1. Collective Emotions in Cyberspace
Short review of Cyberemotions project results
In the name of CYBEREMOTIONS Consortium
Janusz Hołyst, Project Coordinator, Warsaw University of Technology,
jholyst@if.pw.edu.pl
www.cyberemotions.eu
2. Plan
• Emotions, cyberemotions, collective emotions
and collective cyberemotions
• Cyberemotions Project structure
• Main results of various Project layers:
- data collection and classification
- collective character of cyberemotions
and data driven models of cybercommunities
- project ICT outputs
3. There is no agreement what emotions
are but they are important for our life !
8. 8
Twitter
Revolution
Pic. Arab Spring in Egypt 2011
congratulations Egypt the
criminal has left the palace
– a tweet from Egyptian
protest leader Wael Ghonim.
Twitter can help organize. Facebook can
help get the word out. YouTube provides
evidence. Over the past few years, we've
seen that social media can be a powerful
tool in assisting revolutions in countries.
- Cheryl Aguilar, EthnoBlog
Collective emotions in cyber-communities ?
Pic. STOP SHREDDING OUR CONSTITUTION, USA 2012
SOCIAL MEDIA &
US 2012 ELECTIONS
9. Collective Emotions in Cyberspace
European Union Research Project (FP7 FET)
Participant organisation
name
Leaders Country Specialization
Warsaw University of Technology Janusz Hołyst Poland Physics of complex systems
EPF Lausanne Ronan Boulic Switzerland Virtual reality
University of Wolverhampton Michael Thelwall United Kingdom Webometrics
Österreichische Studiengesellschaft
für Kybernetik
Robert Trappl
Marcin Skowron
Austria Human-computer interactions
ETH Zürich Frank Schweitzer
David Garcia
Switzerland Chair of systems design
Jozef Stefan Institute, Ljubljana Bosiljka Tadic Slovenia Physics of complex networks
Jacobs University, Bremen Arvid Kappas Germany Psychophysiology
Technical University Berlin Matthias Trier Germany Dynamic network analysis
Gemius SA Anna Winnicka Poland Online research agency
Large-scale integrating project, ICT Call 3 Science of Complex Systems for Socially
Intelligent ICT. Duration: 1 Feb. 2009 - 31. Jan. 2013. EC funding 3.6 M€
10. Expected impact of CYBEREMOTIONS
• new classes of realistic models of emotionally reacting E-users
• new kind of intelligent self-adapting programs, cyber-tutors, cyber-advisors for e-
communities (long time scale)
• to create theoretical background for the development of the next generation emotionally-
intelligent ICT services using universal methods of complex systems (long time scale) .
CYBEREMOTIONS = data gathering + complex systems methods + ICT outputs
Main aims of Cyberemotions Project were
to understand the process of collective emotions
formation in e-communities
13. WP3 created sentiment analysis softwareWP3 created sentiment analysis software
- Used for research and for light displays- Used for research and for light displays
on the London Eye during the Olympicson the London Eye during the Olympics
SentiStrengthSentiStrength
15. Data collected by Wolverhampton group
BBC Forum BBC “Religion and Ethics” and “World / UK News” message boards starting
from the launch of the website (July 2005 and June 2005 respectively) until the beginning of the
crawl (June 2009).
#comments 2,474,781 #users 18,045 # threads97,946
Digg The analysis spans the months February to April 2009 and consists of all the stories,
comments and users that contributed to the site during this period. The resulting dataset
contains approximately
1.9 million stories, 1.6 million comments and 800 thousand users.
Blog06
crawl of approximately 100,000 blogs and which spans 11 weeks, from 06/12/2005 to
21/02/2006", i.e. the dataset contains webpages from 100,000 different blogs (more than 3
million webpages) . The blogs are from all over the world, although there is an emphasis on
English content
#comments 242,057 #discussions 1219
4 million comments
Detection of collective emotions in cyber-communities
16. Emotions (emotional valence e ={ +1,0,-1})
We define an emotional cluster of size n as a chain of n consecutive messages with
similar sentiment orientations (i.e. negative, positive or neutral).
Emotional clusters
Detection of collective emotions in cyber-communities
17. Emotional homophily of e-communities
α
neepneep )|()|( ≈
The presence of a longer
cluster of coherent
emotional expressions
increases a possibility to
follow the cluster by a
comment with the same
emotion.
Conditional probability for cluster growth increases as a power-law with cluster
length.
18. Collective emotions of cybercommunities detected by various
methods
t
Sentiment Triad Census Analysis
Emotional persitence of IRC chatts
Emotional clustersEmotional avalanches
Hurst eponents
19. 19
p(e|e)
Characteristic exponents α
decay linearly with
conditional probability of
emergence of clusters of size
two
2. Collective emotions in cyber-communities
)|(8.06.0 eep−=α
Minority emotion (less
frequent) posses larger
value of α - the growth
probability is more
dependent on cluster size
Week
interaction
Strong
interaction
Rare emotions create
stronger ties
20. 20
Negative emotions as a fuel for discussion in
cyber communities
Negative emotionBetter not to be here …
A negative emotion results with escape response in real
world
What about the Internet ?
21. 21
|<e>| absolute value of the average emotion
valence of the first 10 comments
<x>
Lenght of a thread 20 40 60 80
Number of comments in a thread
<e>
Average length of a thread as a function of
the absolute value of the average emotion
valence of the first 10 comments
Emotional beginnings of the threads, whether positive or negative,
usually lead to longer discussions
Negative emotion as a fuel for discussions
22.
23. WP6/JSI:Emotional Bots can induce
collective mood
[Ref3]: B. Tadic and M. Suvakov , Arxiv:1305.2741 (2013)
joyBot polarizes network of Agents (red links indicate
positive emotion messages), while miseryBot induces
excess of negative emotion messages (carried by black
links ) [Ref3]
Simulations revealed how
Agents collective emotion
polarizes under the influence
of positive/negative emotion
Bots [Fig.]
Bot's impact on Agents can
be measured; It relies on the
network structure (which
propagates emotion among
Agents) and on the
self-organized nature of the
dynamics (which enhances
correlations)
24. WP6/JSI: Agent-Based Model of Chats
with Emotional Bots
Ref.: V. Gligorijevic, M. Suvakov and B. Tadic, DRAFT (2013)
Experimental data: Users group according to their
similarity in emotional communications with Bot (5
communities, left);
More cohesive groups appear when they are placed in
an interactive environment (simulated, right) [Ref.]
Agent-Based Model with
emotional Agents +
Moderators + Bots developed
& validated
Agents designed with certain
'human' attributes (inferred from the
empirical data )
Experimental emotional Bot
used as input: response of
Agents simulated
25. AustrianResearchInstituteforArtificialIntelligence–OFAI
OFAI , Wien, Interactive Affective Bots
Environment
[user] [web]
Perception
Natural Language
Understanding
Affective Cues
Sentiment class
ANEW: valence,
arousal, dominance
LIWC: affective, ling.
cognitive categories
Action categories,
user_ID, channel_ID
timestamps
Control
Interaction Manager
Dialog Scripting
AIA Report. Module
Simulations
Collective Users
Modelling
Individual User
Modelling
Actuator-Communication Layer
WWWWWW
• Tools for:
• studying affective human-
computer interactions:
- single user, -multiple users
• acquisition of data on users'
sentiment towards entities,
events, processes
• experimental evaluation of
theoretical models
•
• Example realizations of IAB:
• Affect Listener Dialog Participant
• Affective Interaction Analyser
• Affective Supporter and Content
Contributor
26. OFAI, Vienna,Affect Listener
- Development of Affective Dialog Systems
• Evaluation of systems in 5
experimental setups
• Dialog system vs. Wizard of OZ
• dialog realism, chatting
enjoyment, emotional connection
• Effect of system’s affective profile
• positive, negative, neutral
• Effect of interaction context and
roles assigned to the user and
system
• Effect of fine grained communication scenarios
• social sharing of emotions, getting acquainted with someone
• Attention and social interactions context - social exclusion
27. Jacobs University: Social Exclusion by the
Conversational System?
Mean subjective evaluation of attention
paid by the bartender. *** significant
difference at p < .0001 .
***
28. EPFL: emotions in virtual reality
Crowd Visualization Software
1. Two H/W platforms: Desktop and CAVE
2. Two S/W platforms: YaQ and Unity3D
3. Pilot S/W OVS v1.1 accessible online Evaluation and
Validation
WP2
D2.4 Summary
Crowd Visualization with Emotion
1.S/W platform: YaQ
2.Number of virtual humans: 200
3.Display rate: more than 60 fps
31. Selected project achievements
• Emotional responses can be predicted from observation of sentiment
fluctuations in physiological and Twitter data.
• Asymmetry is crucial for emotion animations in facial expressions.
• Sentistrength program used during Olympic Games to monitor daily
moods in UK (display at London Eye).
• Developed affective bots are capable to communicate with humans.
• Emotions can be crucial for leaving of Open Source Community by their
active members (including project leaders).
• Chat Bots developed for data driven models of e-communities can
propagate negative/positive emotions and polarize channel moods.
• Chat Bots can lead to social exclusion of e-community members.
• Universal tools developed for automatic analysis and visualisation of
emotion propagation in social data.
32. Conclusions
.
• We demonstrated that collective emotions do exist
in a broad class of e-communities
• Collective emotional dynamics is vital for the
efficiency and survival of e-communities
• The current technology makes possible to create
bots that can influence human emotions
• Understanding the role of and strategic use of
cyberemotions will be crucial for the future society
because of technological, economical and political
issues.
33. More results will be presented
at next presentations, posters
and www.cyberemotions.eu
CyberEmotions video lectures:
https://www.youtube.com/user/fen
sPW/videos
Notas del editor
1. The dialog system was applied for the acquisition of supplementary data, that help to extend the scope of analysis in: - quantitative way – reaching to the users who do not voice their opinions in the online debates - qualitative way – allowing to conduct a follow-up dialogs, related to the introduced set of topics as well as user’s expressions of affective states 2. Two experimental settings were used for the dialog system evaluation: - Virtual Reality in which the dialog system managed the verbal aspects of communication between a Virtual Human (Virtual Bartender) and users Online web chat environment, typical for several internet communication interfaces 3. In the first interaction settings the system was compared with Wizard-of-Oz settings in terms of: establishing an emotional connection dialog realism providing an enjoyable chatting experience 4. 2 nd round of experiments focused on: - acquisition of data on user’s responses to the introduced topics of interests (announced tax increase, smoking prohibition in the public places, allegations of a bribery related with the organization of UEFA soccer championships) - assessing the effect of the system affective profile in the interaction with users -