8. Conclusions

1. User expectations of services available to the academic community will change as they increasingly experience the advantages (and sometimes the disadvantages) of personalisation in commercial services and in the many applications of Internet banking and finance. Personalisation has succeeded in commerce where it fulfils a core function for the business and where the task is relatively simple and well defined. Many businesses attempting a broad and poorly defined application of personalisation have retrenched, abandoned it or even failed altogether. The businesses that succeed with personalisation are, typically, either "pure" web businesses, such as eBay, expedia, Amazon (notwithstanding the latter's subsequent massive investment in physical order-fulfilment infrastructure), or businesses which already have a very significant investment in ICT like banks and airlines. The UK academic community is not like this, having a far higher proportion of spending on people, a far lower proportion of spending on ICT, and far fewer opportunities to get scale economies. This suggests that the academic community is unlikely in the short term to succeed with broad application of personalisation.
2. Narrow applications of personalisation are far more likely to succeed, particularly where a clear objective has been identified, a business case can be made and control has been or will be established over the relevant, accurate data.
3. Personalisation of institutional information services can offer significant efficiency gains, for staff if they can quickly access institutional information pertinent to their roles, and for students if they can access information for which they would otherwise have to ask in person or in writing.
4. Dalziel points out (quoted above) that generating the individual personal metadata needed to enable learning to be adaptively personalised will be costly and complex; and that manually generating meta data for learning resources (level, content, quality etc) will be prohibitively expensive except perhaps in the very highest volume learning programmes. Automating this metadata generation is possible, but the technology is still at a research stage.
5. Learning environments may use personalisation as one element in an interface presenting a rich mixture of resources to their users. One can imagine something rather like Google's sponsored links, where the sponsored resources are replaced by resources selected by the institution or by subject experts at the RDN. The search results might also feature the top 5 resources in this area as used by others on your course or at your institution. One might draw the analogy of a flexible restaurant menu, where people on course X always get the lecturer chosen items (the dish of the day) but are also shown the results of their own search or profile (the à la carte items).
6. Educational Web sites and services should ensure that access to resources is available to those who cannot use conventional interfaces. Blind users might require audio interfaces or output in Braille, there are many other examples. It is also likely that users will be accessing academic resources with a wider range of devices including small wireless devices, phones, organisers etc. Access for all should be designed by default for all interfaces, but elements of personalisation may be useful in this context.
7. Very little work has been done on user requirements gathering concerning personalisation in general and this is as true in the UK academic community as in other sectors. Much of the evidence is anecdotal, with features being decided upon by developers and champions not as a result of user testing. If, as Nielsen claims, you only need to test on five users to get useful results, then it seems as if some user testing in this area would be cheap and rewarding.
8. Personalising information supply can amount to spoon feeding. Learners who are spoon fed do not develop the required information skills to be able to find information when they are "away from the portal", nor do they develop the critical judgement. The RDN VTS is widely praised for guiding users to acquire essential skills without spoon feeding.
9. Systems which allow "human" personalisation, such as a VLE which allows tutors to configure the system to provide learners with a personalised learning experience, are generally popular with motivated staff and students. However such complex design requires more careful and complex testing than do simpler designs. We should not underestimate the extra demands this puts on the staff involved in customisation and design, which may include role playing a whole suite of different types of students.
10. The growing adoption of standards such WSRP and JSR 168 by suppliers of components such as portal frameworks, will increase the ability to render services as portlets inside a 'fat' portal. However, the fact that the technology enables this does not mean that it will be appropriate for all service types. In particular, it is those services which are designed to have a machine readable interface which will fit best within this framework rather than those where an interface is shoehorned into the framework using techniques such as screen-scraping.
11. Within a portal framework, the benefits of allowing users to customise content and appearance must be weighed against the costs. It makes sense for portal providers to impose some limitations on what can be deleted (or added) and to ensure that certain screen elements are always positioned for maximum visibility. In terms of encouraging users to take advantage of these features, current best practice appears be to offer an initial degree of APOD (based, for example, on course, subject or even a simple 'staff'/ 'student' dichotomy) then allowing users to customise within a restricted set of already targeted resources. There is no compelling evidence to date for the benefits of APUA within the portal contexts, but as open source toolkits offering collaborative filtering and similar functions gain maturity, then the costs of implementing these approaches will fall.
12. The JISC IE provides a logical architecture which should be viewed as a long term goal rather than something which can be achieved immediately. The availability of shared infrastructure services such as user preference services, institutional preference services and service registries would vastly enhance the ability of both JISC services and institutions to offer personalised services. However, such services will take time to be become mature and even then it will take time for real world interoperability issues to be resolved. For example, although, in theory, it has been possible to provide an LDAP directory service to provide common user data to a range of institutional services for many years, and many products claim to be 'LDAP-enabled', in reality even such a simple 'shared service' has proved difficult to implement.
13. Within the JISC IE there is a potential need for personalisation to flow backwards into the content sources as well as across different services within the presentation layer. Such sharing may require a trusted intermediary to strip out personal information and possibly restore it as information comes back from the content provider to the presentation service or to the end user.
14. There are many issues raised by personalisation that require international collaboration and standards to resolve. Naming people as creators and/or authors of resources is one such element, possibly creating a unique ID (persistent email address?), which would be useful across a range of services. Unique and persistent resource identifiers are another. User profiles and preferences will also be very relevant here, with candidate standards such as EduPerson and IMS LIP emerging, but far from established. Where standards are yet to be established we do not recommend establishing national services based on ad hoc standards which may be superseded by international work. As Lynch points out (quoted above) there are many and various pieces of work to be done in this area and we believe it would be useful to fund smaller scale pilot projects with discrete, well defined tasks at this stage.
15. All projects and services are likely to create an interface (presentation layer) in order to make themselves comprehensible and visible to funders and intermediaries if not to end users. However this interface creation has a tendency to become an end in itself. It is important to examine the real or potential function of the project or service. With this in mind, we must also always be aware that complex applications will increasingly be made invisible to the end user by processes involving machines talking to machines. All services should bear this in mind and strive to make their resources available as widely as possible - this does not mean abandoning their valuable IPR - it means expanding the potential market place for products which may use that IPR.
16. Personalisation can either be treated as an extra service - an "add-on" - or as a means of bringing about major innovation around which services should be re-engineered. The latter approach is more likely to succeed and result in major improvements in quality of service than the former.