Context
and Intent in Call Processing
Tom GRAY1, Ramiro LISCANO2, Barry WELLMAN3,
Anabel QUAN-HAASE3, T. RADHAKRISHNAN4, Yongseok CHOI4
1 Pinetel,
Ottawa, Canada
2 Carleton
University, Ottawa, Canada
3 University of
Toronto, Toronto, Canada
4 Concordia
University, Montréal, Canada
Pp. 177-84 in Feature Interactions
in Telecommunications
and Software Systems VII, edited by Daniel Amyot and
Luigi Logrippo.
Amsterdam: IOS Press: 2003.
Abstract. A new feature set suited for IP telephony is
described. The user value of this feature set is discussed in terms of social
science results. This feature set supports natural human collaboration within a
human environment. The use by humans of casual awareness to support
collaboration is discussed and models taken from architectural research are
used to show how casual awareness may be facilitated within the physical and
communication environments. A tripartite policy-based architecture which can
support this new feature set is described.
1
Introduction
Mobility at both at the personal and device levels will be the hallmark of VoIP systems. The network will be able to track user location through both automatic and manual registration services [1]. Personal wireless mobile devices will supply the user with always-on connectivity. These capabilities of VoIP and wireless technologies will force a sea change in the types of features that are supplied by telephony systems. Current systems find much of their customer value by being able to find the called user. Features such as voice mail and the various forms of call forwards were designed to find the called user through both space and time. Next generation systems will be able to locate the user at all times. The current feature set will be a supplanted by a new one that is tailored to the specific capabilities and limitations of VoIP.
This paper is a brief description of a major project that was sponsored by Mitel Networks Corporation. Research was sponsored with a number of academic research groups and undertaken internally to investigate the types of features that will be supplied by VoIP systems. Work was undertaken to determine the basic customer value of IP telephony features and feasible architectures that could provide such services.
2
Features in IP Telephony
2.1
Location Systems and Their Effect on Feature Value
Location awareness is an integral part of the mobility capability of the VoIP architectures and protocols being designed today. This type of awareness raises fundamental issues about the utility and value of the feature set supplied with current communication and specifically telephony systems. If users can be found at all times, the implicit ability of current system systems to filter calls by user location will no longer be available. Users’ collaborators will know that they are locatable by the system and will expect that they will be available for what they consider important calls at all times. However, the uncontrolled use of location awareness and mobility services can be disruptive, as when cell phones ringing and conversations bother others in the vicinity.
Despite such annoyances, the capability of always-on, always-locatable systems can have much merit and can open the possibility for a completely new feature set. These features will enable a new type of customer value whose effect and worth have been the object of many years of social science research. They find their value by the enabling of greater user visibility and thus greater capability for informal and ad hoc interactions. We are already witnessing some of this in the new presence-based Instant Messaging (IM) communication services like Microsoft Messenger, Yahoo Messenger, and ICQ. All of these services are based on the ability of users being able to detect when other users are logged onto their computing device.
2.2
Social Science Background of New Feature Types
In bureaucratic organizations, communication between employees is regulated by the formal structure of the organization [9,10]. Who talks to whom is to some extent prescribed in the organizational chart. The information revolution has brought about changes in organizations’ communication structures leading to new forms of work, such as networked organizations, virtual organizations, and high-performance teams. In these new forms of work, organizations have moved from being bound up in hierarchically arranged, relatively homogeneous, densely knit, bounded groups (“little boxes”) to become social networks [11,12]. In a networked society, boundaries are more permeable, interactions are with diverse others, links switch among multiple networks, and hierarchies are flatter and more complex. In networked organizations, informal communications become essential because they provide employees with important information, new ideas, work opportunities, and influence [13]. Besides leading to the exchange of information, social relationships are responsible for the provision of tacit knowledge: knowledge that is sticky, difficult to transfer and comprises insights and deep understandings [15].
Research
showing the relevance of informal communication for the success of an
organization was published in the 1960’s with the pioneering work of Thomas J.
Allen of MIT [2]. His study shows that the success of R&D projects is
related to the amount of informal interactions both inside as well as outside
of the design team. Rob Cross at IBM’s Institute of Knowledge Management showed
that tacit knowledge and ideas are provided by close working relationships
[14]. It is not the informal communication per se, that seems to be responsible
for the success of teams, but specifically the opportunities for
problem-oriented and unplanned, spontaneous interactions that allow people to
take advantage of the collective knowledge available in the team. In the
architectural field, Bill Hillier provided evidence for the importance of
office space in fostering unplanned interactions [6]. Hillier recognized that
it is casual
visibility that promotes and facilitates informal interactions. His
comparison of high-tech labs shows that open spaces lead to more frequent and
spontaneous conversations between colleagues working on different teams
promoting cross-pollination of ideas and innovation.
AT
IBM, Thomas Erickson, Wendy Kellogg, and Erin Bradner have studied the
communication system, Babble, which promotes through
visibility interactions among employees [15,16]. The study shows that
visibility of employees’ work activities promotes conversations leading to
higher performance. They apply the notion of “social affordances” to describe
how the interface and its ability to depict the social world facilitates
information exchange. Babble has three important features: 1)
a display of all the actors who are online, 2) a display of their current
status of activity (active/inactive), and 3) a display of their involvement
with other actors. In this way, the system creates an interface that allows its
members to recognize each other’s availability and current status of activity,
enabling and inviting toward spontaneous interactions. Communication systems
similar to Babble are especially valuable for
colleagues, who work virtually and thus cannot monitor each other in the same
way that colleagues working co-located can. While many companies still work
co-located, the number of dispersed teams and virtual teams is increasing.
2.3
Features as Proactive Availability
Filters
The results of Allen, Hillier and
other social science researchers are useful for developing the always-on,
always-locatable capabilities of next generation communication systems. These
systems will move beyond the ‘user-finding’ value of current systems to attempt
to engender the types of informal activities as Hillier is doing in his
building designs. Hillier is accomplishing this by increasing the casual
visibility within the office space. New communication systems could do the same
thing by proactively projecting the availability of potential collaborators to
other users. Instead of the typical telephone tag played with today’s systems,
users with a glance at their telephone displays will see who is available for
conversation. They will then be able to make quick informal calls or set up
informal conferences to discuss issues of the moment. Voice mail has turned
POTS into an asynchronous messaging system. These new capabilities will return
the telephone to being a synchronous device for spontaneous conversation.
Availability
has been implicitly defined above as the willingness to talk to someone about a
topic of the caller’s interest. This definition reveals much on how these new
VoIP systems will work. Current systems attempt to locate called users. Next
generation systems will be always aware of where a user is [7]. They will thus
act more as filters that allow users to be visible to their colleagues and yet
protect them from unsuitable interaction. In the physical office that Hillier
is providing, this can be provided by visible clues. Users can indicate their
desire for momentary privacy by the position of their door, their posture, the
volume of their voices etc. Users are in multiple physical locations, all of
which will have rules that indicate the types of interaction that may take
place there. Some of these physical clues have already been implemented into
electronic communication systems like the telepresence project described by
Buxton et al. [8].
These
types of informal interaction can be replicated with the mobile. always-on
capabilities of new systems. Users will be reachable not only at their desk as
in today’s systems but wherever they happen to be. The result of this is that
user interaction in the new systems will take place in multiple locations,
which will necessarily have multiple expectations for proper feature behavior.
Privacy indicators and filters equivalent to those available in the physical
office space will have to be provided. Feature operations that are suited to
users’ private offices will be entirely unsuitable when users are located in
meeting rooms. Operations suited to users working alone in their offices will
be unsuitable when they have important visitors. To be truly useful, an
always-on network must supply means to adapt its operation to these differing
circumstances in the environments of its users. The system will be required to
take action that is consistent with the current situation of users, both the
physical environment and the social context.
2.4
Roles and Groups in the Social Sciences
The social science work discussed
above reveals that users often collaborate in groups and social networks. It is
these groups and networks that provide users with the capability of informal
interaction that brings the benefits that these new communication systems seek.
As a result, communication systems will function best if they operate in
cognizance of the actual structure and functioning of groups and social
networks.
Groups
contain participants, each of whom has a status that is equivalent, for
example, to a job title. These statuses may be formal or informal. Examples
could be: manager, Java guru, secretary, company director, friend, etc. In an
enterprise situation, users will participate in one or more groups and social
networks. As such, they will have one or more statuses that define their
behavior.
Participants
relate to each other in role relationships that join a pair of statuses. Most
importantly to a call-processing discussion, participants’ behaviors in groups
and social networks are defined by the relationships between the roles they
fulfill and the roles of their collaborators. Examples of roles are
boss-secretary, boss–VIP customer, etc. Roles hold the policies that determine
the preferred and suitable interactions between participants. For example, when
people play the role of boss in an organization, their behavior as bosses will
be determined by their status and the status of those with whom they relate.
There will be entirely different behavior and acceptable feature operation in a
boss–secretary interaction than in a boss–VIP customer interaction.
The
functioning of roles is more complex in social networks than in groups. In
groups, users usually have stable statuses and roles, and they usually interact
routinely with the same set of group members. In social networks, users move
between work teams, and they usually relate to different other users in each
team. Moreover, their status and roles in one team may be different than in
others. Hence, awareness of social context is necessary for acceptable feature
operation.
In
both groups and networks, a role is a directional relationship. For example,
the boss-secretary role is very different from the secretary-boss role. The
policies that direct proper behavior are very different in each direction. The
required behavior and operation can be attached to the policies for each role.
Call processing to function within the mobile framework of next generations
must be sensitive to the identity of the roles that the users are currently
playing. Next generation call processing could be improved if it had the means
for extracting this role information automatically as well as specifying a priori specific
rules for classes of statuses (for
example all bosses) and specific statuses (for example the person who is the
immediate boss).
3
Architecture
and Functioning of Context-Aware Call Processing System
3.1
Next Generation Call Processing - CoC and
PoA
The
discussion above indicates the types of features that will be valuable in next
generation systems can be divided into two categories. Existing systems act reactively
to user action to provide services for incoming and outgoing telephone calls.
This reactive service will be maintained in a modified form in next generation
systems. However it will be supplemented by a set of proactive features that
will encourage useful informal discussion within an organization. In our work
we have classified the reactive feature type as CoC (control of communications)
and the proactive feature type as PoA (projection of availability).
Firstly, CoC features will act as call
filters that will be aware of users’ current situations. Secondly, PoA features
will be filters that analyze users’ current situations and project their
availability to potential collaborators based on that determination.
3.2
Context – Call Processing and Presence
Combined
There is much
current interest in the issue of presence and availability. Presence is defined
as users’ current contact information. This could be the directory number of
the telephones in the rooms they are occupying, their wireless numbers, the
directory numbers of the secretaries who can locate them, etc. Availability is
an indication of users’ willingness to converse given their current situations.
There are standards groups and industry
forums dedicated to the development of standards for these services. Because of
the obvious similarity between PoA and presence, we considered carefully the
issue of interaction between presence and call processing systems. We were
initially concerned about the potential for harmful interactions between these
two services. However in working with these systems, we soon realized that this
potential was only illusory. The availability indications provided by presence
systems are an indication of the users current wish to be contacted. However
users can be contacted only if their call processing policies allow it.
No matter what users’ availability
policies might say, if they have set their call processing features to Do Not
Disturb, then they cannot be contacted. Unlike some other standards, there are
no specific presence and availability policies. Rather, users’ current
availability can be determined as a result of a hypothetical call attempt. If
call processing indicates that a call will be completed given the users’
current policies, then they are available and not otherwise. Presence is really
a form of hypothetical call processing. Without this realization call
processing and presence would be very difficult to operate together as a
system. There would be major problems in making their policies compatible.
We have argued that call-processing
features must be aware of users’ current situation. The discussion above has
shown that users’ real presence or availability are determined by their current
call processing policies. This indicates to us that presence and call
processing are really aspects of a single larger composite entity. It is this
entity that will become the focus of next generation call processing. We have
identified this entity with context. Next generation telephony systems will
have their feature set defined beyond call processing and into this new area of
context.
3.3
Physical Context and Human Intent
Context has a research literature of its own. Anind K. Dey, a leading researcher in this field, has defined context [3] as ‘any information that can characterize the situation of an entity. An entity is a person, place or object that is relevant to the interaction between a user and an application including the user and applications themselves’. Note that this definition implicitly defines context to be a relationship that controls behavior much like the role relationship in sociology.
User context can be determined from sensing the environment. Context can determine users’ location, who they are with, what they are doing on the computer system, when it is, etc. The important aspect of this is that surmises may be made about users’ activity in the social context from the sensing of their physical context. That is, from sensing the physical environment, the system may make surmises about which statuses of users are currently active and which roles they are playing. This will allow the selection of features that are appropriate to the current social context. For example, if it is determined that a user is in a meeting room with VIP customers then it can be assumed that she is in the sales status and the current role is the sales-customer role. From this it could be surmised that she does not wish to be disturbed. If the company owner calls this user, then it can be assumed that she is available to take the call no matter when it is.
This is supported by Hillier’s development of his model of human interaction based on building structure. Rules can be created which link the physical context to the intent of the user in the human environment. Just as Hillier’s office space allows for physical clues to indicate a user’s current intent, it is possible for rules to link clues determined by sensors to the user’s intent. This linking of physical sensing to the determination of the user’s human intent will allow for the selection of features that are appropriate to the user’s current circumstances. It is an answer to the issues that wireless and VoIP create when they create the always-on, always-locatable communication systems.
The sensing of the who, where, when, and what in the physical environment will allow the system to surmise the why in the human environment. Next generation call processing will be context-aware so that it can function within the human environment.
3.4
Tripartite Architecture for Context-Aware Call Processing
Figure 1 shows a conceptual diagram of a policy-based, tripartite architecture for context-aware call processing. Policies of different forms will exist in each of the basic three sections shown in the illustration.
Figure 1: Policy-based, tripartite
architecture for context-aware call processing.
Awareness data is raw information about the physical environment. It will be gathered by sensors in users’ environments and by analysis of their interactions with the computer and telephony systems. This information will be analyzed by rules in a context engine that will make assertions about the users’ current situation. For example, it can take into account that a user is in a meeting room with others and that these others have registered at reception as visitors from an important customer.
The second part of the architecture is concerned with feature selection. Given actions by the user or requests from other users the appropriate feature or action to take can be determined. This feature selection will be done by rules that link the physical context to actions appropriate at the human level. This will implement the CoC filters required. Availability can be determined by hypothetical requests to this part of the architecture and projected as PoA indicators as shown. Amer et al’s policy architecture [17] presented at FIW00 gives an indication as to how these first two sections can function.
Third, once the proper feature has been selected it can be sent to the feature execution engine. We have considered using Barbuceanu’s et al [5] OPI policy system in this part of the architecture.
4
Current Work
4.1
Prototype Context-aware Call Processing System
A prototype context-aware call processing system was developed [4]. This system is based on a blackboard architecture implemented on top of a tuple space. This system is based on the tripartite architecture described in this paper. Novel methods of context determination and interaction resolution have been implemented within this system.
4.2
Policy Languages
We are partnering with several university research groups in the development of policy languages for the specification and the detection of feature interaction between policies. We expect that this work can be fitted into the tripartite architecture.
4.3
Context and User Activity Sensing
Work is going on to improve methods of sensing user environments and interpreting this as context. This includes means of sensing user activity on the computer system and physical location sensors to determine users’ location and users’ co-presence (whom they are with).
4.4
Test and Development Systems
A prototype system to simulate natural human interaction in an office environment has been developed. Users are represented by software agents that autonomously interact within a programmable office environment. These user agents are scripted with high-level goals to perform day-to-day activities such as working in a private office, attending meetings, looking for others to deliver messages, etc. The agents are provided with heuristics so that they can collaborate autonomously in performing their individual goals in typically human manners. This system is based on a blackboard architecture. It can thus be coupled directly to the prototype system described in this paper. Work to accomplish this is now going on.
4.5
Native Call Processing Engine
Questions are raised about the scalability of tuple space systems in which large amounts of information are interpreted in the handling of a single call. Work was done to improve the performance of software tuple spaces by the creation of high efficiency operators. To obviate this issue, a hardware-assisted tuple space was developed. This is a hardware device that is capable of matching tuples at hardware memory speed with the speed limitation being the memory’s access time only. This work is now being supplemented to create a hardware policy accelerator in which the policy rules themselves will be matched and executed at memory speed as well. This will enable the development of a native call-processing engine that exploits the possibility of policies.
5
Acknowledgements
We gratefully acknowledge our colleagues at Mitel Networks who were vital in the development of these ideas. We acknowledge the contribution of Katherine Baker and Natalia Balaba for their strong effort in the development of the presence system. We acknowledge the leadership that Daisy Fung and Peter Perry provided us. We acknowledge the contributions of Daniel Amyot, Gunter Mussbacher, Kelvin Steeden, Tonis Kasvand, Serge Mankovskii, Michael Weiss, Babak Esfandiari, Tom Ware, Eliana Peres and all former members of Strategic Technology in the genesis of these ideas. We would like to thank Debbie Pinard for initiating this effort within Mitel. We would also like to acknowledge the academic partners of Mitel Networks who include Ray Buhr, Ahmed Karmouch, Jo Atlee, Luigi Logrippo, Murray Woodside, Ken Turner, Stephan Reiff-Marganiec, Eric Yu, Daniel Gross, Roger Impey, Aurora Diaz, Sue Abu Hakima, Mihai Barbuceanu, Trevor Rainey, Colin Bänger, Hugh McLaren, Dave Athersych, Bernie Pagurek and Nick Dawes. We would like to thank Ernst Munter, Harry Kennedy, the MNS and the RPE for demonstrating how systems really work.
References
[1] Rosenberg,
J. et al, RFC 3261 SIP: Session
Initiation Protocol, http://www.ietf.org/rfc/rfc3261.txt, accessed on
December 15, 2002
[2] Allen,T., Managing the Flow of Technology:
Technology Transfer and the Dissemination of Technological Information Within
the R&D Organization, MIT
Press, Cambridge Massachusetts, 1984
[3] Dey,
A., Abowd, G. Towards a Better
Understanding of Context and Context-Awareness, GVU Technical Report
GIT-GVU-00-18, Graphics, Visualization and Usability Center, Georgia Institute
of Technology, 1999
[4] Choi,
Y. Context Based Delivery of Voice
Messages, Master’s thesis, Department of Computer Science, Concordia
University, Montreal, 2001
[5] Barbuceanu,
M., Gray, T., Mankovski, S., Coordinating
with Obligations, Second International Conference on Autonomous Agents,
Minneapolis 1998.
[6] Hillier,
B. Space is the Machine: A
Configurational Theory of Architecture, Cambridge University Press, 1996
[7] Schilit,B.
Adams,N. and Want.R. Context-Aware
Computing Application. IEEE Workshop on Mobile Computing System and
Application, December 1994.
[8] Buxton,
W. (1995). Integrating the Periphery and
Context: A New Model of Telematics Proceedings of Graphics Interface '95,
239-246.
[9] Monge, P. R., & Contractor, N. S.
(1997). Emergence of communication
networks. In F. M. Jablin & L. L. Putnam (Eds.), Handbook of organizational communication (2nd ed.). Thousand Oaks,
CA: Sage Publications.
[10] Monge, P.
R., & Contractor, N. S. (2001). Theories
of communication networks. Unpublished manuscript.
[11] Hiltz, S.
R., & Wellman, B. (1997). Asynchronous
learning networks as virtual communities. Journal of the ACM, 40(9), 44-49.
[12] Wellman,
B. (2002). Designing the Internet for a
networked society. Communications of
the ACM, 45(5), 91-96.
[13] Uzzi, B.
(1996). The sources and consequences of
embeddedness for the economic performance of organizations: The network effect.
American Sociological Review, 61,
674-698.
[14] Cross, R.,
Rice, R. E., & Parker, A. (forthcoming).
Information seeking in social context: Structural influences and receipt of
information benefits. Organization Science.
[15] Erickson,
T., & Kellogg, W. A. (2000). Social
translucence: An approach to designing systems that support social processes.
ACM Transactions on Computer-Human Interaction, 7(1), 59-83.
[16] Bradner, E.,
Kellogg, W. A., & Erickson, T. (1998). Babble:
Supporting conversation in the workplace. SIGGROUP Bulletin, 19(3), 8-10.
[17] Amer.M.
Karmouch, A. Gray, T. Mankovskii, S. (2000)
A multi-agent architecture for the resolution of feature conflicts in telephony.
Proceedings of the International Conference on Practical Application of
Intelligent Agents and Multi-Agents, 5th (PAAM 2000), April, Manchester, UK,
pages 53-74.