5.1 Context
Perhaps the most difficult challenge is that of context-aware computing. The Smart
Internet should recognize and store information about the context of its users, and
should act on this data by adapting its services automatically as required. Current
approaches to context information modeling and retrieval are still immature and have
tended to fail even in simple applications [18]. In general, health care requires extremely
complex, temporal context information with high assurance of accuracy, as
illustrated in the scenario presented in Chapter 1 of this book. Health care is also an
inherently “risky” domain. Given the progress made in the two decades after publication
of Weiser’s ubicomp vision, it is doubtful that the technical challenges of successfully
modeling, retrieving and reliably acting on complex health-care related
context information will be overcome within at least a decade of this writing.
Of course, this does not mean that particular context-aware functions may not be
used to build a smarter Health Internet in the short term. Existing Intranet context
models and standards that have been developed in the health domain may realistically
be adapted to broad-scale Internet use. One example would be the standards of HL7’s
Clinical Context Object Workgroup (CCOW), which deal with the synchronization of
different health information systems in respect to particular subjects (i.e., patients)
[38]. In this approach, the patient identity is the single point of focus (context information)
shared by all the applications used by the provider.
The challenges moving such simple contextual function from Intranet to Internet
scale are mainly political, rather than technical. As with other forms of Internet-scale
identity management (e.g., single sign on of user identities), controversies may arise
with respect to control and governance of a shared subject identity context manager.
More significant advances in context-based computing for Internet-based health
care applications may be possible by following a “learning” rather than a “modeling”
approach to context enactment. Several promising theories have been developed in
40 J.H. Weber-Jahnke and J. Williams
the field of Computational Intelligence, including Case based reasoning (CBR), a
technique that has found successful applications in health care (e.g., clinical decision
support) as well as other business domains (e.g., recommender systems) [39]. From a
technical point of view, CBR approaches may well be in reach; for example, be companies
like Google and Microsoft may be capable of providing smart search and
dashboard Internet functions based on context data learned from patient cases in
Google Health or Microsoft Health Vault, respectively. However, the political, legal
and sociological issues arising with entrusting profit-driven organizations with highly
confidential, personal data and relying on their services may be insurmountable in the
short run. The development of open standards to store and enact context-information
in a decentralized fashion at organizations trusted by users may overcome some
of these socio-legal-political hurdles. Initiatives for such standards exist in related
areas, e.g., the Liberty Alliance for decentralized identity and trust management
(www.projectliberty.org).
5.2 Dynamic Social Binding
Social networking has provided a foundation from which to support collaboration.
However, current approaches to social networking do not go far enough. Health care
episodes often involve teams of practitioners that coalesce around a patient for a limited
period of time. At present, social networking tools do not support dynamic social
binding – the capability to select other users dynamically to share interactions at a
given level of detail. Current social networking applications allow the formation of
networks through fairly cumbersome means. They are also not geared towards tracking
temporal information about the network, such as the duration for which two users
were joined. In order to provide dynamic social binding, new tools and techniques are
required for both the data structures, and the user interfaces.
In addition, a key requirement of Smart Internet systems concerns the ability to
maintain the user’s tasks, goals and matters of concern as persistent states, in order to
track progress. Social networks provide a type of persistent state, in the form of a
user’s network and profile details. Websites like PatientsLikeMe encourage users to
update various metrics daily, which provides a persistent data store of personal health
information. However, the services available to PatientsLikeMe users are still largely
invoked by the user, who bears the burden of remembering her various tasks and
sub-tasks. Consequently, such sites are frequented primarily by patients with major
existing health conditions (chronic, acute or palliative), while healthy consumers
lack motivation to keep reporting their data. Thus, these sites are less effective for
preventative care.