pa icipan SP6 while ying o esol e an ambigui y, isible in he sequence o long isi s (
≥250ms
) o mul iple p ocess elemen s
shown in di e en colo s be ween he i s and he las isi s o he ambigui y (shown in black)
Page 24 o 38Abbad-Andaloussi e al. P ocess Science (2025) 2:19
Fig. 11 AOIs-o de o e ime plo showing he Ta ge isual beha io o pa icipan KP23 while ying o esol e an ambigui y, isible in he sequence o long isi s (
≥250ms
) o he ew p ocess elemen s
shown in yellow, g een and iole be ween he i s and he las isi s o he ambigui y (shown in black)
Page 31 o 38Abbad-Andaloussi e al. P ocess Science (2025) 2:19
SP3, bu hen shi ing o a Ta ge beha io ( ocused on he connec ed ac i i ies) helped
o igu e some cues o make an assump ion on he in e p e a ion o he ambiguous
model agmen ha led o he comple ion o he ask. The analysis o he en i e da ase
sugges s ha model eade s do no always employ a single s a egy, bu apply mul iple
s a egies while a emp ing o esol e ambigui y. Speci ically, we obse ed Sweep beha -
io in combina ion wi h o he beha io s in 79 ials, Explo e beha io in combina ion
wi h o he beha io s in 60 ials, Ta ge beha io in combina ion wi h o he beha io s in
154 ials, Bounce beha io in combina ion wi h o he beha io s in 42 ials, and Ho e
beha io in combina ion wi h o he beha io s in 122 ials. Besides, in 4 ou o ou 342
ials, he pa icipan s’ beha io was unclea in he plo s due o quali y issues in he
eco ded da a.
Mo eo e , o any gi en ambigui y, we did no consis en ly obse e he same isual
beha io pa e n ac oss he pa icipan s o ou s udy. We also did no obse e a consis-
en adop ion o he same pa e n ac oss he ambigui ies o any gi en pa icipan .
Discussion
Tes ing he e ec s o ambigui y
Ou indings wi h espec o RQ1 (c . In oduc ionsec ion), e eal ha layou , seman-
ic, and syn ac ic ambigui ies, which a ec he p ocess con ol low, exe a signi ican
impac on cogni i e aspec s. These ambigui ies inc ease model eade s’ cogni i e load,
educe hei comp ehension, and esul in a signi ican amoun o isual associa ions,
indica ing heigh ened cogni i e in eg a ion e o (Be a e al. 2019). In con as , lexi-
cal ambigui ies, which a ec he labels, exe a less p onounced e ec . While we could
obse e e ec s in e ms o lowe comp ehension and highe isual associa ions, no clea
e ec s could be obse ed in e ms o cogni i e load. Speci ically, he mean numbe o
ixa ions wi h du a ion
≥250ms
, which a e associa ed wi h men al p ocessing (c . Cog-
ni i e load, comp ehension, and isual associa ionssec ion), did no signi ican ly di e
be ween models wi h and wi hou lexical ambigui ies. A possible explana ion o his ac
is ha ambigui ies in he lexicon may no impose as much cogni i e load as ambigui ies
in he con ol low.
Explo ing ambigui y esol ing beha io s
Ou indings wi h espec o RQ2 (c . In oduc ionsec ion) sugges ha eade s exhibi
di e en beha io s when esol ing ambigui ies. Speci ically, we obse ed and de ined o
he i s ime i e dis inc pa e ns o isual beha io induced by ambigui y in p ocess
models, namely Sweep, Explo e, Ta ge , Bounce, and Ho e . These beha io s sugges di -
e en s a egies o a emp ing o in eg a e in o ma ion o esol e ambigui ies.
Building upon he indings p esen ed in Findings sec ion and applying axial cod-
ing(B yan and Cha maz 2019) (c . Da a collec ion and analysissec ion and he quali-
a i e explo a o y analysis p ocess illus a ed in Fig.5), we examined he ela ionships
be ween he iden i ied isual beha io pa e ns and he le els o a en ion exhibi ed by
he pa icipan s when esol ing ambigui y. He ein, we sugges ha he obse ed ambi-
gui y- esol ing beha io s a e associa ed wi h a ying deg ees o a en ion ocus. Fig-
u e15 illus a es hese deg ees h ough a con inuous scale o a en ion ocus. In he
scale, Sweep e lec s he leas ocused and mos andom a en ion, in ol ing un ocused
isi s spanning he whole p ocess model. Explo e deno es a dispe sed a en ion, as i is
Page 32 o 38Abbad-Andaloussi e al. P ocess Science (2025) 2:19
Fig. 15 Visual beha io codes in a scale o a en ion. Ho e sugges s he mos ocused a en ion when acing ambigui y, while Sweep sugges s he mos andom a en ion
Page 33 o 38Abbad-Andaloussi e al. P ocess Science (2025) 2:19
cha ac e ized by a non- a ge ed explo a ion o mul iple p ocess model elemen s, sug-
ges ing ha he eade s’ a en ion is no ocused owa d speci ic p ocess elemen s bu
owa d gene al con ex ual in o ma ion. Ta ge is associa ed wi h an in e media e deg ee
o a en ion, as model eade s shi be ween a limi ed numbe o p ocess elemen s, sug-
ges ing ha he eade s’ a en ion is ocused owa d iden i ying elemen s ha can p o-
ide disambigua ion cues om a selec ed pool o candida es. Bounce indica es a highe
deg ee o ocused a en ion, as model eade s al e na e be ween he ambiguous elemen s
and a speci ic p ocess elemen likely iden i ied as ele an , sugges ing ha he eade s’
a en ion is ocused owa d applying he in o ma ion ex ac ed om his elemen o
esol e he ambigui y. Finally, Ho e is associa ed wi h he highes le el o ocused a en-
ion, as model eade s concen a e in ensely on he ambiguous p ocess elemen s while
ying o esol e he ambigui y h ough a de ailed elemen examina ion.
As epo ed in Findingssec ion, we did no obse e consis en associa ions be ween
speci ic ambigui y ypes, isual beha io pa e ns, o pa icipan s. This sugges s ha he
adop ion o an ambigui y esolu ion s a egy is a he subjec i e and a ies be ween he
conc e e ambigui ies a model eade is con on ed wi h.
Suppo ing model eade s
Gi en he demons a ed nega i e e ec s o ambigui ies, i becomes c ucial o explo e
ways o be e suppo model eade s–ei he by minimizing ambigui ies in p ocess mod-
els o by enhancing he eade s’ abili y o handle hem. Fo ins ance, au oma ed ech-
niques o label e ac o ing (c .Pi ke e al. (2015)) could help o lessen he p esence
o lexical ambigui y. Howe e , i is un ealis ic o assume ha ambigui ies can be com-
ple ely elimina ed. As a ma e o ac , ce ain ambigui ies a e in en ional, o example
designed o enable lexible in e p e a ion and execu ion o p ocesses (c .F ancesche i
e al. (2023)). Despi e his, he signi ican impac o ambigui ies on ask pe o mance
unde sco es he need o suppo beyond me ely de ec ing modeling e o s. One po en-
ial app oach is os e ing a eedback loop in which modele s and model eade s collabo-
a e o iden i y ambiguous p ocess elemen s. Al e na i ely, p ocess model analysis ools
could be de eloped o au oma ically de ec ambigui ies. While no all ambigui ies migh
be au oma ically de ec ed due o hei inhe en ly subjec i e na u e, one can en ision
ools o he au oma ic de ec ion o ce ain syn ac ic ambigui ies based on o mal de i-
ni ions, such as hose p esen ed in Ambigui y in p ocess modelssec ion o inF ance-
sche i e al. (2023), o lexical ambigui ies based on Na u al Language P ocessing. These
ools migh le e age on ology anno a ions, simila o he echniques p oposed inFan
e al. (2016).
Ano he possibili y is ha , wi h u he de elopmen , he cogni i e e ec s and beha -
io pa e ns associa ed wi h he de ec ion and esolu ion o ambigui y could be le e -
aged o de elop a new gene a ion o con ex adap i e sys ems ha deploy p e- ained
machine lea ning models (e.g.,Abbad-Andaloussi e al. (2024)) o de ec when use s
a e acing ambigui ies and guide hem in disambigua ing he model, helping hem, in
u n, o mo e om andom a en ion o ocused one. This can, o example, be achie ed
by highligh ing con ex ual in o ma ion o displaying addi ional a i ac s such as guid-
ing anno a ions o simula ions. Speci ically, o de ec ambigui ies, he machine lea ning
models can be ained using a supe ised lea ning app oach (Agga wal 2015). Fea u es
de i ed om eye- acking ixa ions, saccades, isi s o AOIs (Holmq is e al. 2011),
Page 34 o 38Abbad-Andaloussi e al. P ocess Science (2025) 2:19
and o he cogni i e and beha io al pa e ns associa ed wi h ambigui y can be collec ed
wi hin de ined ime windows and pai ed wi h labels indica ing whe he use s a e expe-
iencing ambigui y. These labels may be ob ained om hink-aloud p o ocols conduc ed
concu en ly wi h he ask. The models can be hen ained o map he ex ac ed ea-
u es o he co esponding ambigui y label a he le el o each ime window. A un- ime,
incoming da a can be bu e ed in o ime windows, each ed in o he ained machine
lea ning model o p edic he p esence o ambigui y. I an ambigui y is de ec ed, hen
con ex ual suppo can be p o ided by highligh ing ele an in o ma ion o p esen ing
addi ional a i ac s, such as guiding anno a ions o in e ac i e simula ions.
Implica ions
We iden i y implica ions o esea ch, p ac ice, and educa ion unde sco ed by ou
indings.
Implica ions o Resea ch. Rega ding esea ch, u u e empi ical s udies examining
human and cogni i e aspec s in p ocess modeling should accoun o he p esence o
ambigui y in p ocess models. Ambigui ies could unin en ionally be a con ounding ac-
o in luencing esea ch ou comes, making i essen ial o mi iga e hem in he design o
u u e expe imen s. Wi h he AOIs-o de o e ime plo s we in oduced a isualiza ion
ha emphasizes he isual beha io s adop ed by model eade s when con on ed wi h
ambigui y. This pape p esen s a i s a emp a using hese plo s owa d his goal, which
esul ed in he iden i ica ion o key pa e ns desc ibing how ambigui ies a e esol ed. By
cap u ing wo essen ial aspec s o p ocess model comp ehension, i.e., a en ion o di e -
en model pa s and ansi ions be ween hem, hese plo s o e a aluable ool o u u e
s udies. Speci ically, hey can suppo deepe explo a ion o use s’ a en ion dis ibu ion
( h ough he heigh o he ba s e e ing o he ime spen du ing each isi o a p o-
cess model elemen ) and cogni i e in eg a ion p ocesses ( h ough he o de o he ba s
e e ing o hei o de o occu ence o e ime), bo h c i ical o unde s anding p ocess
models(Be a e al. 2019).
Implica ions o P ac ice. Rega ding p ac ice, ou esul s highligh he impo ance o
minimizing ambigui ies in p ocess models as well as p o iding disambigua ion cues o
suppo model eade s. This could be achie ed by en iching p ocess models wi h sup-
plemen a y in o ma ion (c .Abbad-Andaloussi e al. (2021)) o making con ex ual cues
in he models mo e explici , since ou quali a i e s udy sugges s ha model eade s
end o look o disambigua ing cues in p ocess models when hey a e con on ed wi h
ambigui y.
Implica ions o Educa ion. Wi h ega ds o educa ion, ou esul s unde sco e he
impo ance o aising awa eness among p ocess modeling ainees abou he po en ial
ambigui ies in hei models and he impac hese ambigui ies can ha e. T aining p o-
g ams should inco po a e discussions on iden i ying and managing ambigui ies, empow-
e ing ainees wi h he skills o c ea e p ocess models ha can easily be in e p e ed.
Th ea s o alidi y
We ecognize he exis ence o po en ial h ea s o he alidi y o ou s udy. Fi s , in e -
nal alidi y may be h ea ened by he p esence o con ounding ac o s ha canno be
en i ely elimina ed. Howe e , we mi iga ed his h ea by designing a con olled expe -
imen based on a p e-de ined esea ch model (c . S udy designsec ion), ca e ully and
Page 35 o 38Abbad-Andaloussi e al. P ocess Science (2025) 2:19
sys ema ically p epa ing he expe imen ma e ials (c . S udy designsec ion), andomiz-
ing he o de in which he asks we e p esen ed o a oid lea ning and a igue e ec s (c .
Expe imen p ocedu esec ion), and adhe ing o a s ic and uni o mly applied da a col-
lec ion p o ocol du ing all sessions o da a collec ion (c . Expe imen p ocedu esec ion).
A second h ea o in e nal alidi y a ises om e ospec i ely asking pa icipan s o
s a e whe he hey encoun e ed any issues while pe o ming each ask a e i s comple-
ion. Each ask was based on a p ocess model ha was delibe a ely designed o include
an ambigui y ha hinde ed i s execu ion. The e o e, i is possible ha he pa icipan s
migh ha e an icipa ed he p esence o issues in subsequen asks due o his epea ed
ques ioning. Such an icipa ion ep esen s a po en ial con ounding ac o , as i could
in luence hei cogni i e p ocessing. To mi iga e his h ea , we ook se e al measu es.
We ca e ully a oided disclosing he goals o ou s udy o he pa icipan s, and a oided
men ioning ha hey would encoun e issues while analyzing he p ocess models. We
e ained om making he pa icipan s awa e o he concep o ambigui y in p ocess
models, ne e men ioning ambigui y. Addi ionally, we delibe a ely ph ased he ques-
ion in a gene ic manne o a oid sugges ing ha pa icipan s we e being asked o ana-
lyze p ocess models a ec ed by ambigui y (“Name any issues you encoun e ed while
answe ing he ques ion, along wi h how you o e came hem”). Finally, by andomizing
he p esen a ion o he asks (c . S udy designsec ion) we ensu ed ha he inc eased
cogni i e load esul ing om he po en ial an icipa ion o issues was e enly dis ibu ed
ac oss all asks. Ne e heless, in u u e s udies, his h ea could be u he mi iga ed
wi h he inclusion o asks whe e all he models a e ee om ambigui y and by ph as-
ing ques ions excluding nudging e ms such as issues (e.g., “Name any di icul ies you
encoun e ed while answe ing he ques ion, along wi h how you o e came hem”), he eby
minimizing he likelihood o pa icipan s an icipa ing and ac i ely seeking ou p oblems.
Wi h ega ds o ou quali a i e analysis, a po en ial h ea o in e nal alidi y de i es
om he possibili y ha he isual beha io s do no accu a ely cap u e he app oaches
employed by he pa icipan s o esol e ambigui y. To mi iga e his h ea , we iangu-
la ed he isual beha io pa e ns wi h he e ospec i e hink-aloud da a o he s udy
pa icipan s when asked abou he s a egies adop ed o o e come he di icul ies in
sol ing he gi en asks. Ano he h ea s ems om po en ial bias in ou subjec i e in e -
p e a ion o he isual beha io pa e ns, ep esen ed h ough he AOIs-o de o e ime
plo s, and he insigh s, ex ac ed om he hink-aloud da a. To mi iga e his h ea , we
e iewed he coding in o de o alida e ou in e p e a ions (c . Da a collec ion and anal-
ysissec ion). Mo eo e , we ecognize he limi a ion o measu ing only one o he wo
componen s o comp ehension e iciency, namely he esponse ime, neglec ing comp e-
hension accu acy. As explained in Cogni i e load, comp ehension, and isual associa-
ionssec ion, his is due o he impossibili y o measu e comp ehension accu acy due o
he inhe en admissibili y o he mul iple in e p e a ions o ambigui ies.
The ex e nal alidi y o ou s udy may be h ea ened by he inabili y o gene alize he
expe imen esul s due o he sample size o he pa icipan s o he modeling language
used in he p ocess models. To mi iga e hese h ea s, we ec ui ed 44 pa icipan s,
which, o he bes o ou knowledge, posi ions ou s udy among he mos ex ensi e eye-
acking s udies conduc ed in he con ex o p ocess modeling. Rega ding he speci ic
use o BPMN, we a gue ha he in es iga ed ambigui ies a e no exclusi e o BPMN.
Page 36 o 38Abbad-Andaloussi e al. P ocess Science (2025) 2:19
Such ambigui ies could also be encoun e ed in o he impe a i e modeling languages,
such as wo k low ne s and EPC models(Kelle e al. 1992).
Conclusion
Ambigui ies in a p ocess model esul in mul iple po en ial in e p e a ions o he p o-
cess. By le e aging eye- acking, we explo ed how hese ambigui ies in luence model
eade s du ing a ious model comp ehension asks. Ou indings e eal a signi ican
in luence on cogni i e load, comp ehension, and isual associa ions. Addi ionally, ou
esul s indica e he adop ion o speci ic isual beha io pa e ns when sol ing ambigui y
(Sweep, Explo e, Ta ge , Bounce, and Ho e ), which can be associa ed wi h di e en le -
els o a en ion anging om andom o ocused a en ion.
The nega i e e ec s o ambigui y highligh he impo ance o p o iding adequa e
aining and exe cising cau ion o minimize ambigui ies in p ocess models. Addi ionally,
hey unde sco e he need o designing no el au oma ed ools o assis model eade s in
iden i ying ambigui ies and po en ially esol ing hem. In u u e wo k, i is wo hwhile
o in es iga e he ole played by BPM expe ise in he de ec ion o ambigui ies, includ-
ing in ela ion o di e se indus ial backg ounds. Fu he mo e, i is also wo hwhile o
speci ically in es iga e he cogni i e and beha io al impac o in en ional ambigui ies.
Mo eo e , gi en he ichness o ou eye- acking da a, new machine lea ning models can
be de eloped o au oma ically de ec when eade s a e challenged wi h ambigui y, and
guide hem esol ing i . These models will p o ide he ounda ion o a new gene a ion
o ambigui y-awa e con ex -adap i e sys ems.
Acknowledgemen s
We g a e ully hank John K ogs ie o his aluable inpu and insigh ul discussions.
Au ho s’ con ibu ions
A.A.-A., M.F., H.A.L., and B.W. con ibu ed o he wo k’s concep ion. A.A.-A., M.F., and C.S. conduc ed he expe imen .
A.A.-A. and M.F. analyzed he expe imen esul s and w o e he manusc ip . All au ho s con ibu ed o e iewing he
manusc ip .
Funding
This wo k has ecei ed unding om he Swiss Na ional Science Founda ion unde G an No. IZSTZ0_208497
(P oAmbi Ion p ojec ). Amine Abbad-Andaloussi is suppo ed by he In e na ional Pos doc o al Fellowship G an unde
G an No. 1031574 om he Uni e si y o S .Gallen, Swi ze land. Hugo A. López is suppo ed by esea ch g an “Cen e o
Digi al CompliancE (DICE)” (G an No. VIL57420) om VILLUM FONDEN.
Da a a ailabili y
An online appendix con aining supplemen a y ma e ial as de ailed in Da a a ailabili y and ep oducibili ysec ion is
a ailable a h ps://doi.o g/10.5281/zenodo.16738478.
Decla a ions
Compe ing in e es s
The au ho s decla e no compe ing in e es s.
Consen o pa icipa e
The s udy pa icipan s p o ided hei in o med consen o con ibu e wi h hei esponses and o ha e hei eye- acking
and ideo/audio da a eco ded. Pa icipa ion was olun a y, and all pa icipan s we e in o med o hei igh o wi hd aw
a any ime wi hou penal y. The collec ed da a was anonymized o ensu e he p i acy o he pa icipan s and used
solely o he pu poses o his s udy.
Recei ed: 30 Decembe 2024 / Accep ed: 25 July 2025
Re e ences
Abbad-Andaloussi A, Bu a in A, Slaa s T e al (2023a) Complexi y in decla a i e p ocess models: me ics and mul i-modal
assessmen o cogni i e load. Expe Sys Appl 233:120924
Page 37 o 38Abbad-Andaloussi e al. P ocess Science (2025) 2:19
Abbad-Andaloussi A, Lübke D, Webe B (2023b) Conduc ing eye- acking s udies on la ge and in e ac i e p ocess models using
eyemind. So wa eX 24:101564
Abbad-Andaloussi A, Sch eibe C, Webe B (2024) Using eye- acking o de ec sea ch and in e ence du ing p ocess model
comp ehension. In: P oceedings o he 30 h In e na ional Con e ence on Coope a i e In o ma ion Sys ems (in p ess)
Abbad-Andaloussi A, Ze ba o F, Bu a in A e al (2021) Explo ing how use s engage wi h hyb id p ocess a i ac s based on
decla a i e p ocess models: a beha io al analysis based on eye- acking and hink-aloud. So w Sys Model 20:1437–1464
Agga wal CC (2015) Da a mining: he ex book. Sp inge
Amna AR, Poels G (2022) Ambigui y in use s o ies: a sys ema ic li e a u e e iew. In So w Technol 145:106824
An S (2013) Schema heo y in eading. Theo y P ac Lang S ud (TPLS) 3(1)
Ba is a Dua e R, Sil a da Sil ei a D, de Albuque que BV e al (2021) A sys ema ic li e a u e e iew on he usage o eye- acking
in unde s anding p ocess models. Bus P ocess Manag J 27(1):346–367
Be a P, So e P, Pa sons J (2019) Using eye acking o expose cogni i e p ocesses in unde s anding concep ual models. MIS Q
43(4):1105–1126
Be y DM, Kams ies E (2004) Ambigui y in equi emen s speci ica ion. In: Pe spec i es on so wa e equi emen s. Sp inge , pp
7–44
B yan A, Cha maz K (2019) The SAGE Handbook o Cu en De elopmen s in G ounded Theo y. SAGE Publica ions
Caldie a VRBG, Rombach HD (1994) The goal ques ion me ic app oach. Encycl So w Eng 528–532
Campbell DJ (1988) Task complexi y: a e iew and analysis. Acad Manag Re 13(1):40–52
Cha maz K (2006) Cons uc ing g ounded heo y: A p ac ical guide h ough quali a i e analysis. Sage
Chen F, Zhou J, Wang Y e al (2016) Robus mul imodal cogni i e load measu emen . Sp inge
Co adini F, Fe a i A, Fo na i F e al (2018) A guidelines amewo k o unde s andable bpmn models. Da a Knowl Eng
113:129–154
Dijkman RM, Dumas M, Ouyang C (2008) Seman ics and analysis o business p ocess models in bpmn. In So w Technol
50(12):1281–1294
Dju ica D, Kumme TF, Mendling J e al (2023) In es iga ing he impac o ep esen a ion ea u es on decision model comp e-
hension. Eu J In Sys 1–25
Dumas M, LaRosa M, Mendling J e al (2018) P ocess-awa e in o ma ion sys ems. Fundam Bus P ocess Manag 341–369
Fan S, Hua Z, S o ey VC e al (2016) A p ocess on ology based app oach o easing seman ic ambigui y in business p ocess
modeling. Da a Knowl Eng 102:57–77
Figl K (2017) Comp ehension o p ocedu al isual business p ocess models. Bus In Sys Eng 59(1):41–67
Figl K, Laue R (2015) In luence ac o s o local comp ehensibili y o p ocess models. In J Hum-Compu S ud 82:96–110
Figl K, Mendling J, S embeck M (2013) The in luence o no a ional de iciencies on p ocess model comp ehension. J Assoc In
Sys 14(6):1
Figl K, So e P, Webe B (2024) Guiding a en ion in low-based concep ual models h ough consis en low and pa e n is-
ibili y. Decis Suppo Sys 185:114292
Fındık-Coşkunçay D, Çakı MP (2022) An in es iga ion o he ela ionship be ween join isual a en ion and p oduc quali y in
collabo a i e business p ocess modeling: a dual eye- acking s udy. So w Sys Model 21(6):2429–2460
Fo uny J, Pay a ó L (2024) Ambigui y in linguis ics 1. S udia Linguis ica 78(1):1–7
F ancesche i M, Abbad-Andaloussi A, Sch eibe C e al (2024) Explo ing he cogni i e e ec s o ambigui y in p ocess models.
In: In e na ional Con e ence on Business P ocess Managemen . Sp inge , pp 493–510
F ancesche i M, Seige R, López HA e al (2023) A cha ac e isa ion o ambigui y in bpm. In: In e na ional Con e ence on Con-
cep ual Modeling. Sp inge , pp 277–295
Glöckne A, He bold AK (2008) In o ma ion p ocessing in decisions unde isk: E idence o compensa o y s a egies based on
au oma ic p ocesses. MPI collec i e goods p ep in (2008/42)
Gulden J, Bu a in A, Abbad-Andaloussi A e al (2020) F om analy ical pu poses o da a isualiza ions: a decision p ocess guided
by a concep ual amewo k and eye acking. So w Sys Model 19:531–554
Haisjackl C, Pingge a J, So e P e al (2015) Iden i ying quali y issues in bpmn models: an explo a o y s udy. In: In . Wo kshop on
Business P ocess Modeling, De elopmen and Suppo . Sp inge , pp 217–230
Heggse M, K ogs ie J, Wesenbe g H (2015) The in luence o syn ac ic quali y o en e p ise p ocess models on model comp e-
hension. In: CAiSE Fo um. pp 89–96
Hildeb and TT (2018) Ecoknow: E ec i e, co-c ea ed & complian adap i e case managemen o knowledge wo ke s. In: 2018
IEEE 22nd In e na ional En e p ise Dis ibu ed Objec Compu ing Wo kshop (EDOCW). IEEE, pp 9–11
Holmq is K, Nys öm M, Ande sson R e al (2011) Eye T acking: A comp ehensi e guide o me hods and measu es. OUP Ox o d
Jaya aman S, Phillips EK, Chu ch D e al (2024) P i acy and u ili y pe cep ions o social obo s in heal hca e. Compu Hum Beha
A i Hum 2(1):100039
Kelle G, Nü gens M, Schee AW (1992) Seman ische P ozessmodellie ung au de G undlage “e eignisges eue e P ozess-
ke en (EPK)”. Ins i u ü Wi scha sin o ma ik Saa b ücken: Ve ö en lichungen des Ins i u s ü Wi scha sin o ma ik, Ins .
ü Wi scha sin o ma ik
Kindle E (2009) Model-based so wa e enginee ing and p ocess-awa e in o ma ion sys ems. T ansac ions on Pe i Ne s and
O he Models o Concu ency II: Special Issue on Concu ency in P ocess-Awa e In o ma ion Sys ems 27–45
K ogs ie J (2012) Model-based de elopmen and e olu ion o in o ma ion sys ems: A Quali y App oach. Sp inge Science &
Business Media
K ogs ie J (2016) Quali y o business p ocess models. In: Quali y in business p ocess modeling. Sp inge , pp 53–102
Leopold H, Smi no S, Mendling J (2010) Re ac o ing o p ocess model ac i i y labels. In: 15 h In . Con . on Applica ions o
Na u al Language o In o ma ion Sys ems. P oceedings 15. Sp inge , pp 268–276
Lindland OI, Sind e G, Sol be g A (1994) Unde s anding quali y in concep ual modeling. IEEE So w 11(2):42–49
Mendling J, Recke J, Reije s HA (2010) On he usage o labels and icons in business p ocess modeling. In J In Sys Model Des
(IJISMD) 1(2):40–58
Mendling J, Reije s HA, Recke J (2010) Ac i i y labeling in p ocess modeling: empi ical insigh s and ecommenda ions. In Sys
35(4):467–482
Mendling J, Reije s HA, an de Aals WM (2010) Se en p ocess modeling guidelines (7pmg). In So w Technol 52(2):127–136
Page 38 o 38Abbad-Andaloussi e al. P ocess Science (2025) 2:19
Mendling J, S embeck M (2008) In luence ac o s o unde s anding business p ocess models. In: In e na ional Con e ence on
Business In o ma ion Sys ems. Sp inge , pp 142–153
Mendling J, S embeck M, Recke J (2012) Fac o s o p ocess model comp ehension– indings om a se ies o expe imen s.
Decis Suppo Sys 53(1):195–206
OMG (2014) Business P ocess Model and No a ion (BPMN), Ve sion 2.0.2. h ps://www.omg.o g/spec/BPMN/2.0.2. Accessed 19
Dec 2024
Oppl S (2016) A icula ion o wo k p ocess models o o ganiza ional alignmen and in o med in o ma ion sys em design. In
Manag 53(5):591–608
Pe e M (1995) Why looking isn’ always seeing: eade ship skills and g aphical p og amming. Commun ACM 38(6):33–44
Pe usel R, Mendling J (2013) Eye- acking he ac o s o p ocess model comp ehension asks. In: CAiSE 2013, P oceedings 25.
Sp inge , pp 224–239
Pe usel R, Mendling J, Reije s HA (2016) Task-speci ic isual cues o imp o ing p ocess model unde s anding. In So w Technol
79:63–78
Pe usel R, Mendling J, Reije s HA (2017) How isual cogni ion in luences p ocess model comp ehension. Decis Suppo Sys
96:1–16
Pi ke F, Leopold H, Mendling J (2015) Au oma ic de ec ion and esolu ion o lexical ambigui y in p ocess models. IEEE T ans
So w Eng 41(6):526–544
Pus ejo sky J (1998) The gene a i e lexicon. MIT p ess
Ralph P, bin Ali N, Bal es S e al (2021) Empi ical s anda ds o so wa e enginee ing esea ch. h ps://a xi .o g/abs/2010.03525.
Accessed 19 Dec 2024
Reije s HA, Mendling J (2010) A s udy in o he ac o s ha in luence he unde s andabili y o business p ocess models. IEEE
T ans Sys Man Cybe n 41(3):449–462
Reinha z-Be ge I, Figl K, Haugen Ø (2017) In es iga ing s yles in a iabili y modeling: hie a chical s. cons ained s yles. In
So w Technol 87:81–102
Reisig W (2012) Pe i ne s: an in oduc ion, ol4. Sp inge Science & Business Media
Sch eibe C, Abbad-Andaloussi A, Webe B (2024) On he cogni i e and beha io al e ec s o abs ac ion and agmen a ion in
modula ized p ocess models. In Sys 125:102424
Senne A (2011) Ambigui y. S an o d Encyclopedia o Philosophy
Sha a i Z, Soh Z, Guéhéneuc YG (2015) A sys ema ic li e a u e e iew on he usage o eye- acking in so wa e enginee ing. In
So w Technol 67:79–107
Slaa s T, Mukkamala RR, Hildeb and T e al (2013) Ex o ma ics decla a i e case managemen wo k lows as dc g aphs. In: Busi-
ness P ocess Managemen : 11 h In e na ional Con e ence, BPM 2013, Beijing, China, Augus 26-30, 2013. P oceedings.
Sp inge , pp 339–354
Swelle J (2011) Cogni i e load heo y. In: Psychology o lea ning and mo i a ion, ol55. Else ie , pp 37–76
Tu e ken O, Dikici A, Vande ees en I e al (2020) The in luence o using collapsed sub-p ocesses and g oups on he unde s and-
abili y o business p ocess models. Bus In Sys Eng 62:121–141
Vel man J, Jansen C (2005) The ole o ope a o s a e assessmen in adap i e au oma ion. TNO De ence, Secu i y and Sa e y
Soes e be y
Wang W, Chen T, Indulska M e al (2022) Business p ocess and ule in eg a ion app oaches–an empi ical analysis o model
unde s anding. In Sys 104:101901
Webe B, Fische T, Riedl R (2021) B ain and au onomic ne ous sys em ac i i y measu emen in so wa e enginee ing: A
sys ema ic li e a u e e iew. J Sys So w 110946
Wilcoxon F (1945) Indi idual compa isons by anking me hods. Biom Bull 1(6):80–83
Win e M, Neumann H, P yss R e al (2023) De ining gaze pa e ns o p ocess model li e acy-explo ing isual ou ines in p o-
cess models wi h di e se mappings. Expe Sys Appl 213:119217
Wohlin C, Runeson P, Hös M e al (2012) Expe imen a ion in so wa e enginee ing. Sp inge Science & Business Media
Zugal S (2013) Applying Cogni i e Psychology o Imp o ing he C ea ion, Unde s anding and Main enance o Business P ocess
Models. PhD hesis, Uni e si y o Innsb uck
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.
