User Engagement
in the Digital World
Mounia Lalmas
Yahoo! Labs
Barcelona
mounia@acm.org

A bit about myself
 1999-2008: Lecturer (assistant professor) to Professor at
Queen Mary, University of London
 2008-2010 Microsoft Research/RAEng Research
Professor at the University of Glasgow
 2011- Visiting Principal Scientist at Yahoo! Labs
Barcelona
 Research topics
 XML/structured retrieval and evaluation (INEX)
 Quantum theory to model interactive information retrieval
 Aggregated search
 Bridging the digital divide
 Models and measures of user engagement

Outline
 Characteristics and measurement
 (my) Vision and focus
 Some results … and ideas
 What next?

Outline
 Characteristics and measurement
 (my) Vision and focus
 Some results … and ideas
 What next?

User Engagement – connecting three sides
 User engagement is a quality of the user experience
that emphasizes the positive aspects of interaction – in
particular the fact of being captivated by the technology.
 Successful technologies are not just used, they are
engaged with.
user feelings: happy, sad, user mental states: concentrated, user interactions: click, read
excited, bored, … lost, involved, … comment, recommend, buy, …
The emotional, cognitive and behavioural connection
that exists, at any point in time and over time, between a
user and a technological resource
S. Attfield, G. Kazai, M. Lalmas and B. Piwowarski. Towards a science of user engagement (Position Paper),
WSDM Workshop on User Modelling for Web Applications, 2011.

Would a user engage with this web site?
http://www.nhm.ac.uk/

Would a user engage with this web site?
(content)
http://www.amazingthings.org/ (art event calendar)

Would a user engage with this web site?
(aesthetics)
http://www.lowpriceskates.com/ (e-commerce – skating)

Would a user engage with this web site?
(navigation)
http://chiptune.com/ (music repository)

Would a user engage with this web site?
(navigation)
http://www.theosbrinkagency.com/ (photographer)

Characteristics of user engagement (I)
• Users must be focused to be engaged
Focused attention • Distortions in the subjective perception of time used to
measure it
• Emotions experienced by user are intrinsically motivating
Positive Affect • Initial affective hook can induce a desire for exploration, active
discovery or participation
• Sensory, visual appeal of interface stimulates, promote
Aesthetics focused attention
• Linked to design principles (e.g. symmetry, balance, saliency)
• People remember enjoyable, useful, engaging experiences
and want to repeat them
Endurability • Reflected in e.g. the propensity of users to recommend an
experience/a site/a product

Characteristics of user engagement (II)
• Novelty, surprise, unfamiliarity and unexpected
Novelty • Appeal to user curiosity, encourages inquisitive
behavior and promotes repeated engagement
• Richness captures the growth potential of an activity
Richness and control • Control captures the extent to which a person is able
to achieve this growth potential
• Trust is a necessary condition for user engagement
Reputation, trust and
• Implicit contract among people and entities which is
expectation more than technological
Motivation, interests, • Difficulties in setting up “laboratory” style experiments
incentives, and • Why should user engage?
benefits

Forrester Research – The four I’s
• Presence of a user
Involvement • Measured by e.g. number of visitors, time spent
• Action of a user
Interaction • Measured by e.g. CTR, online transaction, uploaded
photos or videos
• Affection or aversion of a user
Intimacy • Measured by e.g. satisfaction rating, sentiment
analysis in blogs, comments, surveys, questionnaires
• Likelihood a user advocates
Influence • Measured by e.g. forwarded content, invitation to join
Measuring Engagement, Forrester Research, June 2008.

Measuring user engagement
Measures
Characteristics
Self-reported Questionnaire, interview, report, Subjective,
engagement
product reaction cards
user study (lab/online)
Mostly
qualita,ve
Cognitive Task-based methods (time spent, Objective,
engagement
follow-on task) user study (lab/(online))
Neurological measures (e.g. EEG) Mostly
quan,ta,ve
Physiological measures (e.g. eye Scalability
an
issue?
tracking, mouse-tracking)
Interaction Web analytics + “data science” Objective,
engagement
data study
Information retrieval metrics + user
models
Quan,ta,ve
Large
scale

Objective measures – Online activities
Proxy of user engagement

Online measures as proxy of user
engagement!
measuring user engagement and interpreting metrics is hard!!

Online measures as proxy of user
engagement!
Multimedia search
activities often driven
by entertainment
needs, not by
information needs
M. Slaney, Precision-Recall Is Wrong for Multimedia, IEEE Multimedia Magazine, 2011.

Outline
 Characteristics and measurement
 (my) Vision and focus
 Some results … and ideas
 What next?

What affect user (& site) engagement?
Web page &
site style?
Web page &
site content?

User engagement… connecting three sides
+ layout + links The three sides
+ saliency + content
+ sentimentality, …
+ emotional
+ cognitive
+ behavioral
user engagement Measurements and methodologies
within and across site + online analytics metrics (dwell time, CTR, …)
+ complex networks metrics
+ new metrics + survival analysis
Goals
+ Models of user engagement + questionnaires, surveys, …
+ Metrics of user engagement + crowd-sourcing
+ biometrics (eye tracking, mouse tracking, …)

User Engagement – connecting three
measurement approaches
eme nt
tio n engag
interac
self-
repo
rted e
ngag
e me
nt

Diagnostic and what we have done
Diagnostic: work exists, but fragmented.
In particular:
o What and how to measure depend on services and goals
o Lack of understanding of how to relate subjective and
objective measures
The rest of this talk:
1. Models of user engagement
2. Attention & affect & saliency
3. Attention & affect & gaze & sentimentality
4. Attention & affect & mouse tracking (results pending)

Outline
 Characteristics and measurement
 (my) Vision and focus
 Some results … and ideas
 What next?

User Engagement – connecting three
measurement approaches
me nt
tio ne ngage
interac
self-r
epor
ted e
ngag
eme
nt

Models of user engagement
Online sites differ concerning their engagement!
Games Search
Users spend Users come
much time per frequently and
visit do not stay long
Social media Special
Users come Users come on
frequently and average once
stay long
Service News
Users visit site, Users come
when needed periodically
Is it possible to model these differences?

Data and Metrics
Interaction data, 2M users, July 2011, 80 US sites
Popularity #Users Number of distinct users
#Visits Number of visits
#Clicks Number of clicks
Activity ClickDepth Average number of page views per visit.
DwellTimeA Average time per visit
Loyalty ActiveDays Number of days a user visited the site
ReturnRate Number of times a user visited the site
DwellTimeL Average time a user spend on the site.

Methodology
General models Time-based models
Dimensions weekdays, weekend
8 metrics 8 metrics per time span
#Dimensions 8 16
Kernel k-means with
Kendall tau rank correlation kernel
Nb of clusters based on eigenvalue distribution of kernel matrix
Significant metric values with Kruskal-Wallis/Bonferonni
#Clusters
(Models) 6 5
Analysing cluster centroids = models

Models of user engagement
[6 general]
• Popularity, activity and loyalty are independent from each other
• Popularity and loyalty are influenced by external and internal factors
 e.g. frequency of publishing new information, events, personal
interests
• Activity depends on the structure of the site
interest-specific
periodic
media
e-commerce
media (daily)
search
models based on engagement metrics only

Time-based [5 models]
Models based on engagement over weekdays and weekend
work-related
daily news
hobbies,
interest-specific
weather
time-based models ≠ general models
next put all and more together! let machine learning tell you more!

Models of user engagement –
Recap & Next
 User engagement is complex and standard
metrics capture only a part of it
 User engagement depends on time (and users)
 First step towards a taxonomy of models of user
engagement … and associated metrics
 Next
 More sites, more models?
 Interaction between sites (online multi-tasking)
 User demographics, time of the day, geo-location, etc
J. Lehmann, M. Lalmas, E. Yom-Tov and G. Dupret. Models of User Engagement, UMAP 2012.

Online multi-tasking
July 2011, 25M sessions
avg session length 26mn (sd 44)
1.7 Yahoo! sites and 4.9 external
(sd 3.1 and 8.6)
leaving a site is
not a “bad thing!”
(fictitious navigation between sites within an online session)
users spend more and more of their online session multi-tasking, e.g. emailing,
reading news, searching for information  ONLINE MULTI-TASKING
navigating between sites, using browser tabs, bookmarks, etc
seamless integration of social networks platforms into many services

User Engagement – connecting three
measurement approaches
me nt
tio ne ngage
interac
self-
repo
rted e
ngag
eme
nt

Saliency, attention and positive affect
 How the visual catchiness (saliency) of
“relevant” information impacts user
engagement metrics such as focused
attention and emotion (affect)
 focused attention refers to the exclusion of
other things
 affect relates to the emotions experienced
during the interaction
 Saliency model of visual attention
developed by Itti and Koch
L. Itti and C. Koch. A saliency-based search mechanism for overt and covert shifts of visual
attention. Vision Research, 40, 2000.

Manipulating saliency
non-salient condition
salient condition
Web page screenshot
Saliency
maps

Study design
 8 tasks = finding latest news or headline on celebrity or
entertainment topic
 Affect measured pre- and post- task using the Positive
e.g. “determined”, “attentive” and Negative e.g. “hostile”, “afraid”
Affect Schedule (PANAS)
 Focused attention measured with 7-item focused
attention subscale e.g. “I was so involved in my news tasks that I
lost track of time”, “I blocked things out around me when I was
completing the news tasks” and perceived time
 Interest level in topics (pre-task) and questionnaire
(post-task) e.g. “I was interested in the content of the web pages”,
“I wanted to find out more about the topics that I encountered on the
web pages”
 189 (90+99) participants from Amazon Mechanical Turk

PANAS (10 positive items and 10 negative items)
 You feel this way right now, that is, at the
present moment
[1 = very slightly or not at all; 2 = a little; 3 = moderately;
4 = quite a bit; 5 = extremely]
[randomize items]
distressed, upset, guilty, scared, hostile,
irritable, ashamed, nervous, jittery, afraid
interested, excited, strong, enthusiastic,
proud, alert, inspired, determined, attentive,
active
D. Watson, L.A. Clark, A. Tellegen. Development and validation of brief measures of positive and negative affect:
The PANAS Scales. Journal of Personality and Social Psychology, 47, 1988.

7-item focused attention subscale
(part of the 31-item user engagement scale)
5-point scale (strong disagree to strong agree)
1. I lost myself in this news tasks experience
2. I was so involved in my news tasks that I lost track of
time
3. I blocked things out around me when I was completing
the news tasks
4. When I was performing these news tasks, I lost track of
the world around me
5. The time I spent performing these news tasks just
slipped away
6. I was absorbed in my news tasks
7. During the news tasks experience I let myself go
H.L. O'Brien. Defining and Measuring Engagement in User Experiences with Technology. PhD Thesis, 2008.

Saliency and positive affect
 When headlines are visually non-salient
 users are slow at finding them, report more
distraction due to web page features, and show a
drop in affect
 When headlines are visually catchy or salient
 user find them faster, report that it is easy to focus,
and maintain positive affect
 Saliency is helpful in task performance,
focusing/avoiding distraction and in
maintaining positive affect

Saliency and focused attention
 Adapted focused attention subscale from the online
shopping domain to entertainment news domain
 Users reported “easier to focus in the salient
condition” BUT no significant improvement in the
focused attention subscale or differences in
perceived time spent on tasks
 User interest in web page content is a good
predictor of focused attention, which in turn is a
good predictor of positive affect

Saliency and user engagement –
Recap & Next
 Interaction of saliency, focused attention, and
affect, together with user interest, is complex
 Next:
 include web page content as a quality of user
engagement in focused attention scale
 more “realistic” user (interactive) reading
experience
 bio-metrics (mouse-tracking, eye-tracking, facial
expression, etc)
L. McCay-Peet, M. Lalmas, V. Navalpakkam. On saliency, affect and focused attention, CHI 2012

User Engagement – connecting three
measurement approaches
me nt
tio ne ngage
interac
self-
repo
rted e
ngag
eme
nt

Gaze, sentimentality, interest … and
user engagement
 News + comments
 Sentiment, interest
 57 users (lab-based)
 Reading task (114)
 Questionnaire (qualitative data)
 Record mouse tracking, eye tracking, facial
expression, EEG signal (quantitative data)
Three metrics: gaze, focus attention and positive affect

Interesting content promote users
engagement metrics
 All three metrics:
 focus attention, positive affect & gaze
 What is the right trade-off?
 news is news 
 Can we predict?
 provider, editor, writter, category, genre, visual aids,
…, sentimentality, …
 Role of user-generated content (comments)
 As measure of engagement?
 To promote engagement?

Lots of sentiments but with
negative connotations!
 Positive effect (and interest, enjoyment and wanted
to know more) correlates
 Positively () with sentimentality (lots of emotions)
 Negatively () with positive polarity (happy news)
SentiStrenght (from -5 to 5 per word)
sentimentality: sum of absolute values (amount of sentiments)
polairity: sum of values (direction of the sentiments: positive vs negative)
M. Thelwall, K. Buckley, G. Paltoglou, Sentiment strength detection for the social web. JASIST, 63,1, 2012.

Effect of comments on user
engagement
 6 ranking of comments:
 most replied, most popular, newest
 sentimentality high, sentimentality low
 polarity plus, polarity minus
 Longer gaze on
 newest and most popular for interesting news
 most replied and high sentimentality for non-interesting
news
 Can we leverage this to prolonge user attention?

Gaze, sentimentality, interest –
Recap and Next
 Interesting and “attractive” content!
 Sentiment as a proxy of focus attention, positive
affect and gaze?
 Next
 Larger-scale study
 Other domains (beyond daily news!)
 Role of social signals (e.g. Facebook, Twitter)
 Lots more data: mouse tracking, EEG, facial
expression
I. Arapakis, M. Lalmas, B. Cambazoglu, M.-C. Marcos, J. Jose. Examining User Engagement through the Prism of
Interest, Sentiment and Gaze, Submitted for Publication, 2012.

User Engagement – connecting three
measurement approaches
me nt
tio ne ngage
interac
self-
repo
rted e
ngag
eme
nt

Mouse tracking … and user
engagement
 400 users from Amazon Mechanical Turk
 Two domains (BBC and Wikipedia)
 Two tasks (reading and quiz)
 “Normal vs Horrible” interface
 Questionnaires (qualitative data)
 Mouse tracking (quantitative data)
 Interaction data (page view, dwell time)
 Results pending! … Hawthorne Effect!!!!!!!!!!

Mouse tracking … and user engagement
(Taxonomy? Correlation vs Causation? Measurement? … )

Outline
 Characteristics and measurement
 (my) Vision and focus
 Some results … and ideas
 What next?

tion
interac ent
em
engag
self-
re
enga ported
gem
ent

Outline
digital world
digital libraries
user engagement
mobile & tablet

Thank you
Collaborators
Ioannis Arapakis
Ricardo Baeza-Yates
Georges Dupret
Janette Lehmann
Lori McCay-Peet
Vidhya Navalpakkam
David Warnock
Elad Yom-Tov
and many others at Yahoo! Labs