A Position paper for the W3C "Mobile Web Initiative" Workshop, November
Contact author David De Roure, University of Southampton, email email@example.com
David De Roure
The mobile devices used to access the Web today are typically interactive handheld devices which are used to access personal and public information, to store personal information and to communicate with others. These devices are an example of the increasing number and variety of pervasive or ubiquitous computing devices which are set to pervade our everyday environment, as parts of ‘wearables’, within everyday artefacts and embedded in the fabric of our living environment – this is the beginning of the realisation of the ubiquitous computing or Ambient Intelligence vision.
The advent of the mobile phone and wireless LAN technologies has exposed users to the benefits of tether-less communication; however, it could be argued that it has done little to break the bind between the service, device and the networking technology. Consequently, many of today’s users associate a particular network and device with a service, for example, the Internet and the PC with Web browsing, GSM and the mobile phone with voice etc. Therefore, as the number of services increase, so to does the complexity of managing these services from a user’s point of view. Furthermore, over time users will be surrounded by a communications environment that is far more heterogeneous than is currently the case.
What does the Web look like in this mobile or ubiquitous environment? This paper addresses this from the perspective of research in 'adaptive information systems'. Our working definition is that an adaptive information system delivers the right information services to the right user(s) at the right time on the right device in the right format and with the right level of intrusiveness. It follows that this requires explicit and machine-processable knowledge about the users, the device context and capabilities, the location in which these users and devices are situated, the content and the dynamic social context. Web solutions for metadata representation provide a mechanism for representing and processing this knowledge, in an interoperable manner.
We observe a trend from users querying the Web using a URL or Web search facility, where very little information about the user's context accompanies the query, through to these queries being augmented by additional context information in order to improve adaptation. To complete this spectrum, ultimately one can envisage a system whereby the context part of the query is sufficiently rich that the user part could be empty - appropriate information is provided automatically based on contextual information (i.e. the context is the query). This may sound like an idealised extreme but it has immediate practical consequences: devices can act proactively and autonomously in order to provide the user with the information they require, for example by pre-processing content or negotiating services prior to the user seeking information.
Some of this can be achieved through Composite Capabilities / Preferences Profile (CC/PP) and User Agent Profile (UAProf) standards. Within the research activities in context modelling there have been a number of other proposals under investigation, with richer definitions of context that include location, user profile and device characteristics but also infrastructure services, user activities, social context and time (see the recent workshop by Indulska and De Roure at UbiComp 2004 on Context Modelling, Management and Reasoning). In the next two sections we illustrate the need for richer context descriptions by looking at the problem from two perspectives: the user and the infrastructure.
User perspective – Context is inside us
Central to the successful adaptive information system is the need to collect and represent a wealth of information about the user such as current activities, skills, interests, personal preferences, privacy requirements and the relationships between people. However, representing a generic notion of context is a hugely complex and difficult task. In addition, there is some doubt to the usefulness of such a generic representation, should one be created, as it is likely that it would not afford the level of detail required by many applications.
Instead, we put forward the case that activity or task based modelling within a given domain would be more appropriate in many situations. Most activities carried out by people are task-oriented, hence the creation of more specific contextual representations is likely to be both easier and more productive.
For example, most working days can be conceptualised as a sequence of different tasks, such as a project meeting, document review, teleconference, patient consultation, or student supervision. Each of these tasks may have a variety of important properties. We propose, therefore, that contextually aware systems can be more usefully constructed in a fashion that is tailored to a specific environment, such as that of an academic research department, hospital ward, or lawyer's practice, through the application of domain specific ontologies and inference rules which may be applied to common infrastructure components.
As a practical example, to represent the activities of an individual we conceive of a schedule containing a reference to each of the tasks or events which that person will be undertaking. For a given domain, it is likely that a high level classification may be made of the typical tasks carried out. These tasks, or events, may be identified as complying with one or more templates, implying that common properties, features and/or resources are required, in a similar fashion to implementing an interface in an object oriented programming language. These high level templates may provide the basis for a hierarchy of more specialised instances of those tasks, defined specifically for the domain in which they are applicable. They may additionally include data such as the location, persons attending, topics of interest, importance and ‘interruptability’ relating to a particular instance of an event or class of events.
System perspective – Context is all around
The ability to automatically personalise service delivery within the mobile and ubiquitous environment is seen as a powerful approach for accommodating the expected growth in network, service and device heterogeneity. Future business opportunities and drivers are predicted to dictate that communications systems ultimately be capable of seamlessly supporting a heterogeneous mix of potentially fixed and wireless network literate devices and integrate them into a user's everyday activities according to their personalised requirements and context.
It is believed that that key challenges in this field are related to handling the dynamics associated with user roaming coupled with the drive to provide seamless, simple services that require little or no user management. Consequently, there are a number of key functional building blocks currently missing from today’s networking solutions that will be needed if this vision is to become realised.
Hence we intriduce the notion of the Personalised Communication Environment (PCE). The PCE is a logical image of a user’s communications capabilities that is dynamically updated as the user roams. The main information sources that are expected to change the PCE are: devices becoming available or unavailable, newly discovered services, changes in connectivity options and changes to the surrounding environment. Traditionally these information sources have remained logically separate and as a result have evolved independently, however, if the mechanisms exist to collate this cross layer information into a PCE for every user then this could potentially enable far more powerful context interpretation to become realised.
To date, although there has been effort on the topic of context management within the international research community, it has remained focused on the information and the devices used to display the information. It could be argued that, looking to the future, users will demand agnostic services and the flexibility to tailor their delivery and display device options dynamically to their own particular preferences. Consequently, it is predicted that there will need to be far more interaction with the surrounding environment, for example, if my door is closed then this means that I do not wish to be contacted. The PCE will provide information on the end user's connectivity, devices, services, and context and can be interrogated prior to establishing communication enabling context management to make decisions based on a user’s task.
In conclusion, this paper has positioned some of the broader issues relating to context management within future heterogeneous networking environments. It has established that a potential evolution for context management is to address the challenges associated with a user’s task in addition to the information and the display device, and to take input from the surrounding environment via a Personalised Communication Environment and use this to influence a user’s Web experience and increase the inherent automation.
Indulska and De Roure, Proceedings of the First International Workshop on Context Representation, Management and Reasoning, UbiComp 2004, Nottingham, UK. See http://pace.dstc.edu.au/ContextWorkshop2004Program.html