Report Homepage

Previous section

Next section


6 - 7
Emergent themes and issues; to personalise?

6.1 The role of portals
6.2 What do users want? What do they need?
6.3 Access and accessibility
6.4 Authentication, Authorisation and Accounting (AAA)
6.5 Profiles and preferences
6.6 Issues outside the scope of the personalisation study
7.1 The business case for personalisation in the UK academic community
7.2 Implementation issues and impediments

6. Emergent themes and issues

6.1 The role of portals

Some respondents refer to 'thin' portals. By this they mean portals whose aim is to deal with authentication issues as far as possible and to bring together diverse types of information from a variety of sources (e.g. publishers, student record systems, library catalogue, Internet resource catalogues, search engines etc) but not to control the subsequent interface to, or look and feel of, that information as the user follows the links. The converse is the 'fat' portal which brings all this information together but also imposes a uniform identity and interface upon it so that it appears to be consistent and within one local or institutional service.

There is disagreement on which of these paths should be followed. At the workshop in Edinburgh we learned from Katie Anstock from Talis that many of Talis' customers (mainly teaching and learning institutions and public libraries) are requiring a fat portal approach. However the institutions to whom we spoke which are currently implementing portal technology seem to have decided on a thin portal. From the University of Edinburgh, Anne Marie Scott's opinion was that fat portals will not currently work well due to lack of screen space and waste of well-designed information.

people have already put a lot of effort into custom building those interfaces which you will be throwing away Anne Marie Scott

Much doubt was expressed about whether a fat portal is achievable with current technologies and whether it could give students what they want. In technological terms, the impact of standards such as WSRP and JSR 168 will be to make it easier both for remote content providers to make their content available in a form that can be rendered by a variety of portals and to make it easier to integrate a number of these 'portlets' into a portal framework. As ready made, off the shelf and standards compliant portlets become readily available, the move towards 'fat' portals will gain momentum. However, this does not address the 'screen real estate' concerns expressed above.

Whether commercial suppliers of VLEs or library systems will build full portlet functionality into their systems remains to be seen. At the moment it seems that the most likely scenario is to apply the user login to draw personalised information, such as learning level or staff/student status, from elsewhere in the institution and then to "drill down" to appropriate pages within environments such as VLEs, library systems and calendar applications and to leave users in those environments from that point on. In reality, it is perhaps not useful to view the issue as either thin or fat. There will be certain classes of application which are intrinsically suited to being rendered as portlets and others that will require their own space. Rather than seeing the portal as the sole application controlling access to resources, the metaphor of portal as 'dashboard', providing a handy reference point to commonly used information, is probably more useful.

Looking into the future, Paul Browning mentioned that Chuck Severance from the Sakai project talks about rich clients replacing browsers. This is a highly disputed area, others point out that there have always been vested interests promoting proprietary alternatives to browsers, and moreover that the presentation layer must work with a variety of clients. Information will have to be served to devices or interfaces designed for disabled users, mobile phones and other small devices. Personal profiles will need to be adaptable to allow for multiple delivery mechanisms depending on receiving device.

6.2 What do users want? What do they need?

Much of our discussions and much of the feedback we received focussed on user wants and needs. Nearly all the data on this is anecdotal and there is a need for some systematic user requirement gathering (and collation of current and future user requirement gathering processes within UK academia).

... in my opinion, users do want and/or need personalization. For example, I am going through a Website redesign process right now at the University. Through ongoing focus group interviews, surveys, usability studies the idea of customization/personalization comes up again and again. While not a dominant theme in the discussions, it is given more than a passing nod. Furthermore, it is impossible to be all things to all people. The usual response to such a situation is to provide a lowest common denominator approach to products and services. Customization and personalization provide means to go beyond the lowest common denominator approach. Eric Lease Morgan

It was pointed out to us that end users in the University of Hull survey did not mention personalisation but they asked for things which could only be delivered using personalisation e.g. "I want my course notes on line".

Do users want, need or use personalisation? From IT support at a large US institution comes an answer:

I think they do [want personalisation], but I don't know if they know they do. We get many contradictory requests from end users about how they would like our on-line services to work. By allowing end users to personalize their own experience with a portal we could permit more variety and flexibility for the end users. Questionnaire respondent

... and from a senior researcher in Canada:

I am not so sure that they want personalization. ... That said, I would suspect that people need personalization and that they would want it if it were taken away from them. Email is an example of personalization (imagine having to search Google to see whether anyone has a message for you), and if email were taken away, people would complain. So I think that, even if there is no stated desire for it, people want and need it, but this depends a lot on how it is deployed. Stephen Downes

There seems to be much anecdotal evidence that, when given the opportunity, most users will not bother to create a profile or to customise their experience substantially. However it seems that what you might call "second tier customisation" is more popular. In other words, if a form of groupisation or adaptive personalisation based on data held elsewhere (APOD) is used to create a profile and thus a ready-to-use customised experience, then users are more prepared to go in to that profile and complete or alter the customisation themselves.

where you have information already then you can offer a personalised system which users will be keener to adapt (than to create one from scratch). Paul Miller

[academic users] don't want to *have to* personalise all the Web-based services they access. It should always be discretionary. But they also will respond positively to applications/tools/etc that make it very easy while guaranteeing privacy. Questionnaire recipient

As can be seen from the product review section of the literature review, the ability to define an initial starting point for users in terms of both format and content, then allow users to customise from that point on, is a common feature of both commercial and open source portals. Another important aspect of this, at least as far as institutions are concerned, is the ability to limit the extent to which users can make changes. In this way, it should be possible to ensure that all users see the 'important announcement' portlet which lets staff and student know about campus closures (10), etc.

Anne Ramsden from The Open University gave a presentation at our London workshop focussing on her work to include personalisation in the OU library portal. Building on work done by Eric Lease Morgan for MyLibrary at Notre Dame (see 4B2.3.1), the OU service offers subject alerts from diverse databases and TOC (table of contents) alerts from e.g. Ingenta, MyOpac, and MyBookbag. Her experience is that users do not use the customisation features to create profiles BUT where OU pre-populates with data then more users, particularly "power users", modify their profile. The pilot has been a success and even before its scheduled finish date the OU are now rolling it out to their student body; Anne feels the key feature is allowing users to gather important "stuff" in one place. An important factor which will ring bells in many finance departments is that it increases the use of resources to which the OU has already subscribed (she found substantial evidence that students in ignorance are paying for resources that institutions have already paid for just because of the multiplicity of entry points and students ignoring local information in favour of crude searching).

Jillian Griffiths has also studied the use of resources by students and found that they are largely dependent on the input (enthusiasm and skills) of the tutor. Students tend to give the same weight to any resource (much agreement around the table) so she feels there is a need for quality, refereed information and a great need for information seeking and evaluation skills to be taught to students via their departments in conjunction with the library. She and others are also worried that completely spoon feeding students with selected resources will prove further disincentive for them to learn information seeking and evaluative skills, and that solving this problem requires training not technical development. As Clifford Lynch points out:

At present many of our information retrieval systems conflate relevance with quality (as do many users), and socially-based systems are particularly prone to this. Some careful work is needed to disentangle these, and this may have important implications for personalization and recommender systems. Lynch op cit

Several respondents cautioned against reliance on systems which restrict a user's view to those resources which the system or its editors consider to be relevant.

How do I know that someone else's judgements will be any good? I want to know about things which are not relevant to my area. I have good information seeking skills. So leave me to it. Interview participant.

Others were keen to point out that users' experience with commercial services like (Amazon, Google, Expedia, eBay, Friends Reunited etc) sets their expectations for services offered by the UK academic community, expectations which will be hard to meet. One went so far as to suggest that if institutions wanted to provide personalised interfaces to appeal to users then users should be able to top up their mobile phones and buy concert tickets and travel tickets from the institutional portal. However we believe that there is room for portals and personalised services which aim to do specialised things well. The academic community has particular needs that can be met by excellent user-involved services, in our opinion it would be a mistake to believe that it is desirable or possible to create a user interface to everything, a one-stop portal.

When building Websites, practice user-centered design. Focus on the tasks users need/want to accomplish, and do not structure it as if it were an organizational chart. Employ principles of usability throughout the entire Website creation and maintenance process. Use customization/personalization to fill in the gaps of a Website's generic interface. Use it to supplement the user experience, especially the experience of frequent users. Eric Lease Morgan

I caution the community against personalizing everything -- it's a waste of time and resources. Focus on user benefits first and foremost - will the personalized function save the user time/money/frustration? Will the user actually take advantage of the personalized function? Will the user have a better experience with the service because it was personalized? If the answer is emphatically yes, you should personalize your service. Dana Dietz

Too much user control may create a new overhead for support staff, if, for example, a user switched off content which the institution is relying on them to make active use of. Too much flexibility in the user interface may make the interface too difficult for some users. Interview participant.

6.3 Access and accessibility

6.3.1 Access

Two particular issues involving access and accessibility need to be noted here. At our workshop held at an FE institution, an interesting bandwidth issue came up during demonstrations and presentations. One attendee remarked:

What is the point of all this stuff if the [FE or ACL] infrastructure is so poor that you cannot use it?

This itself begs the question "How much heavier on bandwidth is a personalised service than an un-personalised one? Do different technical options affect this?" It is certainly a question that designers and implementers should bear in mind. Where broadband is widely available and many students will have experienced it or even have it at home, this deficiency will look stark, a painfully slow connection at their academic institution will damage the credibility of the institution and the services used. Mark Williams noted that at least 60% of FE students do not have a "meaningful" (i.e. usable and used) email address, and that a UK study at an HE institution revealed that 40% of the students had never logged in to their .ac.uk email account (though clearly many of them were using email through an ISP account). Users must not only have tools fit for purpose, but the knowledge and motivation to use them.

6.3.2 Accessibility for all abilities

Accessibility options are important for user choice - for example, enabling large fonts/ high contrast but also for core materials to be printed - [this is a] popular and much requested option. ... personalisation discussion can also be used to examine bad practice - for example where course navigation is overly controlled by the designer, not allowing learners to skip ahead/around ... The way forward is to link the personalisation discussion back into the concept of learner-centred learning, offering choices to learners, and allowing them to take more responsibility for learning. Doug Gowan

Giving users the opportunity to customise a service to suit their preferences, for example by switching off graphics, or changing font-sizes or background colours, can improve the value of the service to the user.

By permitting multiple renderings of the same content, accessibility can be enhanced. This is a key advantage of building personalisation into a service. Perhaps every web site should have a standard service personalisation page. (11) Interview participant.

Customisation of this kind, along with any done by the user through adjustments to their browser, may result in the user's experience of the service being different to that which the provider expects or intends. For this reason a service should be designed so that it continues to function adequately even if a user customizes it in fairly extreme ways.

A particular issue is to ensure that users of assistive technology are able to understand any use a service makes of adaptive personalisation, and that if customisation is a feature of the service, that the customisation interface is itself usable when rendered using assistive technology. The potential use of techniques such as transcoding to render content for display on a variety of devices is discussed in 5.1.2.1.

Personalisation can reveal process problems in administration, and as we saw earlier in the comments on Managed Learning Environments it also demands more from designers, teachers and implementers:

... these personalized courses utilize a level of interactivity that requires course designers to plan and quality-test their courses to an even greater degree than non-personalized courses before they are released to students. Questionnaire respondent

6.4 Authentication, Authorisation and Accounting (AAA)

6.4.1 Software frameworks

It is not appropriate to discuss here the many technical and strategic issues under this heading, but it is important to note that AAA technology is an essential component (or a precondition) for implementing personalisation. In the UK, Athens (run by Eduserv see http://www.eduserv.org.uk/ ) has been a widely used initiative. Athens is a centralised access management system used throughout UK Higher and Further Education for the creation, storage and management of usernames and passwords. An Athens username may be used to access any of the 260 Athens protected resources (although the majority require your institution to have subscribed) - see http://www.athensams.net/dsp/resources.html for a full list. Eduserv Athens see themselves as "the guardians of information about individuals in the UK Higher and Educational community" and hold the current JISC contract for authentication and authorisation until 2006. Athens also offers some technology to allow developers to implement elements of single sign-on access to these resources. Many Athens protected resources use the Athens username as a key to personalisation within their own resources. In the NHS - another key Athens client - the Athens username is designed to be transported to other NHS organisations throughout the individual's career in the NHS. This means that 'favourite journals' and search histories might be retained as the individual moves around the NHS - a junior doctor typically works on six month contracts. Athens have recently introduced organisational and user attributes into the Athens system, starting with the attribute role=student. This has raised a number of worries in the Athens community about 'personal' information being divulged to outsiders. The Athens database allows Athens administrators to record information about their users such as department, number, email address, phone number; however none of this information has been released to outsiders. Athens are currently actively considering options for what they call " Managing Attribute Release Policies within the Athens Identity Management System". Lyn Norris from Athens comments:

The introduction of attributes has brought this concept a little closer; and we certainly believe that an Attribute Release Policy is necessary to allow individuals to control the release of this information.

For a variety of reasons JISC are now encouraging institutions and developers to look at using Shibboleth as a possible future AAA infrastructure for local and national implementations. More information can be found at:

http://shibboleth.internet2.edu/docs/internet2-mace-shibboleth-introduction-200404.pdf

http://www.jnug.ac.uk/presentations/authentication-and-authorisation-0402.pdf

http://www.jisc.ac.uk/uploaded_documents/NotesoftheTownMeeting.doc

Comments on how Shibboleth addresses personalization (note spelling) issues can be found in:

http://shibboleth.internet2.edu/docs/draft-wasley-shibboleth-portals-02.pdf

6.4.2 Single login to multiple resources

It is worth noting that there is a long way to go in this area, even to achieve what might be considered a simple step of allowing users to access resources with a single login. Note the comments of a digital library expert from a leading US institution:

One sign-on and password for all would be ideal ... [the] more current alert services we offer, the more important personalization is ... Personally, I don't think most users know what it is and what is possible ... It would be great if on logon, we can know which department you are from, what subject you are majoring in and your special interests. Then we can offer special services pertaining to that information ... until we have a good authorization and authentication system, it is not worth the effort to implement.

Another, this time from Australia, in a similar vein:

I am yet to see any personalisation that was highly compelling. One of the problems is the lack of Single Sign On between applications which would otherwise make it seamless to have personalisation data passed from one application to another (although the need to address privacy and security are big hurdles to this). Professor James Dalziel

It should be noted here that Athens has allowed an effective single sign to multiple services in several situations - the RDN Subject Portal Project (SPP) is mentioned in section 7.2.1 as an example.

6.5 Profiles and preferences

In order to achieve personalisation, user profiles and preferences must be stored or collected in some form. This issue came up in virtually all the discussions we had and responses we received. It spawns issues of privacy, access, authentication and others that we deal with below.

Any user will have multiple preferences and/or profiles in multiple areas of their Internet usage - work/home/study/leisure etc. One user may be a hospital employee and a student at another institution and a teacher at an evening class elsewhere - who would they trust to hold their information? (12) More commonly a user may be happy to entrust an academic service with preferences for work-related research and teaching information but not about bank details, leisure pursuits, union activity etc.

Whilst AAA is an essential ingredient for sharing appropriate access to resources, the issue of trust goes much wider.

Customization/personalization is a method of building relationships between individual users and institutions providing goods and services, whether they be students and libraries or parents and computer manufacturers. By building trusting relationships everybody benefits. Responsible application of customization and personalization techniques enables the creation of trusting relationships in a networked environment. Doctors help build this trust through doctor/patient relationships. Lawyers facilitate this relationship through lawyer/client privileges. I believe the same thing ought to be fostered between librarians and patrons. In fact, to a great degree it already exists but many patrons do not know it.
Eric Lease Morgan

Academic services will need to take a somewhat different attitude towards their users than commercial organisations. It may be necessary to collect and reuse information about users but there should be clear and open acknowledgement of that so that users understand what is going on. An endless small print agreement to which the user is required to respond "OK" on screen before proceeding is not good enough.

... one should be aware that these kinds of personal information could be valuable in the commercial sector and appropriate means of dealing with that have to explicit ... be frank and honest and tell them exactly what you are doing. Some users like to have "the system" collect information and adopt automatically others like to have full control over what the systems knows about you. Provide options. Be very explicit about user security issues. Anders Ardo

Users should also be encouraged to recognise the issues. While allowing the choice not to be "personalised" the interface must clearly explain the advantages that collecting information may confer upon the end user experience.

The issue in the United States is one of privacy. We have a long history of protecting a user's privacy about both their personal data and about what services or resources they use in the library. To achieve personalization, users will lose that privacy. I don't know if they are aware of that, or how they would react if they did know. But if we start tracking users' behaviors as they use our systems, or even start storing information that they provide us about their use of the library, there is an inherent risk of that data being accessed by others. Questionnaire respondent

The recent growth of systems based on the concept of 'federated identity' (see, for example, the description of the Liberty Alliance in section 5.1.3) raises several issues in this area to which we do not have answers but which are important and merit investigation:

  1. Privacy - what is legally allowed? What legal entity has rights to information about your use of services? Does the data protection legislation specify detailed rights and obligation in this area?
  2. Is your sign up agreement to Amazon or Microsoft (how many of us read them?) more sweeping than anything you sign up with your institution?
  3. Who would you trust to hold your information? Would you want it held in one place with permissions for release of different parts of it to different people? Or would you prefer different institutions to hold different parts of it (with the inefficiencies and uncertainties that may result)? Is there a possible role for institutions which we already trust with personal or financial data to step into this arena?
  4. With particular reference to the issue of trust, how much can be accomplished by consortia or organisations which are not legal entities (JISC, RDN etc)?
  5. How much of your profile or preferences is transferable? Is it the case that most of your preferences will be service-specific and only a small core subset will need to be shared between organisations and services?
  6. Is there a role for a small number of independently verified profile attributes which are freely available but which the user cannot alter (though they can be revealed or hidden by the user)? Level of academic accomplishment might be one.
  7. Could a user preference service be distributed across institutions, in the same way that an LDAP directory service works?
  8. Might users want a bank of profiles so that they can choose which profile to import into (say) an institutional portal or a national portal?
  9. How should information about users be represented in order to facilitate exchange between applications (13)?
  10. How would profiles be updated? Would they suffer from the same fate as "under construction" Web sites, abandoned for years on end?
  11. How would the lack of common terminology and common understanding of vocabularies affect the portability of profiles?

6.6 Issues outside the scope of the personalisation study

Two recommendations arose during the study which are outside the scope of our remit but worth identifying nevertheless:

  1. We are aware that tentative approaches have been made to work with Google at a national and European level in the past with some very limited success, but we believe it is important to work hard to make sure that the leading service in this area is working with some of the best creative minds in the UK academic community to pilot innovative and worthwhile services for the academic community. It was suggested that JISC ensure an approach is made from a sufficiently senior level in government to suggest and effect some collaborative work with Google.
  2. Different publishers are not consistent in the way they make their resources available, nor are the ways individual publishers make their information available stable over time. JISC have approached a number of journal publishers in an attempt to address this issue, it was suggested that this effort is expanded.

7. To personalise? - incentives and impediments

7.1 The business case for personalisation in the UK academic community

In the short term, there are circumstances where there is a clear business case for personalisation:

In the medium term personalisation may bring a transformation of the landscape as Lynch suggests:

Imagine that users could maintain a series of local (personal) databases that defined their trust systems (we have the very primitive beginnings of this in the identity trust codified in systems like PGP or certificate collections in Web browsers); their reputation and rating systems; their preferences of various kinds; interest profiles; and their opinions about objects of various kinds (Web sites, books, sound recordings, films, people, etc). Imagine that there were standards for transferring this kind of information to systems that one wanted to interact with, and also standards for updating the locally held, personal databases to reflect actions taken in remote systems (such as purchasing a book, or visiting a Web site); and, of course, tools to edit the personal databases and to establish rules about their update and dissemination. This would potentially permit instant personalization of any system that one wanted to interact with; it would allow users to visit new sites and immediately obtain potentially useful recommendations from them... Lynch op.cit.

However, this is a large task and one that should be taken on internationally - Lynch goes on to note that:

Moving from this vision to implementation is a very complex research problem, with lots of difficult details, but one that I think merits investigation. I would just note in passing that a good deal of infrastructure and numerous standards are needed to make this work

For the UK academic community some well defined elements of this problem could very well be investigated, but attempts to provide a total solution should be avoided until many more of those elements are in place and internationally agreed.

Our respondents noted that while, within an institution, it should be possible to measure the benefits of personalisation in clearly defined circumstances, it may be very much harder to do that in the case of national or "free" information services such as the RDN.

How do you measure it? - it all looks like cost, without a proven return or even knowing how to prove it. We don't have a record of the users' experience and we have no accounts of any benefits of personalisation. Workshop participant

Several respondents thought that the institutions should be the focus of personalisation with organisations such as the RDN supplying re-usable resources rather than promoting multiple competing individualised interfaces. Others were adamant that end users preferred subject focussed services such as the RDN Hubs and that the push toward institutional portals was institution-led and not supported by end users. It seems inevitable and desirable that there will be multiple interfaces to high quality information and that there will be some degree of competition between them. Where competition within the JISC community leads to restrictive practices, impedes integration and hinders users, it should regulated (see 7.2.1 below).

7.2 Implementation issues and impediments

Eric Lease Morgan notes that implementers often confuse indexers/search engines with relational database applications. It is necessary to appreciate where these technologies are complementary and when to use one over the other. He also identifies use of proprietary "features" as an impediment and exhorts implementers to try to adhere to open standards

The underlying use of XML is an excellent example. By employing open standards implementers can be more flexible and adaptable to change. Proprietary features are short-term solutions locking you into specific vendors. In the long run such things are limiting not empowering.

Adaptive personalisation requires the user's actions to be recorded and stored and, in the case of APOD, to be associated with institutional records. Morgan notes that it is important to be open and to ask permission for this recording from the users.

Implementers should be sure to let people know that this watching/recording is necessary for personalization. "If you tell me a bit about yourself, then I may be able to help you better, but this information will be recorded as opposed to simply being in someone's head." Librarians take this sort of thing very seriously ... and this is sometimes an impediment to providing personalization services in libraries.

Dalziel also points out that it is not only the cost and difficulty of learners taking profiles (see 6.5) with them on their journeys through the Web which is an impediment to personalisation within education. Probably even greater is the cost of assembling the metadata for each resource necessary to enable meaningful personalisation.

It was noted that using Amazon or eBay techniques in academia would need a substantial investment and a lot of users. Past experience suggests that motivating users to participate might be problematic. It appears to be much harder to persuade people to review learning materials than books and music. EDINA reported having great difficulty getting reviews of educational programmes when they tried modelling the process on the IMDb (The Internet Movie Database) - "people will review movies but not educational resources". For their work with SearchLT Engineering, EEVL found that they had to pay people to review learning resources.

Integration between an institutional portal and a third party learning environment is also a challenge, at two levels. Firstly, it may not even be possible to "skin" the learning environment in the colours of the institution. Secondly, a third party's learning environment will not have the same user interface as the institution's web site, so that even if users can have personalised access to "their" area of the learning environment, they will inevitably have to contend with a different user interface. (See 6.1, 5.2.1.3 and 4.1.5 for more on learning environments).

7.2.1 Efficient use and re-use of JISC-funded resources in personalised services

In the context of our discussions, the position of JISC services, and the RDN in particular, was regularly raised by participants. The SPP (Subject Portals Project), demonstrated in Bristol and discussed in Edinburgh and London, was cited as an excellent step forward. The remit for the Subject Portal Demonstrator was to produce a tool for RDN subject gateways to offer a customisable portal service to users. An example of this customisation is the cross-searching feature which can be customised by the portal administrator; the look and feel of the interface and the information sources and other searching options can all be customised by the users. If users login using their Athens password they can see resources which would otherwise not be available to them, showing a single authentication in action across a variety of services.

The Hubs and subject gateways have subject based data which they can match against individually stated preferences. SOSIG, for example, pushes information weekly about new resources to 2,500 people via email, based on user profiles. It also uses 15,000 profiles for "matching" people with people (Likeminds), and people with conferences, workshops etc (Grapevine). Data protection issues mean this knowledge is not shared beyond SOSIG. The broader issue of who owns the service and the personal data associated with it is unclear. The issue of entities without legal status is referred to in section 6.5. Currently different hubs are providing similar services including the same news feeds. At the minimum one would want to share anonymised data about users over the whole of the RDN.

As an academic member of staff, localisation for me means picking resources from an already classified list (e.g. EEVL) NOT classifying and describing them myself. Part of institutional portals or other portal services is relying on tagged information resource descriptions etc. There will be a need for this activity which happens at SPP and Subject Hubs, even if it is invisible. Workshop participant

We want resources [from the JISC Information Environment] to be made available to users ... with "user meaningful" rather than "dumb" integration: this has always been part of the development plan for our portal. But actually doing this is not straightforward. A further complication is the difference between the data which is under our own control, and the data which is under the control of a third party. Interview participant.

The data held by elements of the RDN is potentially highly useful but currently used only in a limited way. This applies not only to personal data but to the resource descriptions themselves. They are a valuable collection of data, created by a human-effort-intensive process which is currently impossible to replicate or automate using technology. Yet even within the UK academic community their use is very far from widespread. We found that those creating institutional portals and those intermediaries serving end-users recognised the value of JISC collections but were frustrated that more JISC effort was not put into integrating them into a variety of other services. One inhibitor that was mentioned several times, by those outside of the RDN as well as those within it was the way that success is measured for such services - the performance indicators. It seems generally agreed that it is vital for the benefits of national investments to be realised (and re-used) on a local level, yet the success of RDN services seems to be measured on statistics for direct end user usage of their (national) services. This seems to be a problem for JISC services in general with "interface blight" being mentioned by users and intermediaries as a symptom. Every project however small, however infra-structural or background their role, feels that it must have its own interface and must attract the maximum number of users to its site. To create simplicity for users we should be reducing the multiplicity of interfaces/services not encouraging competition between services to drive people to "our" interface. If everyone agrees that the RDN resource descriptions should be used as widely as possible then it is counter productive to judge success by measuring Web accesses to individual interfaces - a better measure of success would be if these resources are indeed being used by third parties, institutional portals and other services, nationally and internationally.

7.2.2 Levels and types of learning -suitability of materials

Many people we spoke to were concerned about the persistent problem of levels of knowledge - how do we mark a learner's levels of knowledge in a particular subject and how do we mark a resource's suitability for a particular level of learning. Even within institutions this problem seems difficult, and moving between institutions in further education, higher education and adult and community learning, the problem seems intractable. Several participants thought that personalisation could only have a limited effect until the "level problem" was addressed.

Users want courses and materials that are relevant to them; solve their problems; are at a level that suits them; use examples they can grasp; are concrete and not overly abstract. Doug Gowan

Some e-learning commentators see personalisation as the ultimate goal for e-learning - so every learner can have the perfect individualised pathway through learning resources according to their unique needs. While this is a very attractive idea, it ignores the fact that much of learning is collaborative in nature, and that collaboration presents significant challenges for personalisation as currently conceived. In my view, collaboration learning activity sequences are more valuable than current instances of single learner personalisation, although ideally we need both. The problems of personalisation in a collaborative context are quite complex and not yet well understood. Professor James Dalziel

7.2.3 Vocabularies, Terminologies , Classification and Thesauri

Controlled shared subject vocabularies, used by all actors/services are seen as a key to enabling customisation. This area appears as "Terminology services" in the JISC IE architecture diagram. However, making this sharing happen in a meaningful way is a substantial problem which has had a lot of international money spent on it with no solution in sight. Work done at HILT http://hilt.cdlr.strath.ac.uk/ has recently cast light on some of the problems here. Even within individual RDN Hubs, subject vocabularies, classification schemes and thesauri vary between subject areas and users stoutly resist any homogenising. This is not merely stubbornness or "not-invented-here" syndrome. Users and particularly intermediaries such as subject librarians have very good reasons for wanting to use schemes which have been designed for their subject communities and have had a long history of development and adaptation.

For a user profile or model to be matched with appropriate resources requires that these resources are described in some way which allows them to be matched against the profile. Collection Level Descriptions, mentioned in 5.1.3, are one way of achieving this in the bibliographic domain which may be extensible elsewhere, but they generally rely on metadata being created about each class of resource. Manual creation of metadata about all the resources (e.g. Web pages) which could be included in a personal information landscape is clearly an inhibitor to uptake. Auto-classification based on content (see, for example, the description of the Oracle Portal in 5.2.1.1.1) is still in its infancy but, if proven in real world contexts, could help overcome this hurdle but this is unlikely to gain widespread adoption in the very near future.

Adaptive personalisation may require the sharing between services of records of user behaviour. This sharing itself is a highly problematic area - the data sets will be huge, the data protection issues no less significant and the uncertainty over ownership of the data must be resolved. Amazon or Yahoo can achieve results in this area because they control the whole user experience and have a more sweeping user agreement (see above) than academic services.

FOOTNOTES

  1. At the University of Edinburgh, the highest ever voter turnout at student elections has been explained by some as a result of an advert on the portal (which was seen by all students accessing the online learning environment).
  2. For an example of this see the "Change Layout" option on The Sheffield College's web site - http://www.sheffcol.ac.uk/.
  3. It may be that ePortfolios <http://www.eportfolios.ac.uk/>could form part of the machinery for transferring profiles to achieve personalisation in different contexts - users could share preferences between ePortfolios and other services in both directions, but there are many claims for the future capabilities of ePortfolios and they may be drowned by the weight of expectations.
  4. IMS LIP http://www.imsglobal.org/profiles/index.cfm and EduPerson http://www.educause.edu/eduperson/ were both mentioned

Report Homepage

Previous section

Next section