=Paper=
{{Paper
|id=Vol-2225/paper3
|storemode=property
|title=Risk "Attention" or "Adventure": A Qualitative Study of Novelty and Familiarity in Music Listening
|pdfUrl=https://ceur-ws.org/Vol-2225/paper3.pdf
|volume=Vol-2225
|authors=Vikas Kumar,Sabirat Rubya,Joseph A. Konstan,Loren Terveen
|dblpUrl=https://dblp.org/rec/conf/recsys/KumarRKT18
}}
==Risk "Attention" or "Adventure": A Qualitative Study of Novelty and Familiarity in Music Listening==
https://ceur-ws.org/Vol-2225/paper3.pdf
Risk "attention" or "adventure": A Qualitative Study of Novelty
and Familiarity in Music Listening
Vikas Kumar Sabirat Rubya Joseph A. Konstan Loren Terveen
University of Minnesota
GroupLens Research
Minneapolis, Minnesota, USA
[kumar093,rubya001,konstan,terveen]@umn.edu
ABSTRACT previously unseen options. While prior works [25, 28, 30] based on
While recommendations systems have shown great improvements online music activity logs demonstrated that understanding users’
in generally predicting relevant items, they still face challenges in consumption of novel and familiar options in previous sessions
achieving the delicate balance between novel and familiar options. play a critical role in interpreting users’ needs for novelty, the
Existing works in pursuit to address the challenge have focused on insufficient context about users’ intent in activity logs makes the
the activity logs and algorithms while largely ignoring the user- assumptions imposed by previous methods on what users listen
specific needs and challenges in the balance users seek between to, to be misinterpreted. For instance, a user who likes jazz and
novelty and familiarity. As a result, the assumptions imposed on chooses to listen to a new jazz album could do so for variety of
user actions based on activity logs are limited and could lead to reasons, such as the user wants to catch up on a new album he/she
misinterpretation of users’ needs. To better understand user needs, has not listened to yet, is bored with his or her existing selection,
in this paper, we study users engaged in online music listening or has discovered an artist from a recent jazz event, a friend, or
activity to understand users’ interpretations and rationales in their online media. These factors thus play critical role in users’ selection
novel and familiar music selections. We show that a combination of music and the balance they seek in the amount of novelty or
of factors, both explicit and implicit, such as boredom, need of familiarity in their music.
attention, risk of a bad selection; that play influential role in users’ As a result, in this paper, we use a qualitative approach to study
novel and familiar music selections. We discuss the findings and users while they engage in music listening activity to understand
the implications for user interactions and user modeling that could what and how they balance between the novelty and familiarity in
help better understand what, when, and how users seek the balance their music selection. We use the Contextual Inquiry (CI) method
between novelty or familiarity in their recommendations. where one gathers information about users while they perform
their tasks in the given context. This approach provides a medium
KEYWORDS that allow users a conversation to reflect upon and provide self-
interpretation of their actions for better explanation [21].
Music Recommendation; User Preferences; Qualitative Study; Novel
Through our analysis of about 800 codes generated from 14 such
Versus Familiar;
conversations, we provide insights into user actions and explore
answers to the following three research questions:
1 INTRODUCTION • RQ1: What do users choose? How do users choose? And
The giant ecosystems of music providers, such as Spotify, Apple when and why do they choose to explore novel or familiar
Music, Pandora, Amazon Music, etc. boasts abundance in choices selection?
of content to attract users (e.g. all the music you want in one place) • RQ2: What explicit or implicit factors influence users’
creating tiers of content access for a variety of scenarios [9]. While choices in listening to familiar or novel music and why?
these options promote variety and freedom in choice for users in • RQ3: What are the common challenges that users face
aim to provide better experience, they also pose a challenge to users while seeking familiar or novel music to listen to?
who feel paralyzed and confused with an overwhelming number of
We identify multiple factors — such as the effort in finding new
choices [37].
music, the risk of a bad choice, the potential loss of attention due
Realizing that vast collection of content is both challenging [41]
to new stimuli, boredom from existing selection, the excitement
and limited in their ability to provide great user experience [36],
to explore, as well as the implications of mere exposure [6] from
content providers have adopted more subtle and distinct interpre-
external sources — for the balance participants sought in novelty
tations of users’ taste in form of playlists [46], radio stations [27],
and familiarity in their music selection. We conclude with the design
etc using popular recommendation algorithms. These algorithms
implications of our findings and outline the factors we believe
based on historical consumption data from the action logs strive to
belong in a user model, including when, how, and for whom to
achieve the delicate balance between helping users discover new
balance novelty and familiarity in music recommendation.
music versus helping users find familiar well known options [19].
The existing techniques, however, in their pursuit to find the
balance have largely ignored the user-specific needs. For instance, 2 RELATED WORK
a user might find comfort in familiar choices at times whereas Music is considered more ubiquitous than many other media of
could solicit the excitement or delightful surprise from novel or consumption, such as movies, books, photos, etc. While some view
IntRS Workshop, October, 2018, Vancouver, Canada Kumar et al.
choosing music to listen to as an expression of free will and mun- Compared to other qualitative methods such as surveys and
dane [45], music choice is better understood as a product of inter- questionnaires, contextual inquiry does not suffer from recall bias.
linked social, environmental, cognitive, and biological factors [2, 38]. Surveys and questionnaires assume that users know why they per-
Several common-day tasks involve music, such as walking, cooking, formed a task or what they needed to complete a task. However,
cleaning, working, and relaxing, which have their own complex while engaged in a task like music listening, people do not neces-
sources of meanings and emotions [45]. Various field and lab stud- sarily reflect on what they are listening to, making it difficult to
ies have found multiple reasons why users listen to music, such as, form meaningful interpretations of user actions without the con-
to manage mood [33], to create social identities [18], or to provide text [42]. Instead, CI is effective at uncovering tacit knowledge [21].
a distraction from their surroundings [15] etc. It allows participants to engage in dialogues to reflect on and shape
However, as online music services grow in popularity, the lis- the interpretations of their actions while being in the context of
tening behaviors of people are also changing [31]. The ease of the task.
online-streaming, availability across platforms1 , voice-assistants23 , Due to the qualitative nature of the study, we chose a small
and the inherent psychological and emotional benefits of listen- number of participants to conduct in-depth inquiries with while
ing [15, 43] have made music more accessible and personal to users. they listened to their music in their everyday environments. We
These systems follow listeners’ trail and similar users’ music choices describe the participants, procedure, analysis, and platforms we
to estimate the likelihood of suggesting a similar but new music. studied in the following sections.
However, even with huge amounts of user data, it can often be chal-
lenging to truly understand listeners’ motivations for their music
choices. The data does not imply why listeners behave or listen in
unique ways, especially how and why users seek comfort in their 3.1 Participants and Procedure
selection at times and the excitement to explore new music at other Participants were recruited through informational posters in public
times. places like coffee shops, university libraries, and private workplaces
The challenge to find the delicate balance between novelty and in the cities of Minneapolis and Seattle. Although participation was
familiarity is critical to recommendation systems [23, 51]. An over- voluntary, each participant was required to meet the minimum
emphasis on novelty in recommendations, for instance, can lead criteria as follows: (a) must be 21 years or older, (b) must listen to
to distrust and frustration [13], whereas an under-emphasis on music at least 2-3 times in a month, (c) must be comfortable sharing
novelty can lead to boredom and dissatisfaction [8, 19]. As a re- music experience in an everyday environment, and (d) must be able
sult, plenty of approaches exist, such as topic-diversification [51], to schedule two 1-hour sessions.
item-taxonomy [47], or declustering [49], to introduce novelty or Interested participants contacted researchers via email and were
diversity in recommendation lists. then referred to fill out an online questionnaire to confirm their
The newer deep machine learning systems paired with the huge eligibility. As required by the Institutional Review Board (IRB) [24],
amounts of data are shown to continuously update and change the in the questionnaire participants reviewed the details and require-
behavior of a product to match the expectations of users [11]. While ments of the study, including their eligibility with regards to their
the machine learning based recommendations are most likely the age, their daily music listening patterns, their preferred music plat-
only way to efficiently scale music recommendation for millions of forms, and their preferred times and locations to schedule the in-
users, they also push the focus to infer user preferences from mining terviews. Each participant was compensated with a $30 Amazon
the data and overlook user-specific needs in the system. As a result, gift card upon completion of the final session.
they impose interpretations based on activity logs without sufficient Based on responses to the questionnaire, we used the “purposive”
context of why users make the choices they do. In this work, we sampling approach, in which we include participants from multiple
therefore take the qualitative approach to better understand user- platforms, different professional backgrounds, and with different
specific factors to interpret their actions while they engage in music listening patterns such that a variety of users are represented to
listening activity. confirm or challenge emerging patterns [32].
Participants were then invited to schedule two 1-hour sessions,
3 METHOD each separated by several days and scheduled at different times of
the day to understand users explanations under different contexts,
To study the factors in user actions while they listen to music, we
while keeping the location of the interview the same. During each
use the contextual-inquiry (CI) method. It is a qualitative approach
session, for the first 30 minutes participants picked music to listen
to obtain information by observing and interviewing participants
to while we took notes, with no interruption to the user. In the
while they perform the task in their everyday environment. This
following 30 minutes, we asked users a set of questions to allow
method, adapted from field research techniques in psychology [12],
users to explain and reflect on their actions.
sociology [17], and anthropology [40], has shown to be effective in
Participants’ consent was taken during each interview. Partic-
gaining better insights and understanding of user actions in online
ipation was voluntary and users were given a choice to stop the
media consumption [7, 10, 26].
interview at any time of the observation session. We noted users’
actions, such as browsing, searching, clicking, skipping, and shuf-
1 pandora.com/everywhere fling, while they picked their music to listen to. As music can be
2 spotify.com/us/googlehome a background process while the user focused on other tasks like
3 spotify.com/us/amazonalexa writing, reading, browsing, etc., we asked users to install Last.fm
IntRS Workshop, October, 2018, Vancouver, Canada
Table 1: Participant Details 800 open codes were generated, with each code being reviewed by
two researchers. The number of open codes was then significantly
Platform Age/ Consumption Location reduced in the process of memoing and categorizing using a con-
Gender stant comparison described in affinity mapping [44]. In the process,
every code was compared to others and positioned to reflect its
P1 Apple Music 20-30/M Daily Workplace
affinity to an emerging topic and the research questions. We used
P2 Spotify 20-30/M Daily Workplace
the topics to understand the common themes in responses and
P3 Pandora 40-50/M Once per week Workplace
actions of different participants to answer the research questions.
P4 Spotify 20-30/M Daily Workplace
We also captured themes with disparate responses or opinions to
P5 Spotify 20-30/M 2-3 times per week Transit
gather divergent user perspectives.
P6 Spotify 20-30/F Daily Home
P7 Spotify 20-30/F 2-3 times per week Library/Cafe 3.3 Platforms
Here we discuss the three platforms (Spotify, Apple Music, and Pan-
scrobbler4 (using their own account and consent) to enable song dora) used among our participants. Although these platforms serve
tracking for the music they listened to during the session. This also the similar function of streaming online music, it is critical for this
helped avoid any interruptions while a user was engaged in his/her study to understand the differences and the types of services each
task. They were reminded at the end of session to uninstall the tool. provides. For instance, Spotify and Pandora provide a freemium 5
Table 1 provides a description of the participants, their back- service, while Apple Music provides only a paid subscription-based
grounds, and their music listening patterns. A total of 7 partici- service. Spotify and Apple Music both provide music interactions
pants were selected and 14 interviews (2 for each participant) were that include lists of curated playlists, personalized playlists, al-
conducted in the study. Participants included in the study were bum/artist suggestions, as well as lists of new releases and top
professionals from different backgrounds and included one student charts. Until very recently, Pandora only provided a list of stations
(P7). Participants had varying degrees of interest in music with two that users could create, or they could choose from existing system-
avid listeners (P2 and P6), two with music backgrounds (P1 and P4), suggested stations based on artist, track, mood, genre, etc. Pandora
and three casual listeners (P3, P5, and P7). While our participants also provides a few niche options for users, such as selecting I am
did not span across a wide age range, they do fall within the age tired of this track! to skip a track, compared to only thumbs down
group of 22 to 35 year olds; the age group that has most embraced or next track option in the other two platforms. Finally, all three
online music streaming services [37]. Among the 7 participants, platforms boast an abundance of music options for users with their
four participants were observed while at their workplaces (offices) premium subscriptions, which each of our participants had for
(P1, P2, P3, P4), one while at home [P6], one while working at a their choice of platform. As a result, each session during the study
public place (library) (P7), and one during transit (P5). was ad-free and included all services that the individual platform
provides.
3.2 Analysis
To assess users’ responses, we had to determine what songs were 4 RESULTS
familiar to users and which were novel. However, this was challeng- We now discuss results from the analysis. We first discuss the
ing due to two reasons: 1) the definition of novelty as understood by choices participants made and their intents before listening, fol-
a recommendation system is not how users perceive novelty, and lowed by key factors identified as common themes across partici-
2) every participant was likely to have a subjective interpretation pants that affected their choices of novel and familiar music. Finally,
of novelty in their selection. For example, a song that a user really we conclude with key challenges common across participant re-
likes but has not listened to in awhile can be novel to the user [1]. sponses while they sought novel or familiar music.
Oba et al. in their study of nostalgic experience have shown that
items from the past that a user likes but has not seen or consumed 4.1 Choice of music: What did participants
recently induces nostalgia in the same parts of the brain that are listen to?
active during novel exposure [39]. As a result, we did not ask di- We recognize two specific cases in choices participants made to
rect questions about users’ familiarity or novelty with their music select music to listen to — they either (1) picked music they knew
selection. Instead we asked if the music chosen was (a) listened to they wanted to listen to, that is, a specific artist, an album, or a band
recently, that is within the last few weeks to a month, which we or (2) picked music for which they didn’t have a specific artist or
consider as familiar, or (b) listened to in the past but not in awhile album in mind but instead had a general preference for the kind of
or never listened to before, which we consider as novel in this study. music they wanted to listen to. In the latter case, their choice was
This definition of novelty helps capture the inherent property of specifically to align with specific needs in form of mood, attention,
repetition in music, which is well known to delight users [3, 34]. etc. For instance, “smooth Jazz instruments that won’t interfere
After the completion of the interviews, the 14 hours of voice with my studies.” The participants’ sessions for each case is shown
recordings (two hours for each of the 7 participants) were tran- in Table 2. We discuss both the cases and how participants arrived
scribed into open-codes to capture the individual viewpoints, ratio- at their choices in this section.
nales, and interpretations users shared during the sessions. About
5 A type of service in which a platform provides the music free of charge, but some
4 https://www.last.fm/about/trackmymusic premium features like high-definition music or ad-free music are available for a charge.
IntRS Workshop, October, 2018, Vancouver, Canada Kumar et al.
Table 2: Distinct cases as per user response before they se- songs large enough that they had not listened to the entire playlist
lected the music to listen to during their session. Case 1 yet. When asked about the concern of repetitions, participants men-
is when participants knew what they wanted to listen to, tioned using shuffle to add some uncertainty to the order of songs
whereas Case 2 when participants did not have a specific and that they didn’t mind if some songs repeated over multiple
artist or album in mind but a general preference of what sessions.
they wanted to listen to. R1 and R2 refer to the first and
P2: “I wanted to listen to the playlist I curated
second study session for each participant P. The sessions in
last week. I am excited about the playlist and so
italics are the ones where participants choice of music was
I wanted to listen to it again today"
different than what they were listening to in a session prior
P4: “Spotify created this year-end playlist that I
to the study.
have been listening to for the last couple of days.
Although I have listened to most of the songs in
Case 1 P4-R2, P1-R1-R2, P2-R2, P4-R1, P2-R1 the list, I will continue listening for some time, as
Case 2 P6-R1, P5-R2, P7-R2, P5-R1, P3-R1-R2, P6- I don’t remember listening to the same songs due
R2, P7-R1 to the sheer size of the playlist."
Case 2: When they did not have a specific artist, song, or album
they wanted to listen to, participants cited that their music selec-
Case 1: When participants knew the specific music they wanted to
tion was to align to the specific needs of that hour. For instance,
listen to, they mentioned choosing albums or songs by an artist they
participant P7 cited the benefit of selecting a jazz playlist she had lis-
had recently discovered. Participants mentioned that discovering
tened to earlier in the week to avoid the divided attention between
music they found interesting but different from what they usually
music and an assignment that needed focus. (P7: “Was looking for
listened to was mostly influenced by external sources, such as
something that I can play while doing my assignment and liked how
“found on a TV show,” “saw it trending on BBC Radio One,” etc.
smooth this playlist is, as it has a more monotonous tone that helps
P4: “I found this artist from a TV show and liked me focus.” ).
one of the songs I listened to on YouTube. I liked In another instance, some participants (P6-R2, P7-R1) mentioned
the music and wanted to listen to more of that their desire for a calming and relaxed mood as a reason for their
music on Spotify.” selection.
P2: “I found this DJ from BBC Radio One, which
P6: “I think I selected the playlist because I was
often releases collections of new songs that I like
looking for something easy and warm for the
to explore and listen to to see if I like them.”
mood. It is rainy outside . . . this just fits the at-
Participants mentioned that in some cases discovery was not mosphere I guess. I like some songs in this playlist
only limited to new artists or new bands they had never heard and often choose this playlist to relax.”
before, but also included “new songs they have not yet listened to P7: “I wanted to be in, like, a good mood as I
by a familiar artist” and how mentions from friends or news media have a busy night, so I wanted to listen to some-
helped them find this music. thing happy, upbeat, and generally kind of light,
P4: “I heard that a new release is coming up from and this is supposed to be happier than other
this band, which reminded me of my previous playlists. Also, Christmas is soon and I am a big
favorite album from the band. I just wanted to fan of Christmas music, so I’m kind of getting in
go back to music by the artist before the new a holiday spirit."
release.” In the case that they didn’t have a specific kind of music in mind
P1: “This remix recently came and my friend to listen to, participants also relied on recommendations from the
shared it with me. I really like the remix, although music streaming platform. They picked an album or playlist that
I generally prefer to listen to the original itself.” they were very familiar with but had not recently listened to (P5:
P2: “A new song came from the artist recently “Saw the album in the suggestions and I had not listened to the artist
and my friend, who I believe has a good taste in (Kendrick Lamar) recently. I really like his songs." ). The participant
music, told me about it. So I wanted to listen. I mentioned the induced boredom from his current selection as a
like the song and will likely listen again.” reason to choose something different (P5: “Didn’t want to continue
A few participants (P4, P1, and P2) also mentioned the role ex- what I was listening to as I have been listening to it for few days
ternal events and sources play in reminding them of an old album now." ).
or an artist they had not listened to in awhile (P4: “A game I was Summary: Participants found the music they listened to in var-
playing last night had a tune in the background that reminded me of ious ways. Among these cases, we recognize that the participants’
the band I wanted to listen to today...” ). choice of music, and specifically music different from what they
There were a few sessions (P4-R1, P2-R1) during which par- were listening to, was due to external events or sources. They men-
ticipants knew what they wanted to listen to but did not choose tioned they “discovered an artist from social [media] mentions,”
anything different from what they were currently listening to. Par- they found “new releases from a familiar artist,” their choice was
ticipants chose to continue with an album they were recently lis- an “old favorite that I haven’t listened to in a while,” they had seen
tening to or a playlist they had recently curated or that had a list of a “news story that reminded me of a favorite”, or they were simply
IntRS Workshop, October, 2018, Vancouver, Canada
“bored from their current selection.” Each of these reasons highlight Apart from new music resulting in distraction, a couple of par-
the users’ excitement to either discover a new artist or rediscover ticipants (P3 and P6) mentioned that their past favorite music they
an old favorite from an external source. Whereas for participants have not listened to recently (subset-(b) - novel) resulted in similar
who chose to continue with music they had been listening to, they distractions as it brings their attention to the parts of tunes or lyrics
primarily cited wanting “to continue listening to an existing playlist they like, and the memories associated with the tunes.
that I have not finished yet,” to listen to an “album or list that aligns Participants (P4, P5, and P7) also mentioned the specific type
with the desired mood,” or to find “a playlist that helps maintain or genre of music they prefer for specific attention needs. For ex-
focus on the task at hand.” These users’ rationales suggest they ample, classical or no vocals music to help them zone out from
found comfort in listening to music they were currently listening their surroundings to focus on their work at hand (P4: “..there is
to instead of putting forth the effort to find different music. As instrumental or classical that is not distracting.." ).
such, we observe that the participants’ selections of music were a To summarize, participants’ attention requirements play a unique
combination of specific curiosity needs, moods, or desires. role in their choice of music. Participants preferred to choose music
they recently consumed with high familiarity when they know the
4.2 What explicit and implicit factors attention requirement of the task at hand is high to avoid distrac-
influence users choice of music? tions, and new or forgotten music when it is low. Participants noted
that the effort involved in selecting novel music and the risk of
There are some key factors we identified across users’ responses bad selection are often the causes of divided attention – a possible
for their specific selections of music. explanation of why familiar music help users find comfort and ease
4.2.1 Attention. Music is known to play a crucial role in either in their choice.
aiding or distracting the attention needs of users [14]. Music one 4.2.2 Boredom. Participants frequently mentioned boredom as
likes can help increase focus, while music one doesn’t impedes one of the primary reasons for selecting the music they listened
it [22]. Participants (six out of seven participants) mentioned that to during the session. They mentioned being tired of their current
their degree of attention between music and work play an influential selections as primary factor in selecting music other than what
role in their choices of music. They noted that the type of task at they were currently listening to. Some participants (P1, P2, and P4)
hand affects what kind of music they prefer. For example, a task that mentioned that they sought exploration and hence looked for new
demands high attention, such as reading to absorb new information, music, whereas other participants (P3, P6, and P5) highlighted the
versus a less attention-demanding task, such as browsing online pleasure they sought in playing past favorites they had not recently
social media, results in different selections of music. listened to, which makes them feel nostalgic and relaxed. When
P2: “I don’t like working to new music. To know asked specifically about which novel music would they prefer, new
whether I like an artist/song, I have to really give or past favorites, they cited the available attention and the effort to
it attention. Either figure out the song or do work, find new music as the dependent factors.
can’t do both unless given an attention. As an In addition, irrespective of whether participants chose past fa-
alternative, if at work and likes something then vorites or new music, they commonly mentioned seeking music
will save it for later to give necessary attention." from different genres than they were currently listening to to avoid
boredom.
Participants mentioned that when attention needs are high, they
look for music they frequently play, citing the comfort of listening to P1: “I feel like its harder to find something new I
familiar music in helping them focus (P7: “Yes, today I was wanting could fall in love with in the genre I listened to
to listening something like this because I just wanted to focus and I most. I can more easily find something exciting
have a lot of stuff to do. I listen to it often when I am studying and in a different genre that I don’t usually listen to.”
writing"). And, when attention needs are low, they prefer listening P3: “I did not come across this music (I like) before
to new music (P2:“Okay being distracted as was in mood to discover as I don’t usually listen to this genre.”
new songs”). They discussed how the excitement from the new
For some participants who were multilingual this meant chang-
stimuli led them quickly down a "rabbit hole" of exploring artists’
ing the language of music from what they were currently listening
discography and other similar artists, thus taking attention away
to (P5: “If I get bored from this, I often go back to choose music from
from the task at hand.
language (native) other than English").
P4: “Explicitly wanted something to occupy more
of my attention and listening to new music is an 4.2.3 Sheer Joy of Adventure. While being bored from current
easy way to do that because its a new stimuli. You selections was cited as a cause to select new or forgotten music,
know, may be discovering a new band or artist, users also cited “sheer joy of adventure” as the other reason to choose
that quickly go down very deep rabbit hole" music different from their current selection.
P2: “I am okay being a little distracted today. It P2: “ I didn’t feel like playing my older playlists.
does not happen always but when it happens I I just woke up today in discovery mood and then
am in this way to queue up few things. Picking up looked up a playlist shared by my friend and I
related artists from a new song I like is another found remix of a song that I liked and it all kicked
easy way but also takes the attention away." off from there into finding more related artists."
IntRS Workshop, October, 2018, Vancouver, Canada Kumar et al.
The users’ choices to seek new artist because of sheer joy of ad- P7: “....would like to definitely listen to my fa-
venture emphasize a limitation on the assumption of user model in vorites in the mix of songs in the playlist...”
existing boredom-based novelty recommenders [29, 30]. They only P3: “....instant gratification with your choice of
considered that tendency of user to seek novel music is dependent music when it plays songs you are most familiar
on the boredom of user with current selection, however as our with...”
participants mentioned, it could also be users’ appetite for sheer
joy of adventure. A few participants also changed the music they were listening
To summarize, boredom of participants with their current selec- to during the sessions. This included responding positively to a
tions influences significantly what music users choose to listen to. song, leading to queuing up a couple more songs from the artist
While users look for novel music when bored, they are likely to or other similar artists, or skipping to the next song in the queue.
explore different genres or languages than they were listening to. Participants cited that the primary reason to skip a song is to avoid
In addition, the excitement participants sought in new stimuli is the distraction from the tune or lyrics of the song. One participant
not limited only to boredom, but also the sheer joy of exploration mentioned skipping to reach their favorite song in the playlist
that can lead users to choose novel music. sooner rather than later (P5: “I selected the playlist as I remember
one of my favorite song was in that playlist. I skipped few songs to
4.2.4 Priming. Users choosing novel music mentioned learning reach the favorite song quickly. It would have been nice to have that
about the artist(s) from conversations with friends [P1, P2, P4], TV song play earlier.” )
shows [P5], music blogs [P2], DJs [P2], online social media [P1], Participants who listened to albums, however, cited a distinct
etc. In their effort to recall the external events that affected their tendency to avoid shuffling and skipping. Participants P1 and P4
choices, participants cited sources they trust more than others. For cited their rationale about avoiding shuffling as albums “represent a
example, they mentioned only a few friends who they believe have statement by an artist and I respect the order the artist wants them to
good tastes in music, TV celebrities they like, accounts they follow be heard. I listen to whole album front to back as is and always finish
on social media, and DJs from a reputed radio station, like BBC the album if halfway through” (P1), and “ I rarely shuffle albums. I
Radio One, as sources of influence. know that a lot of times a band will structure albums to kind of have
Apart from external events, participants also mentioned influ- an inherent flow to them. I also do not skip songs in an album even
ence due to recommendations shown on the music platforms (P1: with moderate liking.” (P4).
“often choose new album to listen when platform (Apple Music) sug- To summarize, participants had distinct expectations of the order
gests release of new album from one of my favorite artist") for albums and for playlists. They preferred listening to an album in
These events affected users’ choice in music in ways similar to the same order it was curated, whereas for playlists they preferred
the effects of priming, which reinforces the notion that subtle expo- to shuffle the order of songs with an expectation to listen to their
sure of an entity can cause large effects on the perceived attraction favorites sooner for instant gratification.
for the entity, also referred as perceived familiarity. In social psychol-
ogy, perceived familiarity is defined as the feeling of acquaintance 4.2.6 User Settings. Last but not least, participants P6 and P7
upon mere exposure to an item [48] that leads to a perceived at- highlighted the effects of setting or environment on their selection
traction to the item [35]. Although recommendation systems have of music. Participants discussed how their selection could be dif-
studied ways to model perceived familiarity of users [20], inducing ferent from their current choices based on location, such as the
this familiarity is often bounded to user interactions within the gym, work, home, etc., and the time of day. For instance, some
system [50]. These systems limit the ways to incorporate the effects participants (P2 and P7) preferred listening to mainstream popu-
from sources outside the system–sources in which users place a lar new releases in the evening to avoid disruptions during work
high degree of trust. hours (P2: “Even though I really do like their music (a BBC Radio
One DJ releasing a new playlist), I prefer listening to these while at
4.2.5 Order in the selection of music. Apart from what partici- home in the evening when I play more mainstream stuff”). Similarly,
pants selected, we also asked participants about their preferences participants P1, P4, and P6 mentioned listening to non-vocals at
in the order of songs in the playlists and albums they chose to listen work and fast, upbeat pop music during workout sessions.
to. P6: “Would not choose this music (slow and calm-
Participants who chose playlists mentioned using shuffle to in- ing) for a workout or when thinking quickly to
crease the uncertainty in the order of songs (P4: “I usually shuffle match up the rhythm”
the playlist because I don’t know if this music is most played or 100th
most played, so I usually shuffle if I don’t know there is specific or- Based on the observations of settings across various sessions, we
dering that is going in”). However, even with this uncertainty, a found that participants whose sessions were at the start of the day
few participants who listened to a playlist during the sessions men- when the focus on work is still divided selected more novel options.
tioned the desire to listen to their favorite songs in the playlist However, participants whose sessions were in the afternoon when
sooner rather than later for an “instant gratification” (P3) of their they needed to focus more on work chose familiar music to listen to.
choice in playlist. Other than the attention requirement, we believe another possible
explanation can be the fatigue of the day causing participants in
P4: “Selected a song that I recognized I like in the the afternoon to choose something comfortable without the added
playlist.” effort of finding and exploring something new.
IntRS Workshop, October, 2018, Vancouver, Canada
Users also highlighted that when not interacting with the plat- curated playlists in which songs are expected to be similar to a
form interface, like during workouts or driving, they prefer to pick specific genre suitable for the mood requirement of the hour.
up music they are most familiar with and currently listening to in 4.3.3 Lack of trusted and accessible sources. Participants high-
order to avoid the pain of selection on the smaller screen of mobile lighted trust as an influence in their selection of novel music to
devices. listen to. They relied on sources such as friends whom they believe
P4: “Cause I only have a few songs or albums have good tastes in music and media they follow to help them dis-
downloaded on my phone I listen to them (the cover more novel music (P4: “Finding new song can be hard if don’t
artist) frequently on the bus or gym and that is have the right source.” ). A few participants (P2, P3) also mentioned
one of them.” their frustration with lack of trusted sources, such as BBC Radio
One, on their preferred music listening platforms.
Participants also mentioned the role of mood in their selection
of music, such as upbeat music when happy versus melancholic or
5 DISCUSSION
calm music when occupied with a tedious task [P3, P6, P7].
We conducted a contextual-inquiry study to gain insights into how
P7: None of the songs are favorites, but chose the a small group of participants chose novel and familiar music, what
album due to the mellow and calming nature of factors affect their choices, and the challenges they faced while
the music. seeking novel versus familiar music. In this section, we summa-
rize the insights from the study and discuss the possible design
P3: Would pick up this artist or this type of music implications for effective recommendations.
(90s Alternative Rock) when feeling melancholic,
as they remind me of teenage years. 5.0.1 Balance of Effort, Risk, and Attention. Participants men-
tioned the effort involved in searching for novel music. This effort
Overall, the setting during the session or the context played a contrasts with the comfort they mentioned in continuing to listen
critical role in participants’ choice of music. Most of the participants to their current music selection, or familiar music. To alleviate some
mentioned a specific choice of music under certain settings, such as of the effort of searching for novel music, recommendation systems
fast and upbeat music they are familiar with during workout and aim to introduce novel items directly into their lists of recommen-
gym sessions or exploring new and trending releases during the dations. However, in discussing the effort of searching for novel
evening. music, participants also mentioned two critical factors of this effort
that are often overlooked in the design of recommenders.
4.3 Challenges First, the risk appetite of individual users. Some participants
We asked participants the explicit challenges they face when they mentioned greater appetites to explore newer unknown options
look for music to listen to that is different from what they are than others. In our own results in studying the appetite of users for
currently listening to. We highlight three of these challenges that novel items in Chapter 2, we show that a recommender adaptive to
were commonly expressed in participant responses. the individual differences in novelty consumption is more accurate
4.3.1 Too many options. Music recommenders have evolved in than a traditional one-size-fits-all recommender.
many ways to help users choose music they want to listen to. Spo- Second, the potential attention needs of users. Participants cited
tify, for instance, suggests songs curated into multiple categories how familiar music helps maintain their focus when attention re-
for users to start listening to based on time of day, mood, genre, quirements for the primary task at hand are high, as well as how
recent releases, trending, etc. The multitude of choices are appreci- exploring novel music is avoided to minimize the interruptions
ated by users as they help cater to different needs [16]. However, in their focus-intensive tasks. This aligns with prior studies that
participants in the study mentioned that with plenty of options, have shown that interruptions from peripheral tasks such as music
finding what to listen to becomes harder and that “..it takes a lot listening have huge impacts on the primary task at hand, result-
of energy to find something new to listen to.. - P2”, while wading ing in needing more time to complete the task, committing more
through the myriad of choices. The high effort involved in choosing errors in the task, and experiencing more annoyance and anxi-
among available options was cited as a deterrent to the desire of ety [4, 5]. Studies suggest that delaying such interruptions towards
choosing any new music they have not heard of — “Does not want the phases between the primary tasks causes less disruptive impact.
to spend limited amount of time I have to a song that I have not heard In music listening, these phases could be the intermittent interac-
of” - P3. tions users have with the music service, that is, when users are
4.3.2 Risk of Failure. Participants mentioned the risk involved distracted from their primary tasks. It is at these times that recom-
in choosing something new. mendation systems could introduce or suggest novel items in lists
of recommendations.
P1: “Discovered 3 brand new albums in 1 week 5.0.2 Boredom versus Sheer Joy of Adventure. Participants cited
but only one stuck around. Tried looking for more boredom of their current selection or the sheer joy of adventure
popular Jazz Rock but haven’t found them inter- in exploration of novel music, that is, music different from their
esting.” current selection. However, we noted distinct differences in the
Participants (P6, P7) mentioned the risk of mismatch of their participants’ expertise who sought novel music to avoid boredom
selection with their desired mood as another reason they avoided versus those who sought novel music for adventure. The partici-
selecting new music. In such cases, users preferred listening to pants who mentioned the joy of adventure (P1, P2, and P4) were
IntRS Workshop, October, 2018, Vancouver, Canada Kumar et al.
arguably the ones who take their music seriously, as their interest in willingness to participate in this study, participants are likely more
music went beyond just listening. These were the users who create, comfortable expressing their thoughts than the rest of the general
curate, share, and consume music with others. Also, some of the population. This characteristic of participants could differentiate
artists listened to by these participants were found to be more ob- their preferences for novel and familiar music from the rest of the
scure (ex: Jagga Jazzist with about 56 thousand listeners on Spotify) general population. Finally, their recall accuracy of whether they
than mainstream music. In comparison, the group of participants had listened to specific tracks or artists in the past month could
who cited boredom (P3, P5, P6, and P7) showed more interest in impact the accuracy in determining familiar and novel music.
mainstream music, with preferences for artists like Kendrick Lamar, Second, we discuss limitations with the study design. The mu-
who has about 36 million listeners on Spotify. The latter group of sic language was chosen to be English, limiting any cross-cultural
participants cited reasons to find or explore different music to pri- comparison or inference from the results. Also, there is a likelihood
marily manage mood, avoid boredom, or help achieve focus when of implicit bias due to the selective nature of recruitment that could
distractions surround them. Thus, these different levels of expertise limit the generalizability of the themes across the general popula-
can help systems determine a more accurate appetite for novelty tion. Finally, the number of participants is a small representation of
for individual recommendations. For instance, the tendency of a a wide and diverse range of music listeners and prohibits us from
participant to continue with their choice of music before they again generalizing to the larger population. However, since the study is
felt the urge to shake things up was more evident in the second exploratory in nature, we do not expect this to harm the external
group of participants, whereas the participants who cited the joy validity of our findings and recommendations with a view to inspire
of adventure mentioned seeking novel music more frequently. future work.
5.0.3 Perceived Familiarity, Trust, and Genre of Novel Choices. 6 CONCLUSION
Once users decide that they want novel music, the question that still Recommender systems have become ubiquitous in many online
remains is how do recommenders pick from the plethora of avail- systems, helping users discover both new and forgotten items. As
able options? The participants’ responses highlight three distinct systems grow and more diverse users join systems, it is becoming
insights. First, participants’ likelihood of exposure to the music more crucial to understand the structure and intention of user-
from external sources. Participants mentioned friends, music con- specific needs to provide an engaging and satisfying experience.
certs, social media, and TV shows as some of the sources that led to We conducted a contextual inquiry-based study to understand
their selection of novel music. This is related to the phenomena of participants’ actions and intentions while they seek novel or fa-
mere exposure that results in an aroused curiosity and a perceived miliar music in online music streaming platforms. We observed
familiarity towards previously unheard or unknown items [35]. participants while they listened to music in their everyday settings
Second, participants interested in the discovery of new music and followed up with interviews to expand on factors, such as at-
cared about trust in their sources of music. Participants were clear tention, effort, trust, boredom, and risks, that play a major role
that not every friend, event, or blog is influential in their choices in the users’ choices of novel or familiar music. We identified the
of novel music and that trust and reputations of sources play a challenges participants faced, such as a lack of trustable sources,
crucial role. They mentioned that friends who they believe have an overwhelming number of choices, and the risk of a bad choice,
good tastes in music, celebrities who they like, and music blogs that that drive users to stay within the comforts of familiarity and avoid
they trust, such as BBC Radio One, often led them to explore new uncertain risk-rewards of novelty.
artists. Participants’ emphasis on trust and reputation highlights In order to design effective recommenders, we discussed the
why some new artists were more preferred than others. results and design implications to emphasize the gap in the assump-
Third, the genre of novel music plays a critical role. On the one tions imposed by traditional algorithms on user-specific needs in
hand, the group of participants who mentioned boredom as a cause seeking novel and familiar items. Our results emphasize the goal of
to seek novelty sought to change the genre of music (including lan- recommender algorithms to explore user needs beyond explicit and
guage for multilingual participants) from what they were currently implicit interactions and include in the models the likelihood of the
listening to. However, other participants who sought novel music attention needs of the user, the risk appetite of each user, and the
for sheer joy mentioned exploring new releases in the same genre types of novel music users consume in their sessions. Finally, while
they were currently listening to. The primary difference between this work focuses specifically on music and with limitations on
these participants from the previous group was how important they the number of qualitative observations, our findings speak to the
consider music in their daily consumption. Understanding these challenges in mapping user needs for content providers in multiple
differences in user consumption therefore can help recommender domains such as news, movies, books, etc.
systems identify if novel music to be recommended is required to
belong to a different genre than the user is already listening to. 7 ACKNOWLEDGEMENT
5.1 Limitations We would like to thank participants who agreed to share their
While we summarize the observations and implications from par- time and experience in the study. We also thank the anonymous
ticipants’ responses, we also recognize the limitations of this study. reviewers for their valuable feedback and inputs.
First, we discuss limitations having to do with the participants.
Being under observation could have possibly affected the attention
REFERENCES
[1] Donna Rose Addis, Morris Moscovitch, Adrian P Crawley, and Mary Pat McAn-
of participants who might have selected differently without an drews. 2004. Recollective qualities modulate hippocampal activation during
individual overlooking their picks. Also, by the nature of their autobiographical memory retrieval. Hippocampus 14, 6 (2004), 752–762.
IntRS Workshop, October, 2018, Vancouver, Canada
[2] Gediminas Adomavicius and Alexander Tuzhilin. 2011. Context-aware recom- [28] Komal Kapoor, Vikas Kumar, Loren Terveen, Joseph A Konstan, and Paul Schrater.
mender systems. In Recommender systems handbook. Springer, 217–253. 2015. I like to explore sometimes: Adapting to dynamic user novelty preferences.
[3] Ashton Anderson, Ravi Kumar, Andrew Tomkins, and Sergei Vassilvitskii. 2014. In Proceedings of the 9th ACM Conference on Recommender Systems. ACM, 19–26.
The dynamics of repeat consumption. In Proceedings of the 23rd international con- [29] Komal Kapoor, Nisheeth Srivastava, Jaideep Srivastava, and Paul Schrater. 2013.
ference on World wide web. International World Wide Web Conferences Steering Measuring spontaneous devaluations in user preferences. In Proceedings of the
Committee, 419–430. http://dl.acm.org/citation.cfm?id=2568018 19th ACM SIGKDD international conference on Knowledge discovery and data
[4] Brian P Bailey and Joseph A Konstan. 2006. On the need for attention-aware mining. ACM, 1061–1069.
systems: Measuring effects of interruption on task performance, error rate, and [30] Komal Kapoor, Karthik Subbian, Jaideep Srivastava, and Paul Schrater. 2015.
affective state. Computers in human behavior 22, 4 (2006), 685–708. Just in time recommendations: Modeling the dynamics of boredom in activity
[5] Brian P Bailey, Joseph A Konstan, and John V Carlis. 2000. Measuring the effects streams. In Proceedings of the Eighth ACM International Conference on Web Search
of interruptions on task performance in the user interface. In Systems, Man, and and Data Mining. ACM, 233–242.
Cybernetics, 2000 IEEE International Conference on, Vol. 2. IEEE, 757–762. [31] Tuck W Leong and Peter C Wright. 2013. Revisiting social practices surrounding
[6] John A Bargh. 2006. What have we been priming all these years? On the devel- music. In Proceedings of the SIGCHI conference on human factors in computing
opment, mechanisms, and ecology of nonconscious social behavior. European systems. ACM, 951–960.
journal of social psychology 36, 2 (2006), 147–168. [32] Lorelei Lingard, Mathieu Albert, and Wendy Levinson. 2008. Grounded theory,
[7] Frank Bentley, Crysta Metcalf, and Gunnar Harboe. 2006. Personal vs. com- mixed methods, and action research. Bmj 337, aug07_3 (2008), a567–a567.
mercial content: the similarities between consumer use of photos and music. In [33] Adam J Lonsdale and Adrian C North. 2011. Why do we listen to music? A uses
Proceedings of the SIGCHI conference on Human Factors in computing systems. and gratifications analysis. British Journal of Psychology 102, 1 (2011), 108–134.
ACM, 667–676. [34] Elizabeth Hellmuth Margulis. 2014. On repeat: How music plays the mind. Oxford
[8] Daniel E Berlyne. 1970. Novelty, complexity, and hedonic value. Perception & University Press.
Psychophysics 8, 5 (1970), 279–286. [35] Richard L Moreland and Robert B Zajonc. 1982. Exposure effects in person
[9] Patrick Burkart and Tom McCourt. 2006. Digital music wars: Ownership and perception: Familiarity, similarity, and attraction. Journal of Experimental Social
control of the celestial jukebox. Rowman & Littlefield. Psychology 18, 5 (1982), 395–415.
[10] Sunny Consolvo, Frank R Bentley, Eric B Hekler, and Sayali S Phatak. 2017. [36] Jeremy Wade Morris and Devon Powers. 2015. Control, curation and musical
Mobile user research: A practical guide. Synthesis Lectures on Mobile and Pervasive experience in streaming music services. Creative Industries Journal 8, 2 (2015),
Computing 9, 1 (2017), i–195. 106–122.
[11] Paul Covington, Jay Adams, and Emre Sargin. 2016. Deep neural networks [37] Daniel Nordgård. 2016. Lessons from the worldâĂŹs most advanced market
for youtube recommendations. In Proceedings of the 10th ACM Conference on for music streaming services. Business innovation and disruption in the music
Recommender Systems. ACM, 191–198. industry (2016), 175–191.
[12] Vivian Darroch. 1982. BIOGRAPHICAL NARRATIVE AS THE EXPRESSION OF [38] Adrian C North, David J Hargreaves, and Susan A O’Neill. 2000. The importance
EXISTENCEl. Interpretive human studies: an introduction to phenomenological of music to adolescents. British Journal of Educational Psychology 70, 2 (2000),
research (1982), 215. 255–272.
[13] Michael D. Ekstrand, F. Maxwell Harper, Martijn C. Willemsen, and Joseph A. [39] Kentaro Oba, Madoka Noriuchi, Tomoaki Atomi, Yoshiya Moriguchi, and Yoshi-
Konstan. 2014. User Perception of Differences in Recommender Algorithms. In aki Kikuchi. 2015. Memory and reward systems coproduce âĂŸnostalgicâĂŹ-
Proceedings of the 8th ACM Conference on Recommender Systems (RecSys ’14). ACM, experiences in the brain. Social cognitive and affective neuroscience 11, 7 (2015),
New York, NY, USA, 161–168. DOI:http://dx.doi.org/10.1145/2645710.2645737 1069–1077.
[14] Patricia J Flowers. 2001. Patterns of attention in music listening. Bulletin of the [40] Martin J Packer. 1985. Hermeneutic inquiry in the study of human conduct.
Council for Research in Music Education (2001), 48–59. American Psychologist 40, 10 (1985), 1081.
[15] Walter Gantz, Howard M Gartenberg, Martin L Pearson, and Seth O Schiller. 1978. [41] Andre Paine. 2009. Spotify Hit by Licensing Restrictions. Billboard, January 29
Gratifications and expectations associated with pop music among adolescents. (2009).
Popular Music & Society 6, 1 (1978), 81–89. [42] Marta García Quiñones. 2007. Listening in shuffle mode. Lied und populäre
[16] Mark Glantz. 2016. Internet Radio Adopts a Human Touch: A Study of 12 Kultur/Song and Popular Culture (2007), 11–22.
Streaming Music Services. Journal of Radio & Audio Media 23, 1 (2016), 36–49. [43] Peter J. Rentfrow. 2012. The role of music in everyday life: Current directions
[17] Barney G Glaser and Anselm L Strauss. 2017. Discovery of grounded theory: in the social psychology of music. Social and Personality Psychology Compass 6,
Strategies for qualitative research. Routledge. 5 (2012), 402–416. http://onlinelibrary.wiley.com/doi/10.1111/j.1751-9004.2012.
[18] David J Hargreaves and Adrian C North. 1997. The social psychology of music. 00434.x/full
Oxford University Press. [44] Irving Seidman. 2013. Interviewing as qualitative research: A guide for researchers
[19] Oscar Celma Herrada. 2009. Music recommendation and discovery in the long tail. in education and the social sciences. Teachers college press.
Ph.D. Dissertation. Universitat Pompeu Fabra. [45] John A Sloboda and Susan A OâĂŹNeill. 2001. Emotions in everyday listening
[20] Yoshinori Hijikata, Takuya Shimizu, and Shogo Nishida. 2009. Discovery-oriented to music. Music and emotion: Theory and research (2001), 415–429.
collaborative filtering for improving user satisfaction. In Proceedings of the 14th [46] Nava Tintarev, Christoph Lofi, and Cynthia Liem. 2017. Sequences of Diverse
international conference on Intelligent user interfaces. ACM, 67–76. Song Recommendations: An exploratory study in a commercial system. In Pro-
[21] Karen Holtzblatt and Sandra Jones. 1993. Contextual inquiry: A participatory ceedings of the 25th Conference on User Modeling, Adaptation and Personalization.
technique for system design. Participatory design: Principles and practices (1993), ACM, 391–392.
177–210. [47] Li-Tung Weng, Yue Xu, Yuefeng Li, and Richi Nayak. 2007. Improving recom-
[22] Rong-Hwa Huang and Yi-Nuo Shih. 2011. Effects of background music on mendation novelty based on topic taxonomy. In Web Intelligence and Intelligent
concentration of workers. Work 38, 4 (2011), 383–387. Agent Technology Workshops, 2007 IEEE/WIC/ACM International Conferences on.
[23] Neil Hurley and Mi Zhang. 2011. Novelty and diversity in top-n recommenda- IEEE, 115–118.
tionâĂŞanalysis and evaluation. ACM Transactions on Internet Technology (TOIT) [48] Robert B Zajonc. 2001. Mere exposure: A gateway to the subliminal. Current
10, 4 (2011), 14. http://dl.acm.org/citation.cfm?id=1944341 directions in psychological science 10, 6 (2001), 224–228.
[24] Roberto Ippoliti. 2014. Institutional review board. Encyclopedia of Law and [49] Yuan Cao Zhang, Diarmuid Ó Séaghdha, Daniele Quercia, and Tamas Jambor.
Economics (2014), 1–4. 2012. Auralist: introducing serendipity into music recommendation. In Proceed-
[25] Kurt Jacobson, Vidhya Murali, Edward Newett, Brian Whitman, and Romain Yon. ings of the fifth ACM international conference on Web search and data mining.
2016. Music personalization at spotify. In Proceedings of the 10th ACM Conference ACM, 13–22.
on Recommender Systems. ACM, 373–373. [50] Qian Zhao, Gediminas Adomavicius, F Maxwell Harper, Martijn Willemsen, and
[26] Amalie Enshelm Jensen, Caroline Møller Jægerfelt, Sanne Francis, Birger Larsen, Joseph A Konstan. 2017. Toward better interactions in recommender systems:
and Toine Bogers. 2018. I just scroll through my stuff until I find it or give up: A cycling and serpentining approaches for top-N item lists. In [CSCW2017] Pro-
Contextual Inquiry of PIM on Private Handheld Devices. In Proceedings of the ceedings of the 2017 ACM Conference on Computer Supported Cooperative Work
2018 Conference on Human Information Interaction&Retrieval. ACM, 140–149. and Social Computing.
[27] Iman Kamehkhosh and Dietmar Jannach. 2017. User Perception of Next-Track [51] Cai-Nicolas Ziegler, Sean M. McNee, Joseph A. Konstan, and Georg Lausen. 2005.
Music Recommendations. In Proceedings of the 25th Conference on User Modeling, Improving Recommendation Lists Through Topic Diversification. In Proceedings
Adaptation and Personalization. ACM, 113–121. of the 14th International Conference on World Wide Web (WWW ’05). ACM, New
York, NY, USA, 22–32. DOI:http://dx.doi.org/10.1145/1060745.1060754