INVESTIGATING MUSIC TRACK LIKING
IN THE HALO OF ALBUM COVERS
Oleg Leso a1,2∗Anna Hausbe ge 1,2∗I anna Pshenychna1
Oleksand Sh ydanenko1Olha Yeho o a1Ma kus Schedl1,2
1Ins i u e o Compu a ional Pe cep ion, Johannes Keple Uni e si y Linz, Aus ia
2Linz Ins i u e o Technology, Aus ia
∗Equal con ibu ion
[email p o ec ed]
ABSTRACT
Resea ch on music e ie al and ecommenda ion o en ne-
glec s he ac ha a use ’s esponse o a music ack de-
pends on con ex ual ac o s, such as he composi ion o
he esul s lis , he design o he use in e ace o he ad-
di ional media displayed. Howe e , a body o psycholog-
ical esea ch sugges s ha human pe cep ion and decision
making can be s ongly in luenced by con ex ual ac o s.
In pa icula , an ini ial posi i e aes he ic imp ession o a
p oduc may in luence a buye ’s pe cep ion o i s ea u es
un ela ed o appea ance, such as u ili y o eliabili y, which
is a mani es a ion o a cogni i e bias called he halo e -
ec . The wo k a hand in es iga es whe he an album co e
shown o he lis ene du ing playback can c ea e a halo e -
ec , in luencing he lis ene ’s liking o he ack. We ap-
p oach his ques ion by means o a wo-s age use s udy.
In he i s s age, pa icipan s indi idually a ed a se ies o
album co e s and music snippe s. In he second s age, hey
we e p esen ed wi h music acks and album co e s ( om
hose hey indica ed as un amilia o hem a he i s s age)
a anged in pai s, such ha hei leas liked acks we e
shown wi h hei mos liked album co e s and ice e sa.
The esul s show ha displaying an appealing album co e
while playing a music ack may esul in a highe a ing
o he ack and ice e sa. We also obse e an indica ion
o halo e ec c ea ed by he pe cei ed deg ee o ma ching
be ween he music acks and he album co e shown.
1. INTRODUCTION AND RELATED WORK
The halo e ec is a cogni i e bias ha occu s when an
indi idual’s pe cep ion o a speci ic a ibu e (posi i e o
nega i e) in luences hei opinion on o he un ela ed a -
ibu es [1, 2]. A posi i e halo e ec occu s when a a-
o able imp ession o one a ibu e leads o a posi i e bias
owa d o he a ibu es, whe eas a nega i e halo e ec , o
© O. Leso a, A. Hausbe ge , I. Pshenychna, O. Sh yda-
nenko, O. Yeho o a and M. Schedl. Licensed unde a C ea i e Com-
mons A ibu ion 4.0 In e na ional License (CC BY 4.0). A ibu ion:
O. Leso a, A. Hausbe ge , I. Pshenychna, O. Sh ydanenko, O. Yeho o a
and M. Schedl, “In es iga ing Music T ack Liking in he Halo o Album
Co e s”, in P oc. o he 26 h In . Socie y o Music In o ma ion Re ie al
Con ., Daejeon, Sou h Ko ea, 2025.
“ho n e ec ”, a ises when a nega i e pe cep ion o one as-
pec leads o an o e all un a o able imp ession [2].
Va ious o ms o he halo e ec ha e been s udied,
wi h he aes he ic halo e ec eme ging as one o he mos
p ominen [3]. This e ec desc ibes how an ini ial im-
p ession o aes he ics in luences opinions on aes he ics-
un ela ed a ibu es o a p oduc o a pe son. The aes he ic
halo e ec has been in es iga ed in se e al con ex s, in-
cluding webpage usabili y, p oduc usabili y, and pe cep-
ions o indi iduals. Kwak e al. [3] ind ha isual aes-
he ics o a websi e, in hei s udy, a cha i y websi e, can
c ea e an aes he ic halo, shaping he ini ial pe cep ion o
he si e’s epu a ion and in luencing he e alua ion o o he
websi e a ibu es, such as in o ma ion con en quali y and
usabili y. In hei use s udy, Minge and Thü ing [4] show
e idence o wo halo e ec s in use expe ience wi h dig-
i al audio playe s. Ini ially, pe cei ed usabili y was in lu-
enced by isual aes he ics (hedonic halo), bu a e ask
comple ion, pe cei ed isual a ac i eness was in luenced
by sys em usabili y (p agma ic halo). In he s udy o Gu-
la i e al. [5] AI-based beau y il e s we e shown o inc ease
pe cei ed a ac i eness o almos all indi iduals, ega d-
less o gende , age, and e hnici y. The s udy con i med
he a ac i eness halo e ec o ai s like in elligence and
us wo hiness.
The halo e ec has also been in es iga ed in a ious
en e ainmen domains. In hei esea ch Sahoo e al. [6]
show ha he halo e ec in luences mo ie a ings on Ya-
hoo! Mo ies, whe e use a ings o di e en aspec s (e.g.,
ac ing, s o y) a e co ela ed wi h he o e all a ing. Thei
p oposed ecommenda ion algo i hm, inco po a ing hese
dependencies, shows imp o emen s in ecommenda ion
accu acy. In he domain o music pe o mance e alua ions,
i has been shown ha posi i e i s imp ession can impac
pe o mance a ings [7, 8]. Fu he mo e, esea ch has ex-
amined he in luence o isual appeal on music e alua ion,
showing ha a is a ac i eness can posi i ely a ec lis-
ene s’ a ings o hei songs and cu iosi y abou hei mu-
sic. Ad ian and Ioana [9] explo e how an a is ’s isual
iden i y—such as b anding, appea ance, and p omo ional
ma e ials—in luences audience pe cep ion o hei music,
showing ha s ong b anding enhances memo abili y, en-
gagemen , and pe cei ed quali y. Howe e , i an a is o-
cuses oo much on b anding a he expense o au hen ici y,
182
audiences may become skep ical, which could hu hei
long- e m success. Schaap e al. [10] explo ed o wha ex-
en e alua ions o music a e a ec ed by non-musical cha -
ac e is ics o a is s, speci ically hei gende and pe cei ed
a ac i eness. Thei su ey expe imen e eals a s ong
halo e ec , whe e mo e a ac i e DJs ecei e highe a -
ings o hei music, wi h his bias being sligh ly s onge
o male a is s.
Beyond he a is ’s image, music album a wo k has be-
come a c ucial elemen o music ma ke ing. Album co e
designs c ea e expec a ions, shape i s imp essions, and
e en in luence he olume o music s eaming and sales. A
numbe o s udies show ha he design o an album co e
is connec ed o he way he audience pe cei es he ela ed
music. Le [11] shows ha isual me apho s employed
in album co e a impac he lis ene ’s pe cep ion o he
gen e o he song. Acco ding o Venka esan e al. [12] al-
bum co e a design choices, such as colo schemes and
ypog aphy, a ec he emo ional esponse and expec a ions
o music. Joye and Fennis [13] ound ha ca dinal line o i-
en a ions in album a wo k enhance ma ke pe o mance
and lis ene engagemen .
In he domains o music e ie al and ecommenda ion,
co e a has been le e aged, e.g. o music classi ica-
ion [14, 15], music explo a ion [16] o con en -based ec-
ommenda ion [17], while ecei ing limi ed a en ion in he
con ex o cogni i e biases, and he halo e ec in pa ic-
ula [18]. Meanwhile, use s o audio and ideo s eaming
pla o ms a e o en exposed o album co e a while lis en-
ing o music. This could po en ially c ea e a halo e ec and
in luence hei pe cep ion, in pa icula liking, o music.
Showing ha use s’ p e e ence is dependen on p esence
o absence and quali y o shown album a would mean
ha he e alua ion p ocedu es employed in music e ie al
and ecommenda ion would need o be adjus ed o e lec
such con ex . This pape in es iga es he in luence o al-
bum co e pe cep ion on he liking o a song and examines
he po en ial p esence o a halo and ho n e ec s in lis en-
e s’ e alua ion o music acks. In his s udy we ackle he
ollowing wo esea ch ques ions:
•RQ1: To wha ex en can an album co e , shown du ing
audio playback, in luence he lis ene ’s liking o he music
ack played?
•RQ2: I p esen , does he in luence o an album co e
on music ack liking ollow he pa e ns o he halo and/o
ho n e ec s?
2. METHODS AND MATERIALS
We app oach answe ing he esea ch ques ions by conduc -
ing a wo-s age use s udy. In he s udy we compa e use
liking o un amilia music acks in si ua ions when (1) he
playback is accompanied by an album co e a , and (2) no
addi ional media is shown. To p obe bo h halo and ho n
e ec s, o each use , we pai he mo e app ecia ed music
acks wi h leas appealing album co e s and ice e sa. In
he i s s age o he s udy, he pa icipan s a e a ixed se
o music acks and album co e s. In he second s age we
pai he acks wi h he album co e s as desc ibed abo e
and hen he pa icipan s a e he music acks again (wi h
he espec i e album co e p esen on sc een). A all s ages
we ocus on acks and album co e s he use s conside
un amilia . In he ollowing we desc ibe he p ocedu e in
de ail 1.
2.1 Music acks and album co e s
A he i s s age o he s udy, e e y pa icipan a ed he
same se s o 40 album co e s and 40 music ack snippe s.
Below we de ail he selec ion p ocess o he i ems.
We sou ce he music ack snippe s om he publicly
a ailable Music4All da ase [19], con aining 30-second
audio snippe s, me ada a, ags, and o he in o ma ion
o 109,269 acks. We le e age i s ex ension, he Mu-
sic4All–Onion da ase [20] o assess ack and gen e popu-
la i y based on he numbe o unique use s in e ac ed wi h
each ack. In o de o ensu e di e si y in music acks,
we sample hem om 5 mos popula high-le el gen es
in he da ase . We associa e each ack wi h i s p ima y
gen e, acco ding o Music4All, and agg ega e subgen es
o high-le el gen es using Acous icB ainz Gen e Da ase
[21]. This esul ed in selec ing he acks o m he ollow-
ing gen es: "Rock", "Pop", "Elec onic", "Folk, Wo ld and
Coun y" and "Funk and Soul". F om each gen e we selec
8 leas popula acks acco ding o Music4All–Onion in-
e ac ion da a in o de o inc ease likelihood o pa icipan s
being un amilia wi h he acks. Leng h o each ack was
sho ened om 30 o 15 seconds in he in e es o keeping
he s udy a a easonable leng h.
We sou ce a se o 38,746 album co e s o he acks o
Music4All da ase and manually il e ou ones con aining
ex such as a is name, song i le, o o he di ec iden i i-
ca ion in o ma ion. The il e ing p ocedu e esul ed in 356
album co e s, om which 40 we e sampled andomly.
We limi he se s o he same numbe o 40 i ems each
in o de o (1) be able o c ea e pai s o he a ed i ems,
(2) a oid excessi e a igue o he pa icipan s. Acco ding
o Re illa e al. [22] he op imal use s udy leng h is 15
minu es wi h up o 20 min s ill being accep able. In ou
s udy pa icipan s a e equi ed o lis en o each ack o
a leas 5 seconds be o e submi ing hei e alua ion. This,
combined wi h he esul s o a pilo s udy o 24 pa icipan s
esul ed in an es ima e, ha he i s s age o he s udy is
possible o comple e in 16 minu es (mo i a ed pa icipan s
we e allowed o ake longe , e.g. i hey wished o lis en o
he ull snippe s).
2.2 S udy s age 1: ini ial liking o music acks
The i s s age is di ided in o wo sec ions. The pa ici-
pan s a ed a i s 40 album co e s and in he second sec-
ion 40 music ack snippe s o a du a ion o 15 seconds on
a Like -7 scale in e p e ing 1 as "I do no like i a all." and
7 as "I like i e y much.". The co e s and song snippe s
we e andomly shu led o each pa icipan . Addi ionally,
1The esul ing i em selec ion, addi ional s a is ics and ma e ials
can be ound in he eposi o y h ps://gi hub.com/CPJKU/
music- ack-liking-in-album-a -halo.
P oceedings o he 26 h ISMIR Con e ence, Daejeon, Ko ea, Sep embe 21-25, 2025
183
pa icipan s we e asked whe he hey ecognized he song
("Do you know his song?") and whe he hey we e amil-
ia wi h he album co e o an a is ea u ed on i ("Do
you know his album co e and/o an a is on i ?"), wi h
esponse op ions o "Yes" o "No". Fo he a ing o he
song snippe s, he a ing con ols become ac i e only a e
lis ening o a leas 5 seconds o he song. The pa icipan s
we e only able o p oceed wi h he su ey a e answe ing
bo h ques ions o each album co e o music snippe . A
p e ious s udy [23] suppo s his app oach, con i ming ha
exce p s as sho as 5 seconds a e al eady su icien ly ep-
esen a i e in e ms o p e e ence and amilia i y a ings o
he song as a whole.
2.3 S udy s age 2: ack liking in he halo o album a
Pa icipan s who comple ed he i s s age o he s udy we e
in i ed o ake pa in he second no ea lie han 36 hou s
a e comple ion o he i s s age. This ime span was cho-
sen based on Ebbinghaus’s esea ch [24], which sugges s
ha mos newly acqui ed in o ma ion is o go en wi hin
he i s 24 hou s i no ein o ced. Fo each pa icipan ,
album co e s and songs hey ecognized we e indi idually
emo ed, c ea ing a pe sonalized lis o un amilia i ems.
Then hese lis s we e so ed by hei a ed liking. The lis s
wi h anks o album co e s we e so ed in descending o de
( om 7 o 1), and he lis s wi h songs’ anks we e so ed
in ascending o de ( om 1 o 7). The highes a ed al-
bum co e s we e hen ma ched wi h he lowes a ed song
snippe s. I mul iple i ems had he same a ing, hey we e
shu led be o e ma ching. In cases in which he lis s we e
o unequal leng h, i ems we e emo ed om he middle o
he lis o keeping he highes - and lowes - a ed i ems.
In he second s age, pa icipan s we e p esen ed wi h
pai s consis ing o an album co e and a song snippe , c e-
a ed as desc ibed abo e. The pai s we e shown in an-
dom o de o he pa icipan s. They we e, again, asked
whe he hey ecognized ei he i em ("Do you know his
album co e and/o his song?") o assess whe he he ime
gap p o ided was su icien o o ge ing wha hey p e i-
ously seen and lis ened o. In addi ion, he pa icipan s had
o a e he song again ("How would you a e his song?").
As he album co e s we e assigned o music acks
based on he pa icipan s’ indi idual e alua ion, we in-
clude a ques ion o con ol whe he pe cei ed ma ching
o he album co e o he ack a ec ed he liking o he
ack. The pa icipan s a ed he ma ching be ween e e y
ack-co e pai p esen ed o hem on a Like -7 scale ("In
you opinion, how well does he album co e ma ch he
song?"), in e p e ing 1 as "Does no ma ch a all." and 7 as
"Ma ches e y well". The same p ocedu al ules applied as
in he i s s age: pa icipan s had o lis en o each snippe
o a leas 5 seconds be o e esponding, all ques ions had
o be answe ed be o e p oceeding o he nex page o he
s udy.
2.4 Answe ing he esea ch ques ions
We answe RQ1 by in es iga ing he changes in ack ank-
ing, acco ding o use liking, associa ed wi h showing al-
bum co e s du ing music playback. Fo each use we con-
duc indi idual analysis only on acks hey ma ked as un-
amilia du ing he second s age o he s udy. We also o-
cus on he use s who ended up wi h a leas 10 un amilia
acks in he second s age. On he pa icipan le el, we
compu e Kendall ank co ela ion coe icien (Kendall’s τ)
be ween ack ankings achie ed wi h and wi hou he al-
bum co e shown. Kendall’s τ a ies in he in e al be-
ween -1 and 1, wi h he la e co esponding o he wo
ankings being iden ical and, hus, he album co e ha ing
no e ec on use liking o he acks.
To in es iga e he changes among he ini ially mos and
leas liked acks mo e closely, we compu e he ag eemen
sco e be ween s age 1 and s age 2 ankings o each use ,
while ocusing on he highes and lowes anked ki ems.
The ag eemen sco e gk
T op be ween op ki ems in he wo
anked lis s is compu ed as he he numbe o i ems p esen
among he i s kposi ions in bo h anked lis s (Tk
1and
Tk
2), di ided by he leng h o he lis s, k(Equa ion 1). The
ag eemen sco e o he bo om ki ems is compu ed anal-
ogously.
gk
T op =|Tk
1∩Tk
2|
k,(1)
To ensu e ha i ems wi h he same a ing (and he e o e
ied o he same ank) a e ai ly conside ed in he op and
bo om klis s, he calcula ion is epea ed 100 imes, e e y
ime shu ling he acks wi h he same ank a ound. The
inal ag eemen sco e is compu ed as he a e age o e he
100 uns. To con ol o he po en ial in luence o shu -
ling on i em posi ion, we compu e he s anda d de ia ion
o each i em’s posi ion in he 100 andom shu led lis s o
each use and s age. This e lec s how much an i em’s po-
si ion luc ua es as a esul o shu ling.
We answe RQ2 by in es iga ing Spea man’s ank co -
ela ion be ween he change in he ack a ing (∆ss) and
he appeal o he album co e shown a he second s age
(a1
s). In addi ion, we in es iga e he co ela ion be ween
he change in he ack a ing and he pe cei ed ma ching
be ween he album co e and he associa ed ack (ms).
We compu e ∆ss=s2
s−s1
s, whe e s2
sis he s anda dized
use liking o he ack a he s age 2 o he s udy (when an
album co e was shown along wi h he playback) and s1
sis
he same a he i s s age (no album co e shown). Be o e
he co ela ion analysis we s anda dize all a ings (liking
o songs and album co e s, ma ching o album co e s o
songs) using Z-sco e s anda diza ion o each pa icipan
and each s age o he s udy indi idually.
3. RESULTS
3.1 Pa icipan s
A g oup o 100 pa icipan s, li ing in he Uni ed S a es,
ac i e use s o music s eaming pla o ms, we e ec ui ed
h ough he P oli ic 2pla o m. The s udy was conduc ed in
acco dance wi h P oli ic guidelines. The pa icipan s we e
made awa e o he p ocedu e be o e he s a o he s udy.
2h ps://www.p oli ic.com
P oceedings o he 26 h ISMIR Con e ence, Daejeon, Ko ea, Sep embe 21-25, 2025
184
E e y pa icipan ga e an in o med consen and was made
awa e o hei abili y o wi hd aw hei pa icipa ion a any
momen . The da a was collec ed and s o ed anonymously.
In o al 70 pa icipan s success ully comple ed bo h s ages
o he s udy. The mean age o he pa icipan s was 40.7
(SD = 13.8) a he ime o he s udy; 43 indica ed o be
o he sex emale and 27 male. The h ee mos used mu-
sic s eaming pla o ms epo ed by he pa icipan s we e:
Spo i y (n= 42), YouTube Music (n= 30), and Apple
Music (n= 16). On a e age, pa icipan s use 1.60.9di -
e en music s eaming pla o ms.
Du ing he i s s age o he s udy, on a e age, he pa -
icipan s a ed he song snippe s 3.5(SD = 1.8) and he
album co e s 3.9(SD = 1.7). The pa icipan s indi-
ca ed amilia i y wi h 3.3(SD = 6.2) album co e s and
4.5(SD = 7.1) song snippe s. A e indi idually il-
e ing he amilia i ems ou and c ea ing indi idual se s
o pai s co e - ack, he pa icipan s we e p esen ed wi h
34.9(SD = 7.8) pai s o epea ed e alua ion o he co -
esponding music ack snippe s.
In he second s age o he s udy, he pa icipan s indi-
ca ed ha on a e age hey we e amilia wi h ei he he al-
bum co e o he ack in 3.3(SD = 6.7) i em pai s shown
o hem. In his s age, pa icipan s, again, a ed he song
snippe s on a e age 3.5(SD = 1.7) and he ma ching o
he album co e o he song 3.5(SD = 1.9). A e exclud-
ing all pai s wi h amilia i ems, 31.6(SD = 11.0) da a
poin s (pai s) pe pa icipan we e a ailable o analysis.
On a e age, he pa icipan s lis ened o 13.9(SD =
2.5) seconds o he song snippe s in he i s pa and o
14.1(SD = 2.3) seconds in he second pa .
3.2 RQ1: In luence o shown album co e
We in es iga e he in luence o he shown album co e on
he lis ene ’s liking o he music ack played, by compu -
ing indi idual pe -pa icipan Kendall’s τbe ween he wo
music ack ankings. One – based on hei e alua ion in
he i s s age o he s udy, wi h no album co e shown. An-
o he – based on he e alua ion in he second s age, whe e
along wi h he p e iously mos liked acks leas appealing
album co e s we e shown and ice e sa. We ocus on pa -
icipan s who ma ked a leas 10 songs as un amilia du ing
he second s age o he s udy, esul ing in 65 pa icipan s
wi h 33.8(SD = 7.8) acks pe pa icipan .
The dis ibu ion o Kendall’s τo e pa icipan s is
shown in Figu e 2 (mean τ= 0.39,SD = 0.25,MIN =
−0.32,MAX = 0.74). We obse e ha , on a e age,
he lis s a e only mode a ely co ela ed. This sugges s
ha , while songs a ed highe in s age 1 s ill ended o be
a ed highe in s age 2, showing album co e s no ably a -
ec ed ack anking. Fo 44 pa icipan s we obse e sig-
ni ican (wi h he p- alue h eshold 0.05) posi i e co ela-
ion, showing ha hei liking was mo e consis en in he
p esence o album co e s. Fo he emaining 21 pa ici-
pan s no signi ican co ela ion was de ec ed, highligh ing
he impac o album co e s on hei liking.
Fig. 1 shows he ke nel densi y es ima e (KDE) plo
o he s anda d de ia ion o e he posi ions o each i em
h oughou he shu ling p ocess pai ed wi h he s anda d-
ized sco e gi en by each use in phase 1 and phase 2. The
igu e shows he highes densi y mass a a s anda d de ia-
ion o 2, indica ing ha he posi ion a ies by 2 o mos
i ems by shu ling he i ems o he same sco e g oup.
0 2 4 6 8 10
SD o e posi ion o each i em o e
100 shu les pe use
3
2
1
0
1
2
3
S and. i em sco e
Figu e 1. KDE plo o s anda d de ia ion o e posi ion o
i ems in o de ed lis by ank o e 100 andom shu les o
i em a ings in phase 1 and phase 2.
Table 1 epo s a e age anking ag eemen sco es (and
he co esponding s anda d de ia ions) be ween he wo
ack ankings: o e all pa icipan s and sepa a ely o pa -
icipan s showing p esence/absence o signi ican Kendall
ank co ela ion. We no e ha , on a e age, he wo anked
lis s sha e no mo e han 1 common i em in op and bo om
3 (sco es o 0.18, 0.26), and less han 5 in op and bo om
10 (sco es o 0.44), o pa icipan s showing no signi ican
co ela ion be ween he wo lis s (✗). In case o pa ici-
pan s showing mo e consis en ankings (✓), on a e age
hei wo anked lis s sha e mo e han 1 i em in op and
bo om 3 (0.36, 0.39) and abou 6 i ems in op and bo om
10 (0.58, 0.60).
Key indings: (a) Fo 21 ou o 65 pa icipan s no signi -
ican co ela ion be ween ack ankings wi h and wi hou
an album co e shown de ec ed. This indica es a no able
change in ack liking, in oduced by showing album co -
e s du ing playback. (b) Fo he es , he anking lis s a e
be e aligned. Howe e , no able changes in he ops and
bo oms o he lis s can be obse ed, indica ing a shi in
he o e all music ack liking, in oduced by showing al-
bum co e s du ing playback.
0.2 0.0 0.2 0.4 0.6
Kendall's Tau
0
2
4
6
8
Coun
Figu e 2. Dis ibu ion o Kendall’s τo e he pa icipan s.
P oceedings o he 26 h ISMIR Con e ence, Daejeon, Ko ea, Sep embe 21-25, 2025
185
Table 1. A e age anking ag eemen in op and bo om
k o e all pa icipan s, be ween wo anked ack lis s in
s ages 1 and 2.
Signi ican Top
Kendall’s co ela ion 3 5 10
✓/✗0.310.27 0.370.23 0.530.18
✓0.360.28 0.430.22 0.580.15
✗0.180.23 0.230.20 0.440.21
Bo om
3 5 10
✓/✗0.350.27 0.410.23 0.550.16
✓0.390.27 0.450.22 0.600.13
✗0.260.24 0.320.21 0.440.18
3.3 RQ2: E idence o halo and ho n e ec s
Answe ing RQ2, we compu e Spea man’s ank co ela ion
be ween he change in he s anda dized a ings pa icipan s
assigned o he acks in he wo s ages and he s anda d-
ized p e e ence sco e hey assigned o he album co e s
shown wi h he acks in s age 2. Fo his we conside pa -
icipan s who indica ed a leas 2 pai s as un amilia du ing
he second s age o he s udy, which esul s in 68 pa ici-
pan s and 2,213 da a poin s in o al. Fo he in e p e a ion
o he esul s i is impo an o emembe ha he mos ap-
pealing album co e s we e shown wi h he leas p e e ed
acks and ice e sa.
In Figu e 3, uppe , e e y poin co esponds o a mu-
sic ack, accompanied by an album co e a , shown o
a pa icipan in he second s age o he s udy. The x-axis
shows he s anda dized a ing gi en o he album co e by
he pa icipan in he i s s age. Highe alues show highe
appeal o he co e . The y-axis shows he di e ence in he
s anda dized sco e, gi en by he use o he music ack be-
ween he wo s ages. Posi i e alues mean ha in he sec-
ond s age, wi h he album a shown, he ack was a ed
highe han wi hou he album a . Figu e 3, lowe , ol-
lows he same s uc u e, excep he x-axis shows he pa -
icipan ’s pe cep ion o ma ching be ween he album co e
and he music ack. Highe alues co espond o highe
pe cei ed ma ching.
In Figu e 3, uppe , we obse e ha showing highe a ed
album co e s coincides wi h posi i e di e ence in s an-
da dized song a ings. This indica es ha he songs ac-
companied by mo e appealing album co e s see a highe
inc ease in he liking sco e in s age 2. We in e p e his
inding as an indica ion o he halo e ec . We u he no e,
ha showing less appealing album co e s is ollowed by a
decline in music ack liking in s age 2, indica ing he ho n
e ec . Inspec ing Figu e 3, lowe , we no e sligh posi i e
co ela ion be ween he di e ence in song a ings and he
pe cei ed ma ching o he album co e o he song played.
This indica es ha he pe cei ed deg ee o ma ching be-
ween he album co e and he music ack can po en ially
c ea e i s own halo (ho n) e ec .
Table 2 epo s he Spea man’s Rank Co ela ion be-
ween he s anda dized a ings o he s udy. In he ollow-
2 0 2 4
Album co e a ing
6
4
2
0
2
4
6
Di e ence in song a ings
2 0 2 4
Album co e o song ma ching a ing
6
4
2
0
2
4
6
Di e ence in song a ings
Figu e 3. The di e ence in he s anda dized song a ings
(s age 2 - s age 1) ela ed o he s anda dized album co e
a ings and o he s anda dized pe cei ed album co e o
song ma ching a ings o he second s age.
ing we in e p e signi ican co ela ions a p < 0.01, ma ked
wi h *. The album a ings o s age 1 (a1
s) a e nega i ely
co ela ed wi h he song a ings o s age 1 (s1
s), e lec ing
ou me hodology o combining leas appealing album co -
e s wi h mos ini ially liked acks and ice e sa. We ob-
se e a posi i e co ela ion be ween he album liking a -
ings in s age 1 (a1
s) and he di e ence in s anda dized a -
ings o he songs (∆ss), which is an addi ional e idence
o he halo and ho n e ec s. The nega i e co ela ion be-
ween he song a ings o s age 1 (s1
s) and he di e ence in
song a ings in s age 2 (∆ss) indica es he same due o ou
choice o me hodology. The ma ching a ing o he album
co e o song snippe (ms) shows a sligh posi i e co e-
la ion wi h he di e ence in song a ings (∆ss), indica ing
ha he halo (ho n) e ec is po en ially c ea ed no only by
he appeal o he album co e , bu by he o e all pe cei ed
ha mony be ween he album co e and he music ack.
Key indings: (a) P esence o he halo and ho n e ec s
e idenced by signi ican posi i e co ela ion be ween he
album co e appeal and he di e ence in he ack liking.
(b) Po en ial halo (ho n) e ec c ea ed by di e en deg ees
o pe cei ed misma ch be ween he music ack and he
album co e .
P oceedings o he 26 h ISMIR Con e ence, Daejeon, Ko ea, Sep embe 21-25, 2025
186
Table 2. Spea man’s Rank Co ela ion be ween a ings
o he i s s age o he album co e (a1
s), song snippe
(s1
s), di e ence in s anda dized song a ings (∆ss), and
s anda dized ma ching sco e be ween album and song
(ms). Signi ican alues a e ma ked wi h * (p < 0.01).
a1
ss1
s∆ssms
a1
s1.0∗−0.9∗0.4∗−0.0
s1
s1.0∗−0.5∗0.0
∆ss1.0∗0.2∗
ms1.0∗
4. DISCUSSION
In his s udy, we in es iga e he impac o he halo (ho n)
e ec c ea ed by con ex ual media (album co e s) on he
lis ene ’s liking o he music played. Ou indings sugges
ha such con ex can no ably in luence he lis ene s’ e alu-
a ion o music acks, shaping hei subjec i e expe iences.
One key implica ion o ou esul s is he impo ance o
con ex ual aming in music p esen a ion. Album a wo k,
as shown in his s udy and, po en ially, o he con ex , such
as a is epu a ion, o gen e labeling, can subs an ially
a ec lis ene s’ p e e ences and pe cei ed quali y o mu-
sic. This could be le e aged in domains o music e ie al,
ecommenda ion ( o inc ease use s’ exposu e o new un-
known a is s h ough mo e appealing p esen a ion, o de-
elop mo e p ecise con ex -awa e e alua ion app oaches),
and music educa ion ( o imp o e s uden s’ engagemen ).
We obse e e idence o halo and ho n e ec s in ou
s udy. The posi i e co ela ion be ween album co e ap-
peal and he di e ence in ack liking sco es indica es ha
subjec i ely mo e appealing co e s coincide wi h highe
ack e alua ions and ice e sa. This implies ha a is s
choosing p esen a ion unappealing o hei a ge audience
may cause i o unde alue hei music.
Ou esul s hin on he ole o pe cei ed album- o-
song ma ching. The posi i e co ela ion be ween he pe -
cei ed ma ch and a ing di e ences sugges s ha pa ici-
pan s migh ha e adjus ed hei sco es based on how well
he album co e i s he ack. This in oduces an addi ional
con ex ual ac o shaping lis ene s’ liking o a music ack.
5. LIMITATIONS & FUTURE WORK
In he ollowing we discuss he limi a ions o he wo k a
hand and ou line di ec ions o u he esea ch.
Ou esul s show he in luence o he pe cei ed album o
ack ma ching on he pa icipan s’ e alua ion o he music
snippe . Fu he esea ch is needed o alida e he p es-
ence o he e ec . A me hodology o ien ed on c ea ing
album- ack pai s o a ious deg ees o ma ching would be
equi ed, e.g. aking in o accoun he gen e o bo h he
music ack and he music ac ually associa ed wi h he al-
bum co e . Addi ionally, in ou s udy, he ques ions abou
ack liking and album- ack ma ching we e placed on he
same page, which could cause pa icipan s o associa e he
wo. A mo e dis ibu ed s udy design would help isola e
his con ounding ac o .
The s udy a hand was conduc ed wi h a limi ed sam-
ple o US esiden s ac i ely using music s eaming pla -
o ms. Pa icipan s in o he demog aphic g oups a e likely
o show di e en pa e ns in liking music acks and i s e-
la ion o he album a shown. Fi s ly, people o di e -
en cul u al backg ounds may ha e di e en beau y s an-
da ds and di e en ways o p ocess audio isual in o ma-
ion. Secondly, use s o s eaming pla o ms may display
di e en le els o esponse o isual s imuli, such as al-
bum a and use in e ace, compa ed o people who p e e
o lis en o music o line. Las ly, he a en ion lis ene s pay
o album co e s may a y wi h ac o s beyond demog aph-
ics, such as pe sonali y, music p e e ence o music gen e.
The e o e, he esul s o he s udy a hand canno be gene -
alized a beyond he in es iga ed use g oup. Addi ional
esea ch on la ge mo e ep esen a i e samples is needed
o de e mine mo e obus pa e ns.
S udies show ha he halo e ec weakens a e p o-
longed in e ac ion wi h he i em in he halo [5, 9]. Fu u e
wo k could in es iga e whe he he esul s change i he
pa icipan s lis en o s ic ly 5 o 30 seconds o each ack,
and whe he hei ini ial halo-shaped imp ession las s be-
yond he i s lis ening session.
Finally, i has been shown ha he halo e ec can man-
i es o a di e en ex en , depending on he con ex o he
pa icipan , such as mood [25]. Ha ing addi ional con ol
signal o such con ex ac o s would allow o ob ain addi-
ional insigh s in o halo and ho n e ec s and p opose ways
o le e aging hem in music e ie al and ecommenda ion.
6. CONCLUSION
The wo k a hand in es iga es he halo and he ho n e -
ec s c ea ed by album co e s shown o a lis ene du ing
music playback. In ou use s udy pa icipan s we e lis-
ening o un amilia music snippe s accompanied by un a-
milia album co e s. Based on hei indi idual as e, less
appealing album co e s we e shown while music snippe s
mo e aligned wi h hei p e e ence we e played (and ice
e sa). Ou esul s show ha in oducing album co e s
o he lis ening p ocess no ably changed indi idual mu-
sic ack anking, based on e e y pa icipan ’s liking. We
obse e a signi ican co ela ion be ween he album co e
appeal and imp o emen in anking o associa ed music
acks. We also no e ha , beyond he appeal o he al-
bum co e i sel , he pe cei ed ma ching o misma ching
be ween he co e and he music ack can c ea e a halo
o ho n e ec . Ou esul s sugges ha e alua ion p oce-
du es applied o music ecommenda ion and e ie al sys-
ems should ake in o accoun con ex ual ac o s, such as
media (album co e s) shown in he esul s lis s, as hey
may in luence use liking and, hus, use sa is ac ion.
P oceedings o he 26 h ISMIR Con e ence, Daejeon, Ko ea, Sep embe 21-25, 2025
187
7. ACKNOWLEDGMENTS
This esea ch was unded in whole o in pa by he Aus-
ian Science Fund (FWF): h ps://doi.o g/10.
55776/COE12,h ps://doi.o g/10.55776/
DFH23,h ps://doi.o g/10.55776/P36413.
8. REFERENCES
[1] J. L. Nicolau, J. P. Mellinas, and E. Ma ín-Fuen es,
“The halo e ec : A longi udinal app oach,” Annals o
Tou ism Resea ch, ol. 83, p. 102938, 2020.
[2] N. Noo , S. Be am, F. K. C. Yue , K. Genga ha an, and
M. S. M. Rasidi, “Bias, halo e ec and ho n e ec :
A sys ema ic li e a u e,” In e na ional Jou nal o Aca-
demic Resea ch in Business & Social Sciences, ol. 13,
no. 3, pp. 1116–1140, 2023.
[3] D.-H. Kwak, K. R. Ramamu hy, and D. L. Naza e h,
“Beau i ul is good and good is epu able: Mul iple-
a ibu e cha i y websi e e alua ion and ini ial pe cep-
ions o epu a ion unde he halo e ec ,” Jou nal o he
Associa ion o In o ma ion Sys ems, ol. 20, no. 11,
p. 3, 2019.
[4] M. Minge and M. Thü ing, “Hedonic and p agma ic
halo e ec s a ea ly s ages o use expe ience,” In e na-
ional Jou nal o Human-Compu e S udies, ol. 109,
pp. 13–25, 2018.
[5] A. Gula i, M. Ma ínez-Ga cia, D. Fe nández, M. A.
Lozano, B. Lep i, and N. Oli e , “Wha is beau i ul
is s ill good: he a ac i eness halo e ec in he e a
o beau y il e s,” Royal Socie y open science, ol. 11,
no. 11, p. 240882, 2024.
[6] N. Sahoo, R. K ishnan, G. Duncan, and J. Callan, “Re-
sea ch no e— he halo e ec in mul icomponen a -
ings and i s implica ions o ecommende sys ems:
The case o yahoo! mo ies,” In o ma ion Sys ems Re-
sea ch, ol. 23, no. 1, pp. 231–246, 2012.
[7] G. E. McPhe son and E. Schube , “Measu ing pe -
o mance enhancemen in music,” Musical excellence:
S a egies and echniques o enhance pe o mance,
ol. 4, pp. 61–82, 2004.
[8] N. Zoane i and H. Champion, “Mi iga ing he halo e -
ec : Managing he wow ac o in music pe o mance
assessmen s,” Jou nal o Resea ch in Music Pe o -
mance, 2016.
[9] C. Ad ian and I. Ioana, “The ole o isual iden i y
in music pe cep ion: a alk wi h specialis s on song
likabili y, pe cei ed quali y and emo ional eac ions,”
Kome, ol. 11, no. 1, pp. 110–128, 2023.
[10] J. Schaap, M. Be ghman, and T. Calkins, “A ac i e
people make be e music? how gende and pe cei ed
a ac i eness a ec he e alua ion o elec onic dance
music a is s,” Empi ical S udies o he A s, ol. 41,
no. 1, pp. 284–303, 2023.
[11] V. T. Le, “Visual me apho s on album co e s: An anal-
ysis in o g aphic design’s e ec i eness a con eying
music gen es,” 2020.
[12] T. Venka esan, Q. J. Wang, and C. Spence, “Does he
ype ace on album co e in luence expec a ions and
pe cep ion o music?” Psychology o Aes he ics, C e-
a i i y, and he A s, ol. 16, no. 3, p. 487, 2022.
[13] Y. Joye and B. M. Fennis, “Se ha eco d s aigh ! ca -
dinal line o ien a ions in music album a wo k boos
ma ke pe o mance and music consump ion,” Psy-
chology & Ma ke ing, ol. 41, no. 4, pp. 916–937,
2024.
[14] S. O amas, F. Ba bie i, O. Nie o, and X. Se a, “Mul-
imodal deep lea ning o music gen e classi ica ion,”
T ans. In . Soc. Music. In . Re ., ol. 1, no. 1, pp. 4–
21, 2018.
[15] I. Va olkin and C. McKay, “Mul i-objec i e in es iga-
ion o six ea u e sou ce ypes o mul i-modal music
classi ica ion,” T ans. In . Soc. Music. In . Re ., ol. 5,
no. 1, pp. 1–19, 2022.
[16] A. Leh iniemi and J. Holm, “Easy access o ecom-
menda ion playlis s: selec ing music by explo ing p e-
iew clips in album co e space,” in P oceedings o he
10 h In e na ional Con e ence on Mobile and Ubiqui-
ous Mul imedia, MUM, 2011, pp. 94–99.
[17] Y. Deldjoo, M. Schedl, and P. Knees, “Con en -d i en
music ecommenda ion: E olu ion, s a e o he a ,
and challenges,” Compu . Sci. Re ., ol. 51, p. 100618,
2024.
[18] M. Schedl, O. Leso a, and S. Masoudian, “The im-
po ance o cogni i e biases in he ecommenda ion
ecosys em: E idence o ea u e-posi i e e ec , ikea e -
ec , and cul u al homophily,” in P oceedings o he
11 h Join Wo kshop on In e aces and Human De-
cision Making o Recommende Sys ems co-loca ed
wi h 18 h ACM Con e ence on Recommende Sys ems
(RecSys 2024), Hyb id E en , Ba i, I aly, Oc obe 18,
2024, se . CEUR Wo kshop P oceedings, ol. 3815.
CEUR-WS.o g, 2024, pp. 113–123.
[19] I. A. P. San ana, F. Pinhelli, J. Donini, L. G. Ca ha in,
R. B. Mangolin, Y. M. e Gomes da Cos a, V. D. Fel im,
and M. A. Domingues, “Music4all: A new music
da abase and i s applica ions,” in 2020 In e na ional
Con e ence on Sys ems, Signals and Image P ocessing,
IWSSIP, 2020, pp. 399–404.
[20] M. Mosca i, E. Pa ada-Cabalei o, Y. Deldjoo,
E. Zange le, and M. Schedl, “Music4all-onion–a
la ge-scale mul i- ace ed con en -cen ic music ec-
ommenda ion da ase ,” in P oceedings o he 31s
ACM In e na ional Con e ence on In o ma ion &
Knowledge Managemen , 2022, pp. 4339–4343.
P oceedings o he 26 h ISMIR Con e ence, Daejeon, Ko ea, Sep embe 21-25, 2025
188
[21] D. Bogdano , A. Po e , H. Sch eibe , J. U bano, and
S. O amas, “The acous icb ainz gen e da ase : Mul i-
sou ce, mul i-le el, mul i-label, and la ge-scale,” in
P oceedings o he 20 h In e na ional Socie y o Mu-
sic In o ma ion Re ie al Con e ence, ISMIR 2019,
Del , The Ne he lands, No embe 4-8, 2019, 2019, pp.
360–367.
[22] M. Re illa and J. K. Höhne, “How long do esponden s
hink online su eys should be? new e idence om
wo online panels in ge many,” In e na ional Jou nal
o Ma ke Resea ch, ol. 62, no. 5, pp. 538–545, 2020.
[23] S. J. Philibo e, S. Spi ack, N. H. Spilka, I. Passman,
and P. Wallisch, “The whole is no di e en om i s
pa s: music exce p s a e ep esen a i e o songs,” Mu-
sic Pe cep ion: An In e disciplina y Jou nal, ol. 40,
no. 3, pp. 220–236, 2023.
[24] H. Ebbinghaus, “[image] memo y: A con ibu ion o
expe imen al psychology,” Annals o neu osciences,
ol. 20, no. 4, p. 155, 2013.
[25] J. P. Fo gas, “She jus doesn’ look like a philoso-
phe . . . ? a ec i e in luences on he halo e ec in im-
p ession o ma ion,” Eu opean Jou nal o Social Psy-
chology, ol. 41, no. 7, pp. 812–817, 2011.
P oceedings o he 26 h ISMIR Con e ence, Daejeon, Ko ea, Sep embe 21-25, 2025
189
