UNDERSTANDING USERS OF COMMERCIAL MUSIC
SERVICES THROUGH PERSONAS: DESIGN IMPLICATIONS
Jin Ha Lee
Rachel Price
University of Washington
University of Washington
ABSTRACT
Most of the previous literature on music users’ needs,
habits, and interactions with music information retrieval
(MIR) systems focuses on investigating user groups of
particular demographics or testing the usability of specif-
ic interfaces/systems. In order to improve our understand-
ing of how users’ personalities and characteristics affect
their needs and interactions with MIR systems, we con-
ducted a qualitative user study across multiple commer-
cial music services, utilizing interviews and think-aloud
sessions. Based on the empirical user data, we have de-
veloped seven personas. These personas offer a deeper
understanding of the different types of MIR system users
and the relative importance of various design implications
for each user type. Implications for system design include
a renegotiation of our understanding of desired user en-
gagement, especially with the habit of context-switching,
designing systems for specialized uses, and addressing
user concerns around privacy, transparency, and control.
1. INTRODUCTION
Designing music information retrieval (MIR) systems
such as music recommenders or music management sys-
tems is challenging due to the wide variety of organiza-
tional and listening strategies of music users [3]. Alt-
hough the number of studies on music users, specifically
related to their needs and interactions with MIR systems,
has been increasing since the early 2000s [15], our under-
standing on how to understand and model these users for
system design is still lacking.
Previous studies of MIR system users tend to focus on
investigating needs, perceptions, and opinions of general
users (represented by subjects recruited online or in aca-
demic settings) or specific user groups. Studies involving
specific user groups tend to investigate users based on
particular demographic information or users of particular
MIR systems. However, few studies attempt to categorize
the “personalities” of music listeners surrounding their
interaction behavior on multiple MIR systems. In addi-
tion to demographic information, what kinds of personal
characteristics can we use to model commercial MIR sys-
tem users for system design? Our study aims to fill this
gap in prior research and answer the following questions:
RQ1. What kinds of user personas can we identify
from real users of commercial MIR systems?
RQ2. What are the expressed needs and behavior of
each of these user personas, and what are the implica-
tions for system design for each persona?
Our research will contribute by providing a framework
for understanding users of MIR systems based on their
needs and interaction behavior, beyond typical demo-
graphic information. This will help inform system de-
signers to develop systems that are better targeted for
their user groups representing particular personas, rather
than creating a “one size fits all” mass production model.
2. RELEVANT WORK
2.1 HCI Studies Related to Music
A number of studies in the human computer interaction
(HCI) domain explore different user behavior related to
music discovery or sharing. Most of the literature focuses
on testing the usability of a particular system interface, or
investigating user behavior related to music discovery or
sharing within a particular application.
The literature reflects a growing understanding that
current music listening habits are changing. Voong &
Beale [26] highlight the fact that playlist generation is
done differently now than in the past, whether users cre-
ate playlists by mood, theme, or other criteria. In our re-
search, we aim to understand these criteria that are rele-
vant to users when generating playlists and judging the
playlists created by music services, and how to use those
criteria to influence user experience (UX) design.
The social aspect of music consumption also seems to
be a key area for investigation. Research around social
playlists illustrates how friends can learn more about each
other and can strengthen relationships through under-
standing the preferences of others ([18][21]). Bonhard et
al. [4] further illustrate that “friends from whom we seek
recommendations are not just a source of information for
us: we know their tastes, views and they provide not only
recommendations, but also justification and explanations
for them. (p. 1064)” The impact of new online music re-
positories to people’s music discovery and sharing has
also been discussed in [18].
Some studies looked at the problem of how personality
affects recommenders. Researchers have borrowed theo-
ries from psychology literature about personality, as in
[6], exploring the impact of personality values on users’
needs for recommendation diversity. Their preliminary
research shows a causal relationship between personality
attributes, including openness, conscientiousness, extro-
version, agreeableness, and neuroticism, and users’ diver-
sity preferences when using a recommender system. In
our work, we take a more empirical approach, looking at
© Jin Ha Lee, Rachel Price. Licensed under a Creative
Commons Attribution 4.0 International License (CC BY 4.0). Attribu-
tion: Jin Ha Lee, Rachel Price. “Understanding users of commercial
music services through personas: design implications”, 16th Internation-
al Society for Music Information Retrieval Conference, 2015.
476
user data to understand various types of personas present
in music services users and how the user experience can
be designed to better accommodate these personas.
2.2 User Studies in MIR
Prior studies of MIR system users can be categorized in-
to: 1) empirical investigation of music information needs,
behavior, perceptions, and opinions of humans, 2) exper-
iments, usability testing, interface design involving hu-
mans focusing on a particular MIR system, and 3) analy-
sis of user-generated data such as queries or tags [15].
Of the first category, a few studies focus on “general
music users,” often represented by queries in search en-
gines, or human subjects recruited on various websites or
a game (e.g., [8][17]). A majority of them, however, fo-
cus on a particular group of users based on demographic
information. Several researchers have investigated the
effects of age (e.g., young adults in [14][27]) and nation-
ality [10][12][23]. These studies revealed that age group
and cultural background do affect how people perceive,
use, and search for music. A number of studies also re-
search needs and behaviors of users in specific music-
related professions (e.g., musicologists [2], DJs [20],
film-makers [9]). In order to complement the findings
from these studies, we look beyond demographic infor-
mation and model users based on their goals/behavior
within MIR systems.
A few studies focused on investigating users’ experi-
ences with existing commercial music services, and thus
are more closely related to the current paper. Barrington
et al. [1] and Lamere [13] evaluated the quality of provid-
ed music recommendations or system-generated playlists.
Barrington et al. [1] compare Apple iTunes’ Genius to
two canonical music recommender systems: one based on
artist similarity, and the other on acoustic similarity. They
demonstrate the strength of collaborative filtering com-
bined with musical cues for similarity (similar artists and
other display metadata) and discuss factors that influence
playlist evaluation, such as familiarity, popularity, trans-
parency, and perceived expertise of the system. Lamere
[13] also compares the playlists generated from Google’s
Instant Mix, Apple iTunes, and the Echo Nest Playlist
engine, and notes how personal preference of music or
the context of music can affect the user experience with
music services. Some factors that influence users’ evalua-
tions of playlist (e.g., familiarity, popularity, transparen-
cy) as well as the overall perception of the quality of mu-
sic service (e.g., inexpensiveness, convenience, customi-
zability) were also identified in [1] and [17], respectively.
Celma [5] discusses varied recommendation needs for
four different types of listeners (i.e., savants, enthusiasts,
casuals, indifferents) based on their degrees of interest in
music. Lee & Price [16] also evaluated commercial music
services based on Nielsen’s ten usability heuristics, advo-
cating for more holistic evaluation of MIR systems.
Some studies focused on investigating the factors that
impact people’s music listening or sharing behavior. Baur
et al. [3] analyzed a sample of 310 music listening histo-
ries collected from Last.fm and 48 variables describing
user and music characteristics. They found that temporal
aspects such as seasons and the degree of users’ interests
in novelty were important factors affecting people’s mu-
sic listening behaviors. Additionally, a number of pat-
terns regarding users’ music seeking and consumption
behavior were observed in a large-scale survey [17]: an
increased consumption in mobile streaming services, an
increased desire for serendipitous music discovery and
music videos, as well as a strong desire to customize and
personalize their music experiences.
The scope and approach of our work differ from these
studies on user experience with music services in that we
investigate users of ten different MIR systems (Spotify,
Pandora, YouTube, Songza, SoundCloud, Grooveshark,
Bandcamp, Rdio, Last.fm, iTunes), and we take a qualita-
tive approach, asking questions and observing users’ in-
teractions with MIR systems. Our work aims to build up-
on these studies and provide more detailed information
about how user contexts or characteristics affect actual
usage of music services.
3. RESEARCH DESIGN AND METHODS
Table 1 provides an overview of the methods and activi-
ties used for different phases for this study. The user data
were collected through interviews and think-aloud ses-
sions. All recruited participants were over 18 years old,
and actively use at least one music service/application.
All participants were undergraduate or graduate students
at University of Washington. All the interviews were
conducted between January and March 2014, either in-
person or via Adobe Connect video conferencing. A total
of 40 participants were interviewed and compensated
with a $15 Amazon gift card.
Methods
Activities
User
interview
Semi-structured interview asking about how
participants use music services and how they
evaluate the quality of the services.
Think-
aloud
sessions
Participants narrate their actions out loud as
they use their preferred music service as they
would in a typical session.
Card
sorting
Identify task-based user segments and create
personas for each segment.
Table 1. Overview of the study design
The study session consists of two parts: first, subjects
were interviewed about their preferred music services,
discussing their interactions with the service, how they
navigate the system, why they prefer one service over
others, frustrations they experience with the service, and
how they interact with the service in a typical session.
Secondly, participants were asked to “think-aloud” or
narrate their actions out loud to an investigator as they
use their preferred music service in a typical session.
These tasks include known-item search, browsing al-
bums, artists, or genres, interacting with recommenda-
tions, playlists, and radio stations, and other tasks as they
arose. Each study session, consisting of the interview and
think-aloud, lasted for approximately an hour.
The user data was used to generate a list of behaviors
exhibited around MIR systems. A card sorting activity
Proceedings of the 16th ISMIR Conference, M´alaga, Spain, October 26-30, 2015 477
was used to identify user groups with similar behaviors as
a basis for deriving useful personas. Personas are “hypo-
thetical archetypes of actual users (p.124)” representing
their needs, behavior, and goals which allows for a goal-
directed design of a system [7]. Persona development has
been used to aid design and gain user insights across
many fields [22], and can be beneficial for prioritizing
audiences’ and users’ goals in product development [24].
We created a comprehensive list of user activities from
the interview transcripts and think-aloud activities as well
as the notes taken during observation. A total of 77 user
behaviors related to music services were identified (e.g.,
read reviews, judge others’ tastes, seek recommenda-
tions). Through a card sorting activity, similar behaviors
were grouped, organized, and named. We then attempted
to identify which types of users would show these kinds
of behaviors and tentatively named these user groups
(e.g., genre fans, tech savvy). Afterwards, we identified
two relevant dimensions to express the differences among
these user groups organized by their common behavior,
or “task-based audience segments” [28]: Companionship
(willingness to engage in social aspects of music recom-
mendation and listening: social - neutral - private) and
Investment (willingness to invest time/effort to interact
with the system: positive - neutral - none). As a result, we
derived these seven personas:
Active Curator: Neutral companionship + Positive
investment
Music Epicurean: Social + Positive investment
Guided Listener: Neutral companionship + No in-
vestment
Music Recluse: Private + Neutral investment
Wanderer: Neutral companionship + Neutral in-
vestment
Addict: Private + No investment
Non-believer: Social + Neutral investment
Any user may exhibit a combination of these personas
as they are not mutually exclusive. Each of these per-
sonas is explained in detail in the following section.
4. USER PERSONAS
4.1 Active Curator
This persona takes great pride in their music listening,
and enjoys seeking new music and curating music he/she
is already familiar with. This may come in the form of
playlist creation, “saving” albums in online collections, or
light music “research”, such as previewing songs or tak-
ing recommendations from friends, blogs, and live shows.
Of all the personas, this one is the most actively engaged
with music services (“I’m definitely an active listener
98% of the time.” (P21)).
This persona tends to utilize known-item search along-
side other discovery tools, often searching rather than
browsing (“I [search for song or artist] at least once a
day.” (P26)). An active curator may often find discovery
tools to be disappointing (“I feel like I end up listening to
stuff I already know. It’s a little frustrating” (P1)). They
tend to have higher expectations for music recommenda-
tion services and may not always trust a service to make
good recommendations.
“One of the reasons I use these services is because I’m
looking for linkages from music to music to music…I’m a
little bit pedantic...In fact, I would love to have a little bit
more information [about recommendations].” (P1)
“I would love to see the metadata that goes into choosing
each song…[I’d love to] be able to pick and choose those
attributes, so I could say, ‘ok, I do like those smooth jazz
elements, but I don’t like the saxophone solos.’” (P30)
4.2 Music Epicurean
This persona may be considered a “music snob.” Music
epicureans take an immense amount of pride in the music
they collect and listen to, although they may not neces-
sarily own all that music. Although streaming music is
still an acceptable form of listening, this persona is more
inclined to purchase music after listening to it than other
personas as he/she genuinely cares about sound quality. A
great amount of time is spent “hunting” for new music.
This persona tends to focus on relationships between
bands that may not be typically identified by a music rec-
ommender, such as similar “scene”, overlapping band
members, and a nuanced understanding of genre relation-
ships, and thus expresses dissatisfaction towards the giv-
en recommendations (“It looks like it’s only making rec-
ommendations of artists based on artists.” (P23)).
The music epicurean persona is unlikely to use music
system recommendations; users representing this persona
tend to also represent “The Non-believer” persona de-
scribed below. The Music Epicurean leans on trusted
sources for recommendations, whether it is a small group
of friends with trusted taste or other “vetted” sources.
“I’m very self-directed in listening to music. When I listen
to the radio, it’s KEXP, and it’s usually a really short
amount of time in the morning. I know what I want to listen
to, why am I just going to let a random radio station tell
me?” (P8)
“For me it’s not really worth the time. I think it’s just going
to recommend stuff that’s also tagged [similarly]...I do my
own ways of [finding], and I rely on my friends and people
I write with to recommend stuff...” (P6)
4.3 Guided Listener
The Guided Listener’s most prominent quality is the de-
sire to hand over control of the music to someone else.
This persona mildly enjoys radio’s serendipitous nature,
may have slight preferences over genre or artist, but ulti-
mately just wants to hear something playing. This perso-
na is not picky; he/she may occasionally interact with a
service to indicate preferences or dislikes but will not go
out of his/her way to curate albums or playlists.
This persona may provide “seed” songs or artists to
help a system generate a playlist or radio station, and in-
frequently, will browse new music or artists for fun or out
of boredom. For the most part, the guided listener is a
“set it and forget it” kind of person.
“It’s definitely ‘log in’, get to where I’m going, and it even
goes back to the default station that I was listening to before.
478 Proceedings of the 16th ISMIR Conference, M´alaga, Spain, October 26-30, 2015
I mean, I can get this thing booted up and going within
seconds, and then I’m off doing dishes or whatever, which
contributes to my satisfaction. It’s going to do what I want
it to do immediately. Boom. Off I go.” (P17)
4.4 Music Recluse
The primary characteristic of the Music Recluse is that
he/she is a very private listener; this persona does not
need to discuss his/her music listening habits with many
people, and guards his/her privacy when using a music
recommendation service. The music recluse actively
avoids the social functions of music services like Spotify
or Pandora and considers listening to be very personal.
This persona may have sporadic listening habits, may lis-
ten to music he/she is not proud of or would not want others
to know about. Music recluses do not want people making
assumptions about them based on the music they listen to.
“I would allow zero information. I already think YouTube
is too invasive. They’re already forcing users to create
Google Plus accounts to comment on videos.” (P25)
“I turned off sharing functionality. I made sure that I
wasn’t putting it up on Facebook or sharing it...I definitely
listen to a lot of embarrassing stuff and I don’t want every-
body to know that. And I’m not really part of musical com-
munities or anything, so I don’t feel like scrolling through
my friends’ music gives me any useful information or
songs to listen to.” (P34)
4.5 Non-believer
The non-believer is a persona who does not believe that a
machine can make adequate music recommendations for
a variety of reasons: they do not understand how an algo-
rithm can make “good” recommendations, they are able
to see the limitations of recommendation algorithms, they
prefer getting recommendations from friends, or they
simply have not had good past experience with music
recommendation services. Non-believers also have a ten-
dency to dislike sharing personal information or listening
histories with the service/system because they do not see
the benefit of doing so. This persona often uses human-
curated music services such as Songza or 8-Track,
friends’ playlists, or their own collections, which may or
may not be heavily curated.
“Pandora will give me mainstream blues because it’s simi-
lar rhythmically and in instrumentation, but that’s not the
vibe I’m looking for. It seems like they go off of something
really mechanical. They’re missing out on something and I
don’t know what it would be called, like context, and how
the music makes me feel.” (P23)
4.6 Wanderer
The Wanderer primarily enjoys serendipitous music dis-
covery, and listens to new music with an open mind
(“…when it recommends me things that I never would have
thought of, so I think, ‘yeah, I’ll give it a shot’.” (P11)).
This persona enjoys the discovery process in general as a
fun pastime, and is willing to put in some effort to dis-
cover new music. The wanderer will likely accept rec-
ommendations from a system as equally as she will ac-
cept them from a friend, a blog, or a stranger.
The wanderers tend to listen to music from a wider va-
riety of music genres, although they may also have pre-
ferred favorites. They enjoy discovering music/artists that
are less popular and are willing to listen to new artists or
genres. Wanderers may like recommendations based on
“playful” themes such as “Monday morning” or “Coffee
music.” They are more likely to use a variety of tools and
also new features in the tools they regularly use.
“Honestly, the serendipity of finding new music is what I
enjoy the most. Generally if I’m listening to new music it
will be because a friend recommended it or I came across it
on YouTube through NPR Tiny Desk or something like that.
I prefer that model...I listen to pretty diverse things.” (P13)
4.7 Addict
The Addict exemplifies a known-item searcher and
strongly utilizes a service that features search. This per-
sona may listen to the same song multiple times in a row,
or for a whole week (e.g., “I sort of fixate.” (P1)). This
persona tends to use services like YouTube or Spotify
where it is easy to repeat albums or songs. Their musical
tastes may be all over the map, and they tend to listen to
things on a whim, rather than curating any collections.
They may listen sporadically, for short periods of time,
and rely on easy access to music (web-based) from a va-
riety of devices. The addict typically does not save
his/her preferences by creating playlists for later access.
“I prefer Grooveshark…because I have a tendency to listen
to a song, and then listen to it on repeat until I hate it for-
ever, and Pandora doesn’t let you do that at all, whereas in
Grooveshark you can do that.” (P23)
5. THEMES AND DESIGN IMPLICATIONS
5.1 Engagement, Ownership, and Specialization
Our user data suggest that we may need to rethink the
concept of “engagement” and how that affects peoples’
preferences for music services. If we consider engage-
ment as users interacting with the system by exploring
available features, then while it may be counter-intuitive,
some users have no desire to engage with their preferred
system. The way these users measure the success of the
system is based on how little they have to interact with it.
“As soon as I figured out the basics...as soon as I found
that I could look at some friends’ playlists, and that I could
find a few artists and make a radio station, I just, I was like,
I’m done. I’m done learning how to make this work.” (P1)
“There’s nothing I don’t like about Pandora...It might just
be because I’m content enough...And I think I’m old
enough, you know, I’m 45, I'm not into that 'music is my
world' type of mentality. So it’s not high on my list.” (P17)
A strong satisficing theme was identified among these
users, consistent with the finding in [16]. As long as the
system does what it is “supposed to do”, then it is “good
enough” and users do not expect much more. This is es-
pecially exhibited by participants representing the “guid-
ed listener” persona, who tends to prefer music services
like Pandora. The “addict” also tends to exhibit shallow
engagement with the services. During the interview, it
Proceedings of the 16th ISMIR Conference, M´alaga, Spain, October 26-30, 2015 479
became evident that most participants who can be catego-
rized as guided listeners had never gone beyond the sur-
face level of system. In fact, many participants discov-
ered some of the features offered by their preferred ser-
vice for the first time during the think-aloud sessions.
They tend to have very specific needs and do not explore
the service beyond their immediate needs.
Personas such as active curator and music epicurean
showed higher levels of engagement with the systems and
seemed to have a stronger sense of ownership over their
music collections. Active curators in particular would
spend much time curating playlists even though they do
not technically “own” the music. While guided listeners
would most likely be satisfied with a streaming or sub-
scription-based model, active curators and music epicure-
ans hesitate to abandon the collection-based model. For
this reason, we expect that cloud-based music services
will appeal more to the latter group of users. For them,
providing a way of creating their own access points into
their collection will become an important issue, as the
size of their collection will continue to grow. Organizing
and accessing their collection by play frequency, name of
the person who recommended a track, release date, or us-
er in households where multiple members share the music
service, were some examples that respondents specifical-
ly mentioned as potentially useful.
In order to meet the needs of different personas, it may
make sense to release different versions of the ser-
vice/app so users can decide the appropriate version
based on how much interaction they desire (“If [Spotify]
had a light version then I would use that more. Like
iTunes had a little mini-player, for example.” (P13)).
Based on general observation, it does seem like speciali-
zation works better than generalization; each service def-
initely tends to attract particular types of user personas.
For example, Pandora tends to attract users who do not
want to spend time and effort curating music collections
or listening experiences. On the other hand, Spotify users
tend to invest more time in organizing their collections
and providing input to improve their listening experience.
Although users also rely on Spotify for music recommen-
dations, they tend to be more critical about the results due
to higher expectations. Websites like YouTube also serve
a specific purpose, which is to stream videos, rather than
attempting to work as some sort of Web portal that offers
a variety of services. Many users, especially with need for
known-item searches, will go to YouTube. Users’ strong
desire to customize and personalize their music experi-
ences was also noted in [17].
5.2 Awareness and Preserving User Trails
Another theme emerged around a user’s general aware-
ness within a system. Most users expressed a habit of
“digging” and following “wormholes” while using mid-
to high-level curation tools such as Spotify, Grooveshark,
and YouTube. Many of these systems do a poor job of
indicating the user’s location within the site, or helping
them retrace their steps, which often results in users feel-
ing the sense of “being lost.”
“It’s constant digging. Click, click, scroll...wait, where am
I? Click, scroll. For almost everything I want to do, I can
never get there on the first try, or even if I get there on the
first try, it feels like an accomplishment. Most of the time, I
have an idea of where I am, but I don’t always know how to
get back to where I was.” (P11)
“I feel like I’m not as adventurous in wormholing sometimes
as I can be or want to be, because I’m afraid of getting lost.
If it were a little bit easier to just go back to where you
started from or some sort of chain-of-command of what
you had just done that you could click through (like a
breadcrumb trail), then I probably would feel a little bit
more comfortable.” (P3)
This was also related to the general lack of error ex-
planation in the systems, which would ideally help users
recognize and prevent errors (“It just says, ‘There was an
error.’ I almost never know what's going on when some-
thing goes wrong.” (P11)).
Users who discussed digging, wormholes, and the like,
tended to be those who actively engaged with the service.
This may span across any persona, but there appears to be
a correlation between concern for user trails and engaged
personas like the active curator and the music epicurean.
Ideally the system should support the expression and
preservation of a user trail and use breadcrumb trails to
give users locational clues.
Users also indicated that more transparency over rec-
ommendations would improve their likelihood of trusting
the system. Not knowing why the system wants them to
listen to a particular song made them less inclined to fol-
low the recommendation, especially for the active cura-
tor, non-believer, and music epicurean personas.
“Sometimes I wonder why things are on there. I guess I
need more insight on why I should choose to click on this
thing...if it’s a band I’ve never heard of, I’m not going to
click on it unless there’s a reason for me to...A lot of times
it’s like, ‘You listened to this song by Rihanna once. All of a
sudden we think you should listen to Justin Bieber.’ That
doesn’t work for me.” (P31)
5.3 Privacy Concerns
Several participants discussed privacy concerns around
using music services. Our data suggest that the levels of
privacy concerns are possibly affected by the following
three factors: a) user’s interest in/belief of a machine’s
ability to accurately recommend music, b) level of under-
standing of privacy issues, and c) overall tech savviness.
A user who has a higher interest in/belief of a machine’s
ability, a better understanding of privacy issues, and is
more tech savvy, tended to be more concerned about
sharing their personal information. This trait was exhibit-
ed across personas regardless of music listening habits,
and most dominantly in non-believers.
“When you download the software, the automatic preference
is that Spotify will open every time you turn on your com-
puter. I don’t like that. The first time I ever downloaded
Spotify, that was the reason I didn’t use it [right away]. I
felt like it was hijacking my computer.” (P1)
“I wouldn’t want to give a system more information about
me even if it would provide a perfect playlist, because I still
480 Proceedings of the 16th ISMIR Conference, M´alaga, Spain, October 26-30, 2015
want to have control of that [information]…It’s creepy…I
like having some degree of control and privacy.” (P13)
“I'm split between 'that's really cool' and 'that's kind of
creepy'. If I had the option to control it then that might be
something I accept.” (P30)
Being transparent about information collection and allow-
ing more user control over privacy may help alleviate fears.
This desire for control was also observed in [11], where us-
ers wanted to be in control of logging what they considered
as the most private information. They found that “users pre-
fer sharing some information automatically such as listening
history, sharing some information at will and keeping some
information private (p. 171)” [11]. This aligns with concerns
that arose during our interviews about privacy of infor-
mation or activities. While listeners may be willing to share
listening history, either discretely or publicly, those same
users may be concerned about other information being
shared without their knowledge.
In addition to “what” is being shared, two other aspects
worth noting are the different reaction to “who” is accessing
users’ personal information and the directionality in sharing
information. There seemed to be a distinction between keep-
ing private information from the system versus from other
people. Users exhibiting the music recluse persona, for in-
stance, were much more concerned with the latter aspect.
Music epicureans seemed interested in sharing their music
listening history in a limited social circle (“I talk to about
five people who like the same music as me. I just feel weird
about posting videos on Facebook like ‘Listen to this’.”
(P31)). Also a number of users acted like “lurkers” in that
they wanted to see what other people listen to but did not
want to share their own listening habits with others.
During our work identifying the personas, we initially
thought there may be a persona “Public broadcaster,” some-
one who is very social and publicizes his or her listening
choices. Careful examination of the transcripts, however,
revealed that none of the users interviewed were “public
broadcasters” themselves, but many made mention of that
characteristic in friends or acquaintances who also use digi-
tal music services. Most of the comments alluding to the ex-
istence of this persona described how people have seen this
kind of “broadcasting” behavior on social media (and were
often annoyed by it). We believe that this persona may still
exist, as previous research such as [11] found that their users
were willing to share and seek shared information such as
music listening habits, and some were already publicly do-
ing so on websites such as Last.fm. Although users did want
to keep some information private, music listening history
was not such information. However, it may also be the case
that we are simply seeing other’s music listening history be-
cause of the default setting in some music services to public-
ly share such information, and as previously discussed,
many users do not spend much time trying to master their
service’s feature settings. We plan to further explore this
through a survey with a larger number of music service us-
ers.
5.4 Context-switching
In addition to the different personas, the user’s context
seemed critical in determining which services they use.
“It really depends. If I’m upstairs in the office and coding
data, I generally listen to music that I already know and
like, because I don’t want it to take my focus away. If I am
taking my dog for a walk or going for a drive, I may use the
recommendations just to listen to new songs.” (P13)
This resonates with previous MIR studies discussing
how perceived qualities of music are affected by the con-
text of the user [19], and how mood, activities, and social
context among other factors influence music perception
[25]. There were several aspects of user’s context that
seemed particularly relevant:
1) Level of attention: This was often dependent on
other activities in which users were concurrently engaged
(e.g., driving or working).
2) Level of energy/motivation: This is closely related
to users’ willingness to interact with the system. General-
ly, tech savvy music listeners were more willing to do so,
but depending on the time of the day, this also seemed to
change (e.g., acting passively while fatigued after work).
3) Mood: The user’s mood constantly changes based
on different events he/she is experiencing, and thus, the
user may want to listen to songs with different “feels”.
4) Temporal aspect: This can be seasonal or about the
time of day. Depending on work schedules, the early
morning or evening may be the best time for users to in-
teract with a system. Seasonality also means that users
are engaged in different activities or in seasonal moods.
User needs appear to continually shift depending on
these contextual elements. A system allowing context-
switching based on a combination of system logs of geo
data, device usage, etc. (for attention level and temporal
aspects) and users’ input (for level of energy/motivation
and mood) would be desirable.
6. CONCLUSION AND FUTURE WORK
In this paper, we present seven personas surrounding the
use of commercial music information systems, derived
from user interview data and observation of use sessions.
These personas, each representing specific traits and atti-
tudes of users, will be helpful in designing music infor-
mation systems that are more highly tailored to specific
user groups. Analyzing the user data made it clear that
there is a relationship between persona placements on
spectrums and types of services preferred. For instance, a
user who is an active curator and music recluse would be
more likely to use a “fringe” service such as Songza,
whereas a guided listener user would likely end up rely-
ing on an online radio service like Pandora. Based on us-
ers’ opinions and observations of their interactions with
the services, we discussed several design implications.
In our future work, we plan to expand this study and
test the applicability of these personas with a larger user
population since they were derived from a relatively
small sample. We will verify our results obtained from a
qualitative approach by surveying a larger number of us-
ers to identify appropriate personas reflecting their char-
acteristics, using a stratified user sample based on their
most preferred commercial music service.
Proceedings of the 16th ISMIR Conference, M´alaga, Spain, October 26-30, 2015 481
7. REFERENCES
[1] L. Barrington, O. Reid, and G. Lanckriet: “Smarter
than Genius? human evaluation of music
recommender systems,” Proc. ISMIR, pp. 357-362,
2009.
[2] M. Barthet and S. Dixon: “Ethnographic
observations of musicologists at the British Library:
implications for music information retrieval,” Proc.
ISMIR, pp. 353-358, 2011.
[3] D. Baur, J. Büttgen and A. Butz: “Listening factors:
a large-scale principal components analysis of long-
term music listening histories,” Proc. CHI ’12, pp.
1273-1276, 2012.
[4] P. Bonhard, C. Harries, J. McCarthy and M. A.
Sasse: “Accounting for taste: using profile similarity
to improve recommender systems,” Proc. CHI ’06,
pp. 1057-1066, 2006.
[5] Ò. Celma: “Music recommendation and discovery in
the long tail,” Ph.D. dissertation, Dept. Information
& Communication Tech., UPF, 2008.
[6] L., Chen, W. Wu and L. He: “How personality
influences users’ needs for recommendation
diversity?” Proc. CHI EA ’13, pp. 829-834, 2013.
[7] A. Cooper: The inmates are running the asylum.
Sams, Indianapolis, 1999.
[8] D. P. W. Ellis, B. Whitman, A. Berenzweig and S.
Lawrence: “The quest for ground truth in musical
artist similarity,” Proc. ISMIR, pp. 72-79, 2002.
[9] C. Inskip, A. Macfarlane and P. Rafferty: “Music,
movies and meaning: communication in film-
makers’ search for pre-existing music, and the
implications for music information retrieval,” Proc.
ISMIR, pp. 477-482, 2008.
[10] X. Hu and J. H. Lee: “A cross-cultural study of
music mood perception between American and
Chinese listeners,” Proc. ISMIR, pp. 535-540, 2012.
[11] T. Kärkkäinen, T. Vaittinen and K. Väänänen-
Vainio-Mattila: “I don't mind being logged, but want
to remain in control: a field study of mobile activity
and context logging,” Proc. CHI ’10, pp. 163-172,
2010.
[12] K. Kosta, Y. Song, G. Fazekas and M. B. Sandler:
“A study of cultural dependence of perceived mood
in Greek music,” Proc. ISMIR, pp. 317-322, 2013.
[13] P. Lamere. How good is Google’s Instant Mix?
http://musicmachinery.com/2011/05/14/how-good-
is-googles-instant-mix/
[14] A. Laplante and J. S. Downie: “The utilitarian and
hedonic outcomes of music information seeking in
everyday life,” Library and Information Science
Research, Vol. 33, No. 3, pp. 202-210, 2011.
[15] J. H. Lee and S. J. Cunningham: “Toward an
understanding of the history and impact of user
studies in music information retrieval,” Journal of
Intelligent Information Systems, Vol. 41, pp. 499-
521, 2013.
[16] J. H. Lee and R. Price: “User experience with
commercial music services: an empirical
exploration,” JASIST, DOI: 10.1002/asi.23433,
2015.
[17] J. H. Lee and N. M. Waterman: “Understanding user
requirements for music information services,” Proc.
ISMIR, pp. 253-258, 2012.
[18] T. W. Leong and P. C. Wright: “Revisiting social
practices surrounding music,” Proc. CHI ’13, pp.
951-960, 2013.
[19] M. Lesaffre, B. De Baets, H. De Meyer and J-P.
Martens: “How potential users of music search and
retrieval systems describe the semantic quality of
music,” JASIST, Vol. 59, No. 5, pp. 695-707, 2008.
[20] J. Lingel: “When organization becomes an extension
of your brain: DJs, music libraries and information
practices,” Proc. iConference, pp. 366-376, 2013.
[21] K. Liu and R. A. Reimer: “Social playlist: enabling
touch points and enriching ongoing relationships
through collaborative mobile music listening,” Proc.
MobileHCI 2008, pp. 403-406, 2008.
[22] T. Matthews, T. Judge and S. Whittaker: “How do
designers and user experience professionals actually
perceive and use personas?” Proc. CHI ’12, pp.
1219-1228, 2012.
[23] E. Nettamo, M. Nirhamo and J. Häkkilä: “A cross-
cultural study of mobile music: retrieval,
management and consumption,” Proc. OZCHI ’06,
pp. 87-94, 2006.
[24] L. Nielsen and K. S. Hansen: “Personas is
applicable: a study on the use of personas in
Denmark,” Proc. CHI ’14, pp. 1665-1674, 2014.
[25] M. Schedl: “Ameliorating music recommendation,”
Proc. MoMM’13, pp. 3-9, 2013.
[26] M. Voong and R. Beale: “Music organisation using
colour synaesthesia,” Proc. CHI EA ’07, pp. 1869-
1874, 2007
[27] J. Woelfer and J. H. Lee: “The role of music in the
lives of homeless young people: a preliminary
report,” Proc. ISMIR, pp. 367-372, 2012.
[28] I. Young: Mental models: aligning design strategy
with human behavior. Rosenfeld Media, 2007.
482 Proceedings of the 16th ISMIR Conference, M´alaga, Spain, October 26-30, 2015
