1 Managing Attention in the Social Web: The AtGentNet Approach
Thierry Nabeth, INSEAD – Fontainebleau, France
Nicolas Maisonneuve, INSEAD – Fontainebleau, France

Abstract
By transforming the Web into a massive social space, the Web 2.0 has open a vast set of opportunities
for people to interact with one another using online social networking, blogs, wikis or social
bookmarking. But at the same time such phenomenon has created the conditions of a massive social
interaction overload: people are being overwhelmed by solicitations and opportunities to engage into a
social exchange but they have little means about how to deal effectively with this new level of
interaction. The objective of this chapter is to investigate the use of ICT (Information and
Communication Technologies) to support people in managing their attention at a social level. More
specifically, it consists in the adaptation to a social context, of a general model for supporting attention
that was proposed by Roda and Nabeth (2008) and that relies on supporting attention at four levels:
perception, deliberation, operation and meta-cognition. This chapter also presents how the support of
social attention has been operationalised with the design of an attention aware social platform
AtGentNet, and tested for supporting interactions of communities of learners and professionals. After
discussing the results of the experimentation, this chapter concludes by reflecting on how this
approach can be generalized to support the interaction of people in the social web in general.

Keywords: social attention, social interaction overload, social attention management, social
cognition augmentation, web 2.0.
1.1 INTRODUCTION: ADDRESSING THE SOCIAL INTERACTION OVERLOAD
The Social Web, which refers to people use of Web 2.0 technologies to facilitate social activities (Chi,
2008), has totally reinvented the web as a massive participatory social space. In the Web 2.0
(O’Reilly, 2005), people are engaged in a variety of interaction with others: They use blogs, wikis
(Cunningham and Leuf, 2001) and other social media to participate in the creation of content; they
employ social bookmarking (Golder and Huberman, 2006; Halpin, Robu, and Shepherd, 2006;
Marlow, Naaman, Boyd, Davis, 2006), reputation systems (Resnick, Zeckhauser, Friedman, and
Kuwabara, 2000), RSS feeds (Gill, 2005) and other massively distributed collaborative mechanisms to
filter, share, aggregate and annotate resources; they maintain constant contact with others using instant
messaging or micro-blogging systems; and finally, people use social networking (Boyd and Ellison,
2007) or dating services to help them in managing relationships.

The advantages of a more social Web are undeniable: people’s roles have been transformed from
passive consumers of resources into active participants (much content is generated by users), and this
new setting has considerably augmented the availability, the processing and the circulation of
knowledge (since everyone is potentially an active participant in these knowledge processes). Besides,
these processes exist without supervision, relying on concepts such as emergence or collective
intelligence (Tapscott and Williams, 2007; O’Reilly, 2005; Weiss, 2005; McAfee, 2006) to make it
happen without apparent effort.
Yet at the same time, this new setting has come at a cost: Whereas in the Web 1.0 people were
overwhelmed with information, the Web 2.0 has come with a tremendous level of interaction, and
people are now subject to massive social interaction overload. Participating in social media such as
blogs and Wikis, as a consumer and even more as a contributor represents a time consuming activity
(Perez, 2008). People are also subject to a high level of social solicitations originating from multiple
sources and available in multiple forms (invitation to connect in online social networking services,
invitation to chat, invitation to become part of a group, invitation to comment, etc.) and providing a
fragmented (quasi schizophrenic) perception of the environment. They are also kept in a state of
“continuous partial attention” (Stone, 2006) reinforced by the fear of being disconnecting from the
social sphere and becoming marginalized. Finally people are under constant pressure to develop their
social relationships (Granovetter, 1973) since the associated social capital (Burt, 1997) is
acknowledged as a determining factor of success in the knowledge economy (Nardi, Whittaker &
Schwarz, 2000; Thomas, Kellogg, and Erickson, 2001).
Managing all these social interactions and solicitations in the social web represents a daunting and
seemingly impossible task: how to sort out the different solicitations and limit the level of interaction
knowing that some one them will be critical? Which level of time and attention to dedicate to these
social interactions without risking overlooking something important or having one’s attention totally
consumed in unproductive activities? In particular when we know that in the Web 2.0, people
cognitive capabilities have not changed: the short term or working memory of the human brain is still
limited to a maximum capacity to manipulate 7 +- 2 concepts (Miller, 1956) and, more importantly
related to social interaction, the maximum number of stable interpersonal relationships that a person
can effectively manage is 150 (Dunbar, 1992).
The objective of this paper is to provide some answers to these questions in the context of a social
platform AtGentNet that was elaborated to support the interactions of groups of learners engaged in a
blended learning program.
The first part of this paper briefly describes principles for supporting attention, based on a model of
attention support at four levels: perception, deliberation, operation and meta-cognition. We then
describe how these principles can be adapted to support attention in a social context. In the next
section we present how these principles have been operationalised in the social platform AtGentNet,
via the design of an architecture and of a series of mechanisms supporting social attention at the

different levels. We then report on how this platform was tested for supporting the online interaction
of a group of learners engaged in a blended learning program. Finally, we conclude by reflecting how
this approach can be generalized to support the interaction of people in the social web in general, and
we identify lines of further evolutions.
1.2 MANAGING ATTENTION
1.2.1 Attention in the Information & Knowledge Economy
The questions of social interaction overload can barely be considered as totally new and only
happening with the creation of the so called social web, since similar questions had already surfaced as
a major concern related to the advent of information and knowledge economy (Drucker, 1999).
The old industrial economy was characterized by the scarcity of information, and by an important cost
of accessing it. At this time, people and organizations had no difficulties managing information and
the critical factor for success of companies was the capacity to access capital. The advent of the
knowledge economy at the end of the XXe century totally transformed this situation: In the
information & knowledge economy, information is abundant and even overflowing (Goldhaber, 1997).
Knowledge is also the subject continuous regeneration and transformation (Senge, 1990), and is also
socially constructed via social interaction (Argyris and Schon, 1978; Kogut, 2008). In this context, the
new factor of success has become the capability to process effectively huge amount of information and
knowledge, with limited resource constituted mainly by people time, i.e. the attention they can
dedicate to accomplish their tasks. In the information economy, attention has therefore become one of
the most important focus of people and organizations (Ocasio, 1997; Simon, 1971), the most
successful organizations being the ones the most able at allocating their attention effectively and
therefore the more capable at dealing with information overflow (Davenport and Beck, 2001) and to
adapt to a changing environment (Ocasio, 1993). The importance of attention for knowledge intensive
activities has thus been acknowledged in a variety of sectors such as venture capitalism (Gifford,
1997), libraries (Bridges, 2008) or advertising (Huberman and Wu, 2007).
The advent of the social web has further transformed the situation by putting social interaction in the
centre. In the social web, opportunities to interact with others have flourished everywhere: from the
blogs in which people can engage in interactions with others; in wikis in which they are asked to
contribute to the creation of collective knowledge; in microblogging applications (such as Twitter) that
are used to express and to follow people’s current thinking; in online social networking services where
people develop and maintain their relationships. In the social web, the abundance of information has
been complemented by the abundance of social interaction, attention remaining however the scarcest
resource. In this context, the more successful organizations will be the ones in which people are able
to deal the most effectively with a high and rich level of social interactions and to optimize the use of
attention for supporting social interaction.

1.2.2 Supporting attention: a four level model
Different models and mechanisms have been proposed to support attention, first at an organizational
level by finding approaches and methods for helping companies and people to manage their attention
in a more effective way and second at the technical level by using ICT (Information and
Communication Technologies) to support them more operationally in managing their attention.
At the organizational level, Davenport and Beck (2001) have proposed an approach and an assessment
tool (AttentionScape) helping individuals and organizations to determine and to optimize how they
allocate their attention along tree axes (Beck and Davenport, 2001): (1) Aversive / attractive: Aversive
attention is paid when people are afraid of the consequences of not paying attention. Attractive
attention is given to elements people like and expect to be pleasant.; (2) Captive / voluntary: People
pay voluntary attention to things they find innately interesting, but attention is held captive when
people have something thrust upon them; (3) Front of mind / back of mind: Front of mind is related to
an active and conscious allocation of cognitive resources (like reading a textbook), whereas back of
mind is related to a partial, unconscious allocation of these resources (like listening music in the
background whiles reading a book). This notion of peripheral attention is also present in the concept
of “continuous partial attention” coined by Linda Stone (2006) to describe the dominant mode of
attention nowadays where people are under a continuous state of vigilance given the use of the new
communication tools (SMS, Chat, messengers). Linda Stone also makes a distinction between two
modes of management of people‘s time: The tactical mode where “all is about tasks prioritization” and
“optimization efficiency”; and the strategic mode where “all is about intention, making choices as to
what does and does not get done”. She stresses that the tactical mode is mostly in use today, and calls
for a development of the strategic mode. Finally, Ashkenas (2007) proposes to address the attention
overload by working on the reduction of complexity in organization by: (1) Making simplification a
goal; (2) Simplify the organizational structure; (3) Prune and simplify products and services; (4)
Discipline business and governance processes; (5) Simplify personal patterns.
At the technical level, different approaches have been proposed to use ICT to support attention. Bier
and colleagues (Bier, Stone, Pier, Buxton, & DeRose, 1993, p.73) have proposed the Magic Lens
concept as “filters that modify the presentation of application objects to reveal hidden information, to
enhance data of interest, or to suppress distracting information”. More generally ‘attentive user
interfaces’ have been proposed in order to increase the effectiveness of human computer interaction
(Vertegaal, 2003). More specifically four main types of attentive user interfaces have been identified
(Vertegaal, et al., 2006): Visual attention, turn management, interruption decision interfaces; and
visual detail management interfaces. Huberman and Wu (2008) have introduced automatic
configuration mechanism generating the most relevant information being presented to limited attention
users. Finally, Anicic, Stojanovic, and Apostolou (2008) propose the use of Enterprise Attention
Management Systems that consist in attention aware platforms such as Workflow, Content

Management System (CMS) and the GroupWare System (GWS). Their attention model relies on the
idea of proactively supporting the user in reacting on changes respecting the user’s context and
preferences. Practically, their system is an event-based system tracking and mining events (such as a
new document has been added or someone opened a document), and managing alerts in an attention
effective way.
A brief analysis of these different approaches and tools indicates that attention can be supported in a
variety of ways such as by filtering the noise, by making the information more relevant (via
personalization), by minimizing distraction by better managing notification, by reducing the required
level of vigilance (e.g. through the use of personal organizers), or helping people to assess the level
and the nature of attention that people dedicate in their activities (e.g. what is done with
AttentionScape).
Roda and Nabeth (2008) have proposed a holistic framework for integrating in a single model the
different means of supporting attention. This model relies on the idea of supporting attention at four
different levels: (1) the perception level; (2) the deliberative level; (3) the operational level; (4) the
meta-cognitive level. The support of attention at the perceptual level consists in enhancing people’s
perceptive capabilities. It relies on the idea of filtering the irrelevant or less important information, of
emphasizing the most important one, and of presenting interruptions at the appropriate level of
prominence. The support of attention at the deliberative level consists in helping people in their
decision making. The support of attention at the operational level consists in reducing the effort
needed to accomplish tasks. For instance some mechanisms may automate a task or reduce the number
of steps required to accomplish a task. Finally the meta-cognitive support consists in helping people to
improve their attention allocation practices. Thus mechanisms can be used to allow people to assess
their current practices, such as visualizing how they allocate their attention and to situate their
practices. Other mechanisms can allow people to experiment with new practices. Finally, other
mechanisms can be used to stimulate motivation, for instance by increasing the perception of self-
efficacy (Bandura, 1994), or by providing the means to compare their practices with others.
The next section will show how this four levels model can be adapted to the support of attention in a
social context.
1.3 MANAGING SOCIAL ATTENTION
In this section, we are going to present how the four levels model of supporting attention of Roda
and Nabeth (2008) can be adapted to support the management of attention in a social context. More
specifically, for each level (perception, deliberation, operation and meta-cognition), we will introduce
a set of mechanisms that can be used to support people in better allocating their attention when
interacting with others. However, before proceeding further, we would like first to make a rapid
overview of theories and works that may be relevant to the support of attention in a social context and
that will inform the work presented here.

1.3.1 Relevant theories and research
Psychology and Sociology have proposed a number of theories that can be relevant to the management
of attention in a social context, and in particular that associate limitations or a cognitive effort to the
establishment and the maintenance of social relationships. First, the anthropologist Robin Dunbar
(1992) has found a maximum number of 150 as the number of stable interpersonal relationships that
humans can effectively manage due to some limitation in the brain, indicating that there exists some
constraints on the number of relationships that can be managed by a single person that will be difficult
to overcome. A consequence is that it is illusory to expect to support people to manage a very large
number of relationships at a given time, unless some of these relationships are very loose, and actually
not really used. Thus the value of the hundreds of relationships that some persons record in online
networking services, or the important number of people that they may ‘follow’ in micro-blogging
systems such as Twitter, is most probably be very limited. Second, there exists a number of theories
that have tried to understand how people manage their relationships and that induce some limits in
their number such as the social exchange theory (Thibaut and Kelley, 1959) and the social behavior as
exchange theory (Homans, 1958). Both these theories represent models derived from the application
of the economic theories of rational choice that associate a cost to establishing and maintaining a
social relationship. More specifically, the social exchange theory proposes that voluntary relationships
depend on receiving satisfactory outcomes, and that a person’s commitment to an existing relationship
is proportional to his/her satisfaction in this relationship and to the investment he/she has already put
in this relationship and it is inversely proportional to potential alternative relationships. A relationship
therefore generates value but it also has a cost, and there exists a limit to which the cost of a relation
overcomes the benefit. The social behavior as exchange theory is even more radical about this, since it
explains how people interact socially as a process of negotiated exchanges between parties. We can
presume, since negotiating has a cost, that people will find some limit on the number of people with
whom they can establish a fruitful relationship.
Another stride of research is related to the augmentation of social cognition. Social cognition
augmentation (Chi et al., 2008) consists in “enhancing the ability of a group to remember, think, and
reason”, increase usability and facilitate the navigation in the information via social imitation and
mimesis (Erickson, 2009) with the concept of social navigation (Dieberger, Dourish,.Höök, Resnick,
and Wexelblat, 2000) and social foreaging (Chi, Pirolli, & Lam, 2007; Giraldeau and Caraco, 2000).
These mechanisms also influence the level of the motivation of individuals and of groups via social
stimulation happening via social comparison (Harper, Li, Chen, Konstan, 2007), via reinforcing the
perception of self-efficacy in a social context (Banduras, 2001) by displaying the value of contribution
(Rashid et al., 2006), or by allowing the expose of personal information (Tufekci, 2008) and flattering
people ego (Joinson, 2008; Nishikant, Konstan & Terveen, 2005). More generally these mechanisms
intervene for supporting collective intelligence (Yuan, Chen, Wang, Du, 2007).

Finally, on a more operational perspective, we can observe the integration in many social media
platforms (such as online social networking, micro-blogging, Wikis) of a number of mechanisms
increasing the effectiveness of managing a large number of interactions. For instance, the social
networking system LinkedIn or the micro-blogging system Twitter aggregate in a single linefeed all
the activities originating from the acquaintances of a user, reducing the cognitive effort to keep track
of the crowd of the acquaintances of this user. Wikipedia offers the possibility via the watch list, to
track in a single place the changes happening in a set of pages that the user puts in observation,
reducing considerably the effort required to follow the resources that are the more relevant to this user.
Finally, we can mention personal information management (PIM) systems such as electronic address
book, or more simply the SMS in mobile phone, that definitively allow a user to be more attention
effective related to the management of social relationships. In all these cases however, the support of
social attention has not been the driving force for the design of these tools, but only the desire to
increase personal efficiency.
1.3.2 An application of the four level model
1.3.2.1 Enhancing social perception
The enhancement of the perception of the social activity has been the object of numerous researches in
particular with the work of Kellogg, and Erickson (2001) and the concept of social translucence.
Social translucence consists in making participants and their activities visible to one another. The role
of social translucence is to inform, to create awareness and to enforce accountability (Erickson &
Kellogg, 2003). By enhancing the social perception, it also contributes to the coordination of groups as
well as stimulating participation (Vassileva and Sun, 2007). A variety of mechanisms can be used, and
have been invented as part of the web 2.0, to make the social activity more visible. Erickson (2009)
refers to these mechanisms as social proxies, i.e. “minimalist graphical representation that portrays
socially salient aspects of an online interaction”. Examples of such mechanisms include: presence
features in instant messaging systems, notification in email systems, list of contributors in online
communities and the level of their contributions and items that are the most popular in online
communities, social connectedness and life activity feeds in online social networking services, tagging
in collaborative bookmarking services (Marlow, Naaman, Boyd, Davis, 2006).
1.3.2.2 Supporting deliberation
The support for the deliberation process consists in assisting users in choosing the most effective
approach to adopt for interacting with others. Guidance can be offered by mechanisms based on high
level visualization helping the decision process. It can also suggest approaches maximizing the impact
of the actions such as the application of principles derived from the diffusion of innovation theory
(Rogers, 2003) or the use of viral marketing techniques (Subramani & Rajagopalan, 2003).

Mechanisms may include: (1) methods and tools to analyze social networks (Cross, Borgatti and
Parker, 2002) so as to identify the most important nodes; (2) reputation indicators (Resnick,
Zeckhauser, Friedman, and Kuwabara, 2000) allowing to assess the risks attached to a potential
interaction and decide the value of engaging into a social exchange (Thibaut and Kelley, 1959); or (3)
indicators allowing to select information and actions having maximum impact but requiring a
minimum of effort. In the later cases, users may be guided by elements such as novelty or popularity
of the items (Wu and Huberman, 2008).
1.3.2.3 Providing operational support
The support of operation is aimed at helping to reduce the cognitive effort required to accomplish a
task related to a social interaction by reducing the number of steps required or by automating a task.
An example of such tasks is the affiliation to a group that can be suggested to a user after having
observed some level of interest of this user towards this group. By making more visible people
interests or by the use of introductions and recommendation mechanisms the cognitive effort related to
the weaving of social ties is reduced. Other approaches may consists in selecting a tool reducing the
amount of effort to communicate, such as it is the case with the use of social media like blogs, Wikis
or RSS feeds. The usage of these tools will result in a reduction of the effort from the
“communicator”, and less distraction for target users. Other examples includes organizer services
allowing people to be assisted in managing their agenda, reducing their ‘back of mind’ cognitive load
from the necessity to remember appointments, or birthday date of friends. Finally in a similar line,
watch mechanisms such as the ones that can be found in Wikipedia, will allow a user to keep track of
the changes of a set of resources without dedicating too much of his attention.
1.3.2.4 Meta-cognition
The support to meta-cognition is related to all the mechanisms that provide people the possibility to
get aware of their own behaviours when interacting with others, so as to learn new interaction
practices that are more effective. It also includes elements intervening in people motivation to interact
or to participate. The first category of mechanism includes tools allowing them to assess how they
have dedicated their time in social related activities, such as AttentionScape system previously
mentioned, but also statistical tools visualizing their activities. Mechanisms indicating the popularity
of their contributions, and more generally their level of impact may also be used as a means to assess
the effectiveness of their current practice. It will also intervene at the level of motivation of the users
since it contributes to increase the perception of self-efficacy (Bandura, 2001). This effect has for
instance been demonstrated by Huberman, Romero and Wu (2008) in the case of user generated
content on YouTube: “the productivity exhibited in crowdsourcing exhibits a strong positive
dependence on attention, measured by the number of downloads. Conversely, a lack of attention leads
to a decrease in the number of videos uploaded and the consequent drop in productivity”.

1.3.2.5 Discussion and summary
A mechanism may provide attentional support at different levels at the same time. Thus a
mechanisms contributing to make the social activity more visible will intervene both in: enhancing the
user perception, helping user’s decision making; and providing him feedback allowing him improving
his practice. This makes the applicability to a social context of our four level models of attention
support less straightforward than if we had been able to assign to each mechanism to a single level.
However, the application of this model remains useful since it allows ensuring that attention support is
done at all levels, and not only at the levels of perception as it is usually the case in systems focussing
essentially on user interface.
The following table provides a summary of the support of attention at different levels:
Level Support of social attention Examples of mechanisms
Perception Make the social activity more visible
(Social translucence). Filter the
information and personalize the
interaction.
Presence mechanisms; Notification
mechanisms; Popularity indicators (such as
most accessed document)
Deliberation Inform and guide users in deciding about
actions that are the most attention
effective (require less effort and have the
most impact).
Reputation indicators; involvement
indicators (e.g. most active contributors);
social network visualization and analysis;
impact indicators (e.g. ‘who reads me’);
tactical advices.
Operation Reduce the cognitive effort required to
accomplish a task or conduct an activity,
and for instance of the need to
remember.
Automated group affiliation; watch lists;
adoption of more attention effective
communication tools (such as blogs, rss);
organizer services.
Meta-
cognition
Help the users to assess the attention
effort in conducting some activities so as
to help them to learn to become more
attention effective.
Display of statistics of attention allocation;
comparison with other people practices;
indicators of the level of impact; strategic
advices;

Table 1: Supporting attention at different levels

1.4 ATGENTNET: A SOCIAL ATTENTION EFFECTIVE PLATFORM
1.4.1 Overview
AtGentNet is a Web 2.0 social platform elaborated in the context of the EU research project
AtGentive aiming at supporting users’ social attention in community platforms (Roda and Nabeth,
2006). This platform is used to support both collaboration and social networking in different contexts
such as blended learning (Nabeth, Karlsson, Angehrn and Maisonneuve, 2008), knowledge exchange
in communities, and collaboration in groups. AtGentNet incorporates a mix of features supporting
users in interacting and “networking” with others in a way that is social attention effective i.e. helping
to reduce the level of attention that a user must dedicate in this interaction.
The platform appears as a community portal (see Figure 1), and provides the communication and
social network infrastructure supporting the interaction of members and well as their networking.


Figure 1: a snapshot of the platform1

It includes forums, bulletin boards, chat space, search, tagging, profiles, groups, membership
management. This platform also includes a series of mechanisms providing basic attention support,
and in particular translucence mechanisms, such as the display of the statistics of activities of the
participants, the list of the most popular resources, the list of the most active contributors, the list of
the more regular visitors, etc. This platform is enhanced by an intelligent agent generating different

1
Other snapshots of the platform are available at :
http://www.calt.insead.edu/LivingLab/AtGentive/Wiki/?AtGentNet

interventions that are proposed in one of the portlets or delivered via an artificial character embodying
the artificial agent. More details about these mechanisms will be provided later in this chapter.
1.4.2 The Design
AtgentNet is composed of two main elements: (a) a collaborative / social platform providing a
variety of means (including an artificial character) supporting people interaction and networking; (b)
an external agent module able to provide proactive and intelligent support to attention. This agent
module consists in a perception layer and a reasoning layer. The function of the perception layer is to
observe and to mines the activity of the community originating from the platform. The reasoning layer
use this (digested) information to generate a series of advanced proactive interventions.

users
P
e
r
c
e
p
t
i
o
n
L
a
y
e
r
1
2
User A
User B
Reasonning Layer
Platform
Data
Exporter
State
AtGentNet
network &
collaboration
platform
CHI
API
Data capture &
fusion
3
State
4
agent

Figure 2: the AtGentNet overall architecture

This architecture allows a very good decoupling between a community platform and the advanced
mechanism. The platform is indeed only aware of the agent module via the calls done to its API. This
agent-based approach, also adopted in (Nabeth, Roda, Angehrn, Mittal, 2005) to stimulate
participation in virtual communities, moves totally the intelligence to an external module increasing
the modularity and thus the reusability since both the platform and the agent module can be used
independently in a totally different context. Yet, the two systems appear seamlessly integrated into an
agent-enhanced social platform as presented in Figure 1.

1.4.2.1 The platform
The platform was implemented in the Lotus Domino technology, which provides an environment well
adapted for the design of collaborative applications. The application logic was implemented mainly
using the Java programming language.
This platform provides a ReST (Representational State Transfer) based API (Application
Programming Interface) (Fielding & Taylor, 2002) opening its data and activities to any external third-
party services willing to monitoring the community’s activity, and allowing also to receive any
attention-related interventions from them. More precisely, this API provides:

Access to the resources of the community: This API gives access (through an XML
representation) to a variety items of the platform such as: (1) the documents, including both
the content as well as meta-information such as associated tags, dates and statistical
information; (2) the user’s profiles with identification, basic description, dates, statistical
information, competence and interest, (3) and general information about the community e.g.
the list of members or the most popular documents.

Access to the activity of the community: the platform exports via several Atom RSS feeds
representing the events happening in the platform such as (1) document-centric events (all the
activity related to a specific document e.g. creation, view, comment); (2) user (or group)-
centric events i.e. all the activity of a given member (view of a user’s profile, creation, tagging
or comment of a document); (3) tag-centric events all the activity associated to a tag
(document tagged).

The possibility to execute external interventions: External components can create interventions
to notify users about information or make suggestions (guidance). Interventions may later be
delivered in the form of a message displayed in a portlet, via the availability of new
information in the platform (information that has been inferred or fetch from other sources), or
the execution of an animation by an artificial character (an embodied agent).
The subscription to the community’s activity (a list of events) allows an external entity to monitor in
real-time the community’s activity. The decoding of this information is facilitated by the tagging of
each event according to different aspects: the type of actions, the resources and users involved in the
interaction. Such tagging allows to get a higher level of semantic of events more easily processible by
machines and visualised by humans (advanced RSS readers are taking advantage of the tags when
visualising information).
1.4.2.2 The external agent
The agent is an external component that has been implemented using the Java programming language,
and that runs on a Tomcat web server. The platform and the agents for the test were running on
different servers. As indicated previously, this agent was developed separately so as to reduce the
complexity of the whole system and provide the maximum of flexibility.

This external agent observes the activity of the community and generates attention-related
interventions. An intervention is constituted of a predicate (corresponding to the identification of a
particular state), and of an action or suggestion. For instance if a new user has not completed his
profile, the agent send a reminder to suggest to complete it. The external agent can generate different
types of interventions from basic to advanced ones such as a reminder of a task completion
(completing a user profile), the observation and notification of collective or individual bursts of
attention related to a given resource, or advices about user’ practices related to his/her attention
management, (e.g. diversifying user’s attention toward other topics or users, or, on the contrary, its
reduction).
The cycle of functioning of this agent can be summarized by the following steps (see Figure 2):
1- The users interact with the platform generating digital traces (Latour, 2007);
2- The agent observes the community’s activity using AtgentNet’s API;
3- This agent (perception layer) mines this information (Anjewierden & Efimova, 2006;
Wolpers, Najjar, Verbert & Duval, 2007; May, George & Prévôt, 2008) and conducts
operations such as data fusion, pattern matching making use of heuristics, user profiling
(Nabeth, 2008);
4- Finally it generates a personalized intervention (Pierrakos, Paliouras, Papatheodorou &
Spyropoulos, 2003; Mobasher, 2007) and sent it to the platform using the API. The platform
then executes this intervention, displaying the result using either an embodied character, or a
porlet.

1.4.3 Mechanisms supporting social attention in AtGentNet
AtGentNet integrates a variety of mechanisms supporting attention implemented directly at
platform level, or via the agent module, and designed to provide a user-centric perspective of the
system. In the previous paragraph, we have already mentioned some of these mechanisms, such as the
display of information contributing to the social translucence, or the intervention of a proactive agent
proposing some guidance.
1.4.3.1 Social perception support
Social translucence: AtGentNet is designed so as to make the social dimension and social activity
visible (social translucence) and to reinforce the idea that the environment is primarily about social
interaction and networking and not about managing documents. Pictures of participants are therefore
displayed everywhere. Besides, traces of the activities of members and indicators of popularity of
resources are largely displayed to convey the feeling of people participation. For instance the
homepage displays the list of members recently connected as well as the list of the most popular
documents. The display of a document includes the pictures of the author and of the last readers. The

display of the page associated to a tag includes the document annotated with this tag but also the
pictures of the authors who have the most used this tag. Finally many statistics are available such as
the listing the most active contributors, or the listing of people who connect the more often.
Abstract view of the social dimension of a list of items: The platform is also able to abstract
information at the community level according to the three dimensions: people, resources and concepts
(i.e. tags). For instance, when a list of postings is displayed, a member can immediately get an
aggregated view of the most popular authors of the items in this list, of the concepts associated these
items, an aggregated view, and the most popular item included in this list.
Reporting of bursts of attention: The agent module informs members of abnormal level of attention of
the documents they have authored. This category of mechanism may be extended in the future, to
notify members of an abnormal level of attention given to concepts they have shown interest to.
1.4.3.2 Deliberation support
Assessing the social impact of their actions: This platform provides the possibility for members to
assess the social impact of their actions. For instance, when posting a document, an author can see not
only the number of visits but also who the readers are. At the platform level (Figure 3), members can
also know which are the people they are getting the most attention from (“people who read me” or
“people having visited my profile”).

Figure 3: Who read me, who do I read

Such information allows a user to decide more confidently to engage or not interactions with another
one. At the community level, this information may be useful in determine the most effective tactics.

For instance targeting people having already the attention of a user (e.g. have read several documents
authored by this user) will probably be more effective than targeting people that appear that do not
appear to be aware about this user.
Assessing the attention given to others: The platform makes each member aware of people he/she is
dedicating most of his/her attention to. For instance in the user profile (Figure 3), the platform displays
the members receiving most of the user attention (“people I read”). The platform also allows users to
determine documents or concepts (tag) to which they are dedicating most of their attention as a
consumer (“tags of my last reading”) or as a contributor (“tags of my last postings”) (Figure 4).

Figure 4: Stated and observed competences and interests

Network visualization: Network visualization displaying the relationships between members and
resources (such as authorship of a document, readership to a document, contribution to concepts,
social relationships, etc.) provides an alternative way to navigate the maze of information and to
facilitate the decision making process.
1.4.3.3 Operational support
AtGentNet makes available a number of mechanisms for operational support such as assisting users in
accomplishing a task, or reducing the number of steps that they need to execute an operation. For

instance, in the last case, the platform is designed so as to allow users to access rapidly the most
relevant information (such as a list of the last unread posts available from the home page, or access to
the information using the tags).
Reminder to complete a task: The agent module intervenes to provide assistance to users in
accomplishing some tasks, such as in managing assignment in reading documents, or in reminding
them of the importance of completing their user profile.
Presentation of the community by an embodied character: The agent detects the creation of new
accounts and then propose a presentation of the platform to the new users using an embodied character
(cf. Figure 1). The agent may also notify the community as a whole of a new member, facilitating the
establishment of social relationships.
Aggregation of information: The exporting of RSS feeds also contributes to reduce the cognitive load
of users. RSS readers, by aggregating different sources of information in a single place, reduce the
number of pages that a user need to consult.
Watch list: AtGentNet implements the watch list feature, allowing a user to track in a single place, the
activities related to documents, tags or people. A user may use this feature to know who is reading /
replying to some documents, or to know of the activity of the members of a group (e.g. the date of
their last connection, the number of documents they have created)
Tracking inattentive members: Furthermore, the agent can notify an author about (inattentive)
members who have not yet read a document assigned to them. This feature can be especially useful to
teachers for supervising students.
1.4.3.4 Meta-cognition
Finally, many of the mechanisms provided by AtGentNet allow users to self reflect about their
behaviour in a perspective of self-improvement, but also contribute to their motivation (and therefore
influencing the level of attention they dedicate in interacting with others).
More specifically, these mechanisms include the social translucence visualisation tools, the assessment
of the impact of actions, or the availability of many statistics indicating the usage of the platform (how
often the person connects, how much she contributes and when, or what are her different types of
actions).
Users’ intention versus users’ attention: it is important to distinguish between the information that
people declare (such as acquaintances, or competence), and the information that reflect their real
actions, this later being the one that really matters (Huberman, Romero Dand Wu, 2008). For instance
this is the case when a user has stated a particular interest on a topic but never reads a document on
this topic, or when a user has indicated an expertise on a subject but never provides any input on this
subject. The platform AtGentNet displays in one page the aggregated list of tags of the documents
that a user has read or has authored (Figure 4). The display of this information in parallel with the one

that the user has explicitly declared (such as stated competence, stated interest, or list of
acquaintances) helps to identify dissonances in her attention allocation.
Tracking the diversity of social attention: The agent also notifies users when they read only
contributions generated by the same group of members, since this is associated with low social
diversity. The objective of this intervention is to augment the cohesion and the social cognition of the
whole community.
1.4.3.5 Summary
The following table summarizes (incompletely) and categorizes by level (P: perception, D:
deliberation, O: operation, MC: meta-cognition) the different mechanisms.

mechanism P D O MC Comments (and examples)
Picture of participants X x P: the use of pictures of participants in many different
places as traces of their activities
MC: Contribute to the motivation.
Abstract view of the
social dimension of a
list of items (tags or
documents)
X P: Making visible the social dimension of the
resources. Facilitating social serendipity (via social
navigation).
Reporting of burst of
attention at the
community and
individual levels

x X x P: Notifying an abnormal high level of concentration
of attention from the community or from a user
D: Engage more confidently an interaction with people
seeming interested
MC: such burst of attention is an indicator in the
assessment of his influence in the community or among
certain members (individual level)
Showing members
getting user’s attention
(list of “people I read”)
+
Showing the social
impact of a user (list of
“people who read me”)
x X X P: social translucence: showing the related social
impact at the community and member level of
contributions of a user.
D: Engaging more confidently interactions and
augmenting globally the social cognition of the
community thanks to social reciprocity factors (if A
knows that B pays attention to him, by reciprocity he
will be influenced in his decision making by paying
more attention to B).
M: provide a way for a user to evaluate and improve
his strategy of communication
Visualizing the social x X P: allowing users to discover hidden social relationship
D: such tool supports users to identify the more useful

Network contacts.
Reminder about
completing the user
profile
X O: reducing the cognitive effort to remember
Presentation of the
community
X O: Educate users about how to use the platform.
Watch Mechanism

X O: Facilitating and reducing the cognitive effort to
follow of a discussion
Tracking inattentive
members
x X P: for the author: provide a view of the inattentive
members of his posting (i.e. show the non activity)
MC: for the inattentive members: such mechanism
provide a way to be aware of their behaviour and
change their attention management
Assessing user intention
vs. user observed
attention
X MC: Such tool provides a way to detect good or bad
practices in attention management by showing
similarity or dissonance between what was declared by
user and his real foci of attention.
Tracking social diversity X MC: assess his management of his social attention and
provide suggestions to reduce or open his social
attention to other members.
Basic statistics about
attention allocation
X MC: some general statistics (number of posts read,
write, tag used etc, people watched) help the use to
assess his attention allocation

The table below summarizes the different (proactive) agent interventions:
Role Condition Action Attention support
Presentation:
Presentation of the
platform: Generation of
guidance about how to use
the platform (using an
artificial character)
Detection of an event
about the creation of an
user's account
Propose a tutorial
about the platform
(using an artificial
character)
An overall presentation of
the platform will help the
user to be more effective in
using the platform.
Profile completion:
Reminder about the
completion of the user's
profile
At the 2nd connection
of a new user to the
platform
Remind user to
complete his/her
profile (notification)
A good user profile is
important to get the
attention of others.
Deadline approaching:
Alert inattentive members
to read a posting close to its
A few days before the
deadline date of a
posting, detection of the
Remind each
inattentive member to
read the document.
Reminder reduces the
cognitive load related to
back of head attention

deadline.

inattentive members
who haven't read a
document yet
(notification)
Tracking inattention:
Report to the author of a
posting the set of inattentive
members to it after a given
deadline.
At the posting's deadline
date, detection of the
members who haven't
still viewed the posting.
Report to the author' of
the posting about the
set of inattentive
members.
Assist the user in managing
the completion of
assignment to a given
audience.
Tracking readership:
Report to the author that
someone responded to
his/her posting
Detection of a response
(event "creating" +
relationship with a
parent posting)
Notification to the
author.
Contribute to the awareness
of the impact of a
contribution. Notification of
relevant information.
Collective burst of
attention:
Abnormal collective interest
for a posting or for a user’s
profile
Detection of an
abnormal high audience
for a posting (or a user
profile).
Suggestion to have a
look at this resource.
Contribute to collective
awareness.
Personal burst of
attention:
Abnormal interest from a
specific user for a posting
or for a user’s profile
Detection of an
abnormal high number
of view of a user for a
posting (or a user
profile).
Notification to the
author of the resource
of this special interest
from a user
Contribute to introduce or
reinforce link between
members . Contribute to the
awareness of the impact of
an individual contribution
Low social diversity:
A lack of openness to others
was observed

Detection of the
diversity of the user's
attention the last 2
weeks.
Suggestion to open-up
to other people
Contribute to create
awareness to others.
Table 2: agent interventions
1.5 ATGENTNET: APPLICATION IN A BLENDED LEARNING SCENARIO
1.5.1 The Context
The AtGentNet platform, and the mechanisms supporting attention in a social context, have been
tested during a pilot test that took place during a period of 6 months in 2007, in the context of the ITM
(International Trade Management) vocational training programme (Nabeth, Karlsson, Angehrn and
Maisonneuve, 2008). In this programme the participants attend to a series of local seminars, of joint
international seminars, and of monthly meeting with an expert coach. The objective of these meetings
is the elaboration of an export business plan.
The role of AtGentNet was to provide a learning platform to be used between the different
sessions, so as to make the physical sessions more effective. This platform therefore enables the

transformation of this programme into a blended programme, i.e. taking place both off-line and online.
This platform was used before the sessions to inform the participants, to deliver them some materials,
but also to contribute to the familiarisation of the participants with one another and with the faculty
members. The assumption was that this familiarisation process would create trust that would
contribute to the motivation of the participants, and would help in the building of a common
understanding and facilitate communication (Clark & Brennan, 1991). This platform was also used
after the sessions to consolidate the work conducted during the sessions, and to support knowledge
exchange and confrontation of ideas afterwards. The AtGentNet platform appeared to be particularly
adapted to support the context of the ITM programme: the participants of this programme are
“isolated” because of their geographical location, because they travel a significant part of their time
and because of the size of their organisation (SMEs) that makes them unlikely to exchange knowledge
with colleagues of similar expertise.
This pilot involved the participation of sixty people from seven countries (Greece, Hungary,
Iceland, Lithuania, Namibia, Norway, Slovenia, South Africa, and Sweden), and seven faculty
members (from Denmark, China, France, South Africa, The Netherland, and the UK), and was
launched at the first International seminar that took place at Lidköping, Sweeden in May 2007. This
first seminar provided the opportunity to present the participants with the ITM concept and to
introduce them to the platform. Participants were explained that they would be able to use the platform
as an interaction space to use between the seminars to strengthen their social relationships and to
exchange knowledge (such as experiences) once they have returned to their respective countries. After
this seminar, participants were organised into groups and received an account to connect to the
platform. The first group was provided access to a legacy collaborative platform that only provided
basic communication capability, and was not analysed. The second group (the control group) was
provided with an access to a restricted version of the new social network platform only offering a
subset of the functionalities. Finally the last group (the experimental group) was given access to the
full functionalities of the new platform, and in particular to the more advanced mechanisms supporting
attention (proactive agent interventions; watch list; advanced indicators).
Different actions were then initiated to stimulate the participants in engaging into an interaction so
as to generate a maximum of data for our analysis such as: a first phase of familiarisation; a series of
small light assignments related to the course; and finally an online role playing collaborative business
game was organised to boost participation. This game had been designed in the context of another
European research project L2C (Angehrn & Nabeth, 2006). The data collected for the analysis during
this AtGentNet pilot consisted in the log files of the activities of the users; the responses to a series of
questionnaires filled by participants; and some post-trial telephone interviews. The data was analysed
using statistical analysis (for the log files), but more qualitative methods were also used since the
small size of the sample did not always allowed concluding in a way that would be statistically
significant, but also so as to allow a higher level of analysis (Rudman & Zajicek, 2007).

1.5.2 The results
Before analysing the results, it is important to mention that one of the main issues for the utilisation
of social platforms is the question of participation. The profile of the population participating to the
test consisting is busy managers made the problem even more difficult: This category of participants
connects in the first place to the platform only if they consider that they are receiving tangible value
from this interaction. It is for this reason that the project decided to organise a simulation game, so as
to overcome the cold start effect and allowing the participants to assess the value of the platform in the
context of a very engaging activity.
A very noticeable difference was observed in the patterns of system use before and after the
simulation game was organised: before this game, the participants of the experimental group (who had
accessed to the more sophisticated mechanisms) were consistently showing more activity than the
participants of the control group (who only had access to the basic mechanisms). The observation of
the difference in usage between the two groups after the simulation indicated a less pronounced
difference.
The advanced mechanisms (provided by the agent module) mainly appear to play a role in
increasing the perceived value of the platform, and therefore the likeliness from a participant to
reconnect later. This can be considered as relatively disappointing, since we can imagine that this
element will fade once the novelty effect is over, and that the project was investigating the possibility
to provide advance support to attention (and not only the basic and passive support to attention). More
work need to be accomplished to design such mechanisms, but we can also expect that more maturity
from the users will improve the situation.
The data extracted from the log files provided a number of other findings. “Lurking” to other
people profiles (as an indicator of social interest) represents a behavioural pattern that is very
important, and extensively used: people connect to the platform not so much to interact with others but
to get information about others. This behaviour was later confirmed on the Alumni platform that was
set-up: people connect to the platform firstly to get information about other users. This is something
that is consistent with the large adoption of online social networking services such as LinkedIn or
FaceBook that we can observe today.
The data extracted from the questionnaires and the telephone interviews helped to refine, to
elaborate or to add to the previous findings. First the participants described themselves as busy people
strongly involved with their regular work, and unable to justify dedicating time to an activity not
generating tangible value. The comparison of the answers originating from the two groups indicates
that the more advanced platform helps the understanding of the use of the platform, but also eases the
access to the documents, confirming users’ interest in mechanisms helping them to be more efficient.
The participants of the two groups liked the ability of the platform to help them maintaining business
and social connection with the other participants. They also expressed their interest in being able to get

in touch with the lecturers, as well as having the possibility to collaborate with colleagues they had
met at the seminar. However, on the latest point, the observation shows that they engaged in these
activities mostly when they had been organised and generated clear benefit (such as in the case of the
game), and were reluctant to engage into informal interactions.
Finally, all the participants indicated that they found the interface too complex. The reason of this
complexity has its origin in an underestimation of the level of distraction of displaying too much
information in a page following the desire to provide a maximum of social translucence to the system.
The interface of second version of the prototype was simplified so as to address this problem.

1.6 DISCUSSION AND CONCLUSION
This project generated a number of findings that could be derived from the design of a social
platform supporting attention at the social level and from the empirical study that was conducted to
evaluate this platform in the context of supporting a blended learning programme.
1.6.1.1 Attention: an abstract concept
First, we have to point out the difficulty we experienced “manipulating” the concept of attention in
a way that is concrete enough and that makes sense to users, and that can be “mapped” into a technical
implementation. Attention refers to a relatively abstract concept that is grounded in cognitive sciences,
and that may not easily be transformed into artefacts directly manipulated by end-users, and be later be
evaluated. Thus, if people may understand what the concept of social information overload means, and
even experiment it in their real life, they may not be ready to incorporate this concept straightaway to
their “day to day” thinking, but would rather like to think in terms of concrete mechanisms that will
help them be more effective in their activities. Therefore in this project, we designed a platform
supporting people in “being more effective interacting with others”, and not branded as a platform
“supporting their social attention”. This orientation was further reinforced by the difficulty for us to
measure “attention”. Attention, even when it is long term attention, can not easily be measured
directly, since much of the associated data in not currently accessible, and would probably require the
monitoring of the brain itself, or the capture of people action over long period of time. On the other
hand, the capture and the exploitation of people activities (i.e. available as the digital traces that people
leave when they interact) that we were able to conduct in AtGentNet already proved us it was possible
to design meaningful systems supporting attention with the existing data. Of course progress will need
to be made on this subject, and in particular in knowing how to extract from these digital traces the
data representing people attention. We have some reasons to believe that we are only at the beginning,
and that attention aware systems will increasingly be available: The development of the Web 2.0, and
more generally of the information society, is making available an increasing large amount of traces
that can be exploited to monitor people activities, even if some restriction linked to the protection to

people privacy may apply. The Web 2.0 has also created the conditions, in which people have to be
very effective at managing their interaction with others, and we do not expect this to change in the
future.
1.6.1.2 Implementing the four levels model
This project offered us the possibility to design an attention aware system based on the
operationalisation of the four model of attention proposed by Roda and Nabeth (2008). Our conclusion
is that the application of this model is functioning, but appears to have been less straightforward than
anticipated.
We were able to verify that the support of social attention according to the four levels (perception,
deliberation, operation and meta-cognition) was making sense since we could find for each level a set
of meaningful mechanisms supporting attention. We were also able to implement meaningful
mechanisms at the four different levels. The technical architecture appears to have been functioning
very well, and according to our expectation. In particular the separation of the advanced mechanisms
using an external agent, from the platform that was providing only the communication capability
proved to be valuable.
However, we were confronted to a number of findings we had not anticipated.
First, we found that mechanisms may provide support to attention to several levels at the same
time. Actually, this is coming not too much as a surprise, but it makes the application of the four level
of attention support a little more complex than what we had originally thought.
Second, and more interestingly, we found that the support of attention at different levels may
sometime conflict, and that more work would need to be conducted to understand better the
relationship between the support of attention at different levels. Practically in the first version of the
prototype, we made an important use of translucence mechanisms. Our intention was to inform the
user as much as possible using a minimum of steps. In other words we wanted to maximise the support
at the operational level, and reduce the number of steps required for accessing this information.
Unfortunately this approach proved to be counter productive since it generated information overload,
and resulted in reducing the “quality of the perception” of the users. At the same time we are
observing that web 2.0 sites are proposing richer and more complex interface. Online networking
services such as LinkedIn or FaceBook display a large quantity and variety of information on a same
page such as the activity stream, suggestions to connects to other people, or applications (such as
slides). It will be interesting to follow this evolution towards richer and more complex user interfaces,
and if people will learn to use them effectively or on the contrary, if the provider of these services will
have to go back to the design to simpler interfaces.
Third found that the less sophisticated mechanisms such as the ones contributing to the social
translucence have proved to be the most effective. This is something somewhat disappointing if we
consider that the first objective of this research was to investigate how to provide advanced support to

attention, and in particular to explore how proactive and intelligent mechanisms implemented by an
agent could be used for this purpose. Yet at the same time were able to observe that the most advanced
mechanisms were having a positive impact. Besides, we have to admit that we only implemented a
limited set of advance mechanisms supporting attention, and that the mining and the exploitation of
digital traces (a core element of our research) is a subject under important investigation by the research
community, (Anjewierden & Efimova, 2006; Wolpers, Najjar, Verbert & Duval, 2007; May, George
& Prévôt, 2008) and which represents a very promising direction for further work.

1.7 REFERENCES:
Angehrn Albert A., Nabeth Thierry (2006); The L2C Project: Learning to Collaborate through
advanced SmallWorld Simulations; ECTEL 2006; First European Technology Enhanced Learning
Conference, 1-4 October 2006, Crete, Greece
Anicic, D., N. Stojanovic, and D. Apostolou (2008), Enterprise Attention Management System, in M.
Hepp, N. Stojanovic, K. Hinkelmann, D. Karagiannis, and R. Klein (Eds) Proceedings of the 3rd
International Workshop on Semantic Business Process Management (SBPM 2008) in conjunction
with the 4th European Semantic Web Conference (ESWC 2008) Tenerife, Spain, June 2, 2008
Anjewierden Anjo, Efimova Lilia (2006); Understanding Weblog Communities Through Digital
Traces: A Framework, a Tool and an Example; On the Move to Meaningful Internet Systems 2006:
OTM 2006 Workshops (2006), pp. 279-289.
Argyris, C., & Schon, D. (1978). Organizational learning: A theory of action perspective. Reading,
MA: Addison-Wesley.
Ashkenas, Ron. (2007), Simplicity-Minded Management: A practical guide to stripping complexity
out of your organization, Harvard Business Review 85 (12) : 101-109
Bandura, A. (1994), Self-efficacy. In V. S. Ramachaudran (Ed.), Encyclopedia of human behavior
(Vol. 4, pp. 71-81). New York: Academic Press.
Bandura, A. (2001), Social Cognitive Theory: An Agentic Perspective, Annual Review of Psychology
52 : 1-26.
Beck John C. and Davenport Thomas H. (2001); How Corporate Leaders Can Help their Companies
Manage the Scarcest Resource of All: Attention; Accenture Point of View, August 2001. url
http://www.accenture.com/Global/Research_and_Insights/Outlook/By_Alphabet/HowAttention.htm
(last accessed the 20 November 2008)
Bier, E. A., Stone, M. C., Pier, K., Buxton, W., & DeRose, T. D. (1993, September), Toolglass and
magic lenses: The see-through interface, Proceedings 20th annual conference on Computer graphics
and interactive techniques, 73-80.
boyd danah and Ellison Nicole (2007), Social Network Sites: Definition, History, and Scholarship,
Journal of Computer-Mediated Communication, 13 (1), article 11.

Bridges, K. (2008); Librarians and the Attention Economy; Library Philosophy and Practice, March
2008. Url: http://libr.unl.edu:2000/LPP/bridges3.htm (last accessed: 19 November 2008)
Burt Ronald (1997); The contingent value of social capital; Administrative Science Quarterly 42, No.
2, 339–365 (1997).
Chi, Ed H. (2008), The Social Web: Research and Opportunities, IEEE Computer 41(9) : 88-91.
Chi Ed H., Pirolli Peter, Suh Bongwon, Kittur Aniket, Pendleton Bryan, Mytkowicz Todd (2008),
Augmented Social Cognition, In Proceedings of AAAI Spring Symposium on Social Information
Processing. AAAI Press, March 26 2008.
Chi, E., Pirolli, P., & Lam, S. K. (2007). Aspects of augmented social cognition: Social information
foraging and social search. Proceedings of the Human-Computer Interaction International
Conference.
Clark, H. and Brennan, S. (1991), Grounding in Communication, In L.B. Resnick, J.M. Levine, and
S.D. Teasley (eds.), Perspectives on Socially-Shared Cognition. APA, Washington, D.C.
Cross, R., Borgatti, S. P. and Parker, A. (2002), Making Invisible Work Visible: Using Social
Network Analysis To Support Strategic Collaboration, California Management Review 44 (2) : 25-
46.
Cunningham W. and B. Leuf (2001), The Wiki Way. Quick Collaboration on the Web, Addison-
Wesley.
Davenport, T. H., & Beck, J. (2001), The Attention Economy, Harvard Business School Press.
Dieberger, A., P. Dourish, K. Höök, P. Resnick, and A. Wexelblat (2000), Social navigation:
techniques for building more usable systems, Interactions 7 (6) : 36-45.
Drucker, P. F. (1999), Knowledge Worker Productivity: The Biggest Challenge, California
Management Review 1 (2) : 79-94.
Dunbar, Robin (1992), Neocortex size as a constraint on group size in primates, Journal of Human
Evolution 20 : 469-493.
Erickson, T. & Kellogg, W. (2003), Social Translucence: Using Minimalist Visualizations of Social
Activity to Support Collective Interaction, in Designing Information Spaces: The Social Navigation
Approach. (Eds. Höök, K., Benyon, D. and Munro, J.) Springer, 2003, pp 17-42
Erickson, T. (2009), ‘Social’ Systems: Designing Digital Systems that Support Social Intelligence, AI
& Society 23 (2) : 147-166.
Fielding, Roy T.; Taylor, Richard N. (2002), Principled Design of the Modern Web Architecture,
ACM Transactions on Internet Technology 2 (2): 115–150
Gifford, Sharon (1997), Limited attention and the role of the venture capitalist, Journal of Business
Venturing 12 (6) : 459-482
Gill, K.E. (2005); Blogging, RSS and the Information Landscape: A Look at Online News, WWW
2005, Chiba, Japan.
Giraldeau L.-A. and Caraco T. (2000), Social Foraging Theory, Princeton University Press.

Golder Scott and A. Huberman Bernardo. (2006), Usage Patterns of Collaborative Tagging Systems,
Journal of Information Science 32 (2) : 198-208.
Granovetter, M. (1973), The Strength of Weak Ties, American Journal of Sociology 78 (6) : 1360-80.
Halpin, H., Robu, V., Shepherd, H. (2006), The Dynamics and Semantics of Collaborative Tagging,
In: Möller, K.; de Waard, A.; Cayzer, S.; Koivunen, M.-R.; Sintek, M.; Handschuh, S.: SAAW’06,
CEUR Workshop Proceedings.
Harper M., X. Li, Chen Y., Konstan J. (2007), Social Comparisons to Motivate Contributions to an
Online Community, Persuasive Technology conference, Palo Alto, CA.
Homans, George C. (1958), Social Behavior as Exchange, American Journal of Sociology 63 : 597-
606.
Huberman, Bernardo, Wu, Fang, (2008), The Economics of Attention: Maximizing User Value in
Information-Rich Environments, Advances in Complex Systems 11 (4) : 487-496
Huberman Bernardo A., Romero Daniel M. and Wu Fang (2008), Crowdsourcing, Attention and
Productivity, Working Paper HP Labs, 2008-09-11. url
http://www.hpl.hp.com/research/idl/results.html (accessed the 20 November 2008)
Huberman Bernardo A., Romero Daniel M. and Wu Fang (2008b), Social networks that matter:
Twitter under the microscope, Working Paper HP Labs, 2008-12-05. url
http://www.hpl.hp.com/research/idl/results.html (accessed the 20 November 2008)
Huberman, Bernardo, Wu, Fang, (2007), Comparative Advantage and Efficient Advertising in the
Attention Economy, 22nd European Conference on Operational Research (EURO XXII) (2007)
Joinson, Adam N. (2008), Looking at, looking up or keeping up with people? Motives and use of
Facebook, SIGCHI 2008, 1027-1036.
Kogut, Bruce (2008), Knowledge, Options, and Institutions, Oxford University Press
Latour Bruno (2007), Beware your imagination leaves digital traces, Times Higher Literary
Supplement, 6th April 2007
Marlow, C., Naaman, M., Boyd, D., Davis, M. (2006), Position Paper, Tagging, Taxonomy, Flickr,
Article, ToRead, Collaborative Web Tagging Workshop (at WWW 2006), Edinburgh, Scotland,
May 22
May Madeth, George Sébastien, Prévôt Patrick (2008), A Closer Look at Tracking Human &
Computer Interactions in Web-Based Communications, International Journal of Interactive
Technology and Smart Education (ITSE) 5 (3) : 170-188
McAfee, A. P. (2006), Enterprise 2.0: The Dawn of Emergent Collaboration, Sloan Management
Review 47 (3) : 21-28
Miller, G. (1956), The magical number seven, plus or minus two, Psychological Review 63 : 81-97.
Mobasher B. (2007), Data Mining for Web Personalization, In: Brusilovsky, P., Kobsa, A.,, Nejdl, W.
(Eds.): The Adaptive Web, Methods and Strategies of Web Personalization. Lecture Notes in
Computer Science 4321 Springer

Nabeth Thierry (2008), User Profiling for Attention Support for School and Work, Book Chapter in
Mireille Hildebrandt and Serge Gutwirth Editors (2008), Profiling the European Citizen; Springer
Nabeth Thierry, Karlsson Hasse, Angehrn Albert A., Maisonneuve Nicolas (2008), A Social Network
Platform for Vocational Learning in the ITM Worldwide Network, IST Africa 2008, Windhoek,
Namibia, 14 - 16 May 2008
Nabeth, Thierry, Roda Claudia (2006), Les espaces sociaux virtuels: approches, pratiques émergentes
et perspectives, Invited chapter in the book Capital Immatériel, Connaissance et Performance,
Ahmed Bounfour Editor, L'Harmattan. pp. 225-265
Nabeth Thierry, Roda Claudia, Angehrn Albert A., Mittal Pradeep Kumar (2005), Using artificial
agents to stimulate participation in virtual communities, IADIS International Conference CELDA
(Cognition and Exploratory Learning in Digital Age)
Nardi, Steve Whittaker, and Heinrich Schwarz (2000), It's Not What You Know, It's Who You Know:
Work in the Information Age by Bonnie, First Monday 5 (5) URL:
http://firstmonday.org/issues/issue5_5/nardi/index.html
Nishikant Kapoor, Konstan Joseph, & Terveen Loren (2005), How Peer Photos Influence Member
Participation in Online Communities, Proceedings of ACM CHI 2005, April 2–7, 2005, Portland,
Oregon, USA
Ocasio, William (1993), The structuring of organizational attention and the enactment of economic
adversity : a reconciliation of theories of failure-induced change and threat-rigidity, Working paper
(Sloan School of Management), pp. 3577-93. url http://hdl.handle.net/1721.1/2473 (accessed the 20
November 2008)
Ocasio, William (1997), Towards an attention-based view of the firm, Strategic Management. Journal
18 : 187-206.
O'Reilly T. (2005), What Is Web 2.0 Design Patterns and Business Models for the Next Generation of
Software by Tim O'Reilly, 09/30/2005 url:
http://www.oreillynet.com/pub/a/oreilly/tim/news/2005/09/30/what-is-web-20.html
Perez Sarah (2008), Real People Don't Have Time for Social Media, ReadWriteWeb, April 16, 2008
url: http://www.readwriteweb.com/archives/real_people_dont_have_time_for_social_media.php
Pierrakos D., Paliouras G., Papatheodorou C., Spyropoulos C.D. (2003), Web Usage Mining as a Tool
for Personalization: A Survey, In: User Modeling and User-Adapted Interaction 13 : 311-372,
Kluwer Academic Publishers
Rashid, A. M., Ling, K., Tassone, R. D., Resnick, P., Kraut, R., & Riedl, J. (2006), Motivating
participation by displaying the value of contribution, In CHI '06: Proceedings of the SIGCHI
conference on Human Factors in computing systems, pp. 955-958. New York, NY, USA: ACM
Press.
Rudman Paul, Zajicek Mary (2007), Final Evaluation Report, AtGentive Project, Deliverable D4.4.
Available at: http://www.calt.insead.edu/LivingLab/AtGentive/Wiki/?D4.4

Roda C., Nabeth T. (2006), Poster: The AtGentive project: Attentive Agents for Collaborative
Learners, ECTEL 2006, First European Technology Enhanced Learning Conference, 1-4 October
2006, Crete
Roda Claudia, Nabeth Thierry (2008), Attention management in organizations: Four levels of support
in information systems, In A. Bounfour (Ed.), Innovation and Information Systems. Routledge
(Advanced Research Series in Management).
Resnick, Paul, Zeckhauser, Richard, Friedman, Eric, and Kuwabara, Ko (2000), Reputation Systems;
Communications of the ACM 43 (12) : 45-48.
Rogers, E. (2003), Diffusion of Innovations, (5th ed.), New York: Free Press.
Senge, Peter (1990), The Fifth Discipline: The Art and Practice of the Learning Organization,
Doubleday
Simon, H. A. (1971), Designing organizations for an information rich world, In: M. Greenberger (ed.),
Computers, Communications and the Public Interest (Johns Hopkins Press, 1971), pp. 38–52.
Stone, Linda (2006), Linda stone's thoughts on attention and specifically, continuous partial attention,
url http://continuouspartialattention.jot.com/WikiHome (accessed the 20 November 2008).
Subramani M.R, Rajagopalan B. (2003), Knowledge-sharing and influence in online social networks
via viral marketing, Communications of the ACM, 46, (12) : 300-307.
Tapscott D., Williams A. D. (2007), Wikinomics: How Mass Collaboration Changes Everything, New
York: Penguin.
Thomas, J. C., Kellogg, W. A., and Erickson, T. (2001), The knowledge management puzzle: human
and social factors in knowledge management, IBM System Journal 40 (4) : 863-884.
Thibaut, J. W., and Kelley, H. H. (1959), The social psychology of groups; New York: Wiley.
Tufekci Zeynep (2008), Can You See Me Now? Audience and Disclosure Regulation in Online Social
Network Sites, Bulletin of Science, Technology and Society.
Vassileva J., Sun L. (2007), Using Community Visualization to Stimulate Participation in Online
Communities, e-Service Journal 6 (1) : 3-40. (special Issue on Groupware).
Vertegaal, Roel (2003), Attentive User Interfaces, Communications of the ACM 46 (3) : 30-33
Vertegaal, Roel; J. S. Shell, Chen D. and Mamuji A. (2006), Designing for augmented attention:
Towards a framework for attentive user interfaces, Computers in Human Behavior 22 (4): 771-789
Weiss A. (2005), The Power of Collective Intelligence, netWorker 9 (3): 16-23.
Wolpers Martin, Najjar Jehad, Verbert Katrien, Duval Erik (2007), Tracking Actual Usage: the
Attention Metadata Approach, Educational Technology & Society 10 (3) : 106-121, Special Issue on
"Advanced Technologies for Life-Long Learning".
Wu, F. and Huberman, B. A. (2008), Popularity, novelty and attention, In Proceedings of the 9th ACM
Conference on Electronic Commerce (Chicago, Il, USA, July 08 - 12, 2008). EC '08. ACM, New
York, NY, 240-245.

Yuan, W., Chen, Y., Wang, R., Du, Z., (2007), Collective Intelligence in Knowledge Management, in
IFIP International Federation for Information Processing, Volume 254, Research and Practical
Issues of Enterprise Information Systems II Volume 1, eds. L. Xu, Tjoa A., Chaudhry S. (Boston:
Springer), pp. 651-655.