Jennifer E. Jobst
LIS 385T
Due October 22,
2002
Submitted October
22, 2002
In the world of e-commerce, personalization and customization is becoming the business model of choice. Many of the most successful e-business web sites use personalization and customization with the goal of creating a strong working relationship with their clients and end-users (Riecken, 2000 August). However, issues such as privacy, the difficulty of providing an appropriately personalized page, and the fact that many users prefer to remain anonymous, are problems that must be addressed for personalization to be successful. From the standpoint of an information architect, these issues and others all play a role in the design of information for a personalized web site or application.
There are two terms commonly used when discussing an interface designed for a specific user or user demographic: customization and personalization. Although some might use these terms interchangeably, they refer to two very different aspects of design. Customization, as defined by Jacob Nielsen (1998), is used to describe interface attributes that are user controlled; for example, background color or pictures, font size, icon placement, etc. Users often customize their interfaces because of personal preference they may want a specific background picture on their desktop, and they may prefer their icons to be on the right-hand side of the desktop instead of the left-hand side. Users might also choose to customize an interface to compensate for a disability. For example, a user with low vision might choose to increase the screen resolution so that icons and fonts will appear larger, and might turn on audio alerts so that error and warning messages are read aloud.
Personalization, on the other hand, Nielsen (1998) defines as interface attributes that are computer driven, often by an algorithm or user model based on the userıs selections, habits, demographic information, etc. The two concepts are closely related and the dividing line between the two can be vague. For example, suppose a user selects an option in his or her Preferences labeled, ³I have trouble seeing the screen.² Selecting this option then increases the font and icon size, changes the display to a high-contrast color scheme, and reads all error messages and alerts aloud. Since the user did not directly choose these specific settings, the way in which the user interface was customized could be considered personalization.
There are a number of different reasons to personalize an interface (IBM High-Volume Web site team, 2000). A company might want to cater to different levels of user expertise by showing different users different information. For example, suppose a software engineer visits a vendor site looking for the specifications of a certain product. The engineer wants detailed technical information, whereas a manager from the same company might want more of an overview of the product, along with a cost analysis so that manager can decide whether or not the product can be purchased with this yearıs budget.
Another reason to personalize might be to provide different interfaces for different levels of authorization. For example, the web master of a secure site will almost certainly see different options than would a regular user, since the web master must have the ability to add and remove users, track their usage, control passwords, and other administrative tasks. A regular user does not need these options and therefore should not see them.
Often it is in a businessıs best interests to cater to the userıs habits, which is another reason to personalize. For example, amazon.com tracks each clientıs buying habits and from the information collected recommends products that might be of interest to that client. The personalized page that amazon.com serves up for a client includes the clientıs name at the top of the page, and just under the clientıs name is a list labeled ³Your Recommendations,² which shows a list of products that amazon.com recommends based on the clientıs past purchases.
Finally, a web site or application might be personalized to allow for multiple languages or nationalities. In the case of large, multinational corporationıs web sites, the server might look at the default language setting on the clientıs browser and serve up the web site in the same language.
There are different reasons that a web site might be personalized, and there are also different aspects that can be personalized. The two most common aspects to personalize are structure and content (Rossi, Schwabe, Guimaraes, 2001 May). Personalizing the structure of an application or web site means altering the location of available links as well as the location of the content (text and images). This aspect of personalization involves placing the links and content that are determined to be the most relevant to the user in a prominent place. Thus, different users may see different layouts of the interface. Amazon.com makes use of structural personalization, as the ³Book Recommendations² link for a client is placed near the top middle of the page, and other items on the page are places so that the clientıs eye is drawn to the list of recommendations.
Content personalization, on the other hand, changes the content of a web site depending on the user (Rossi, Schwabe, Guimaraes, 2001 May). Amazon.com also makes use of this aspect of personalization, since Amazon builds each clientıs ³Book Recommendations² list based on that clientıs previous purchases.
One aspect of personalization that is becoming more important in is that of delivery. As different delivery devices become available, such as PDAs, cell phones, in-car navigation systems, and others, content and structure will have to be personalized for these devices (Billsus, Brunk, Evans, Gladish, Pazzani, 2002). Currently, many web servers check to see which browser is requesting information (for example, Netscape, Internet Explorer, Mozilla, Opera, etc.), and possibly the resolution of the delivery device as well, before serving up an appropriate page (or requesting that the user upgrade or use a different browser). Of course, if a cell phone or PDA is the delivery device, graphics, advertisements, and other bandwidth or graphic intensive information may be too much to download or display, thus the server must tailor its data for the device. This aspect of personalization, while currently in its infancy, will become more of an issue as more wireless devices are used to communicate and transact business.
Although businesses are developing more and more personalized and customized web sites for a variety of reasons, there are a number of issues that must be considered when developing such a site. Information architects must understand each of these problems and how they affect usersı web usage rational in order to effectively apply new and existing frameworks in an effort to minimize these issues.
The first issue to consider is that users often do not want to provide personal information that could help to personalize the site to best suit their needs. Instead, many users prefer to remain anonymous (GVUıs WWW Surveying Team, 1998). Studies of the my.yahoo.com web site, which is one of the most prominent, best-rated sites for personalization and customization, show that the majority of users take exactly what is given to them and only use the default my.yahoo.com page (Manber, Patel, Robison, 2000 August). Only the most experienced power users are likely to take the time to customize or personalize, and only then if they have to provide a minimum of personal information. Based on these and other findings, an information architect cannot assume that users will spend the time to provide the information necessary to personalize or customize a site. To counteract this problem, Manber, Patel, and Robison (2000 August) suggest that the default page must be extremely well designed, under the assumption that most users will only use that page. Another way to bypass the lack of information provided by the user is to have intelligent agents collect information about usage habits and develop a user model from that information (Bogonikolos, Makris, Tsakalidis, Vassiliadis, n.d.). The user model can then be mapped to appropriate content and context to provide personalized structure and/or content (Instone, 2000).
One of the key reasons users do not want to personalize is that they are concerned about privacy. To provide the most appropriate personalization for each user, information about that user is required. This means that users must give up some aspects of their privacy if they want personalization. Many users are simply not willing to give up their personal information, despite the potential benefits. In the 10th WWW User Survey (GVUıs WWW Surveying Team, 1998)
In other studies in the e-commerce world (Kobsa, 2002), respondents:
reported being concerned about threats to their privacy when using the Internet; being concerned about divulging personal information, online; and being concerned about being tracked online. . The [sic] reported leaving Web sites that required registration information; having entered false registration information; and haying [sic] refrained from shopping online due to privacy concerns, or having bought less. . They want Internet sites to ask for permission to use personal data and are willing to give out personal data for getting something valuable in return.
It is clear from these studies and others that users are concerned for their privacy, that many want anonymity online, and that users want to control what information they give out and when. Thus, the information architect must realize that if users do not feel that the information requested, whether for personalization or for other means, is absolutely necessary, the user may refuse to give it. The potential lack of user input raises the question as to whether or not personal information such as usage or shopping habits should be tracked by intelligent agents, as discussed in the previous section, so that the user does not have to input any information (Instone, 2000 December). Of course, the information gathered by such agents is considered by many to be private, and some users may take exception to this information being collected without their knowledge, consent, or control. Indeed, these data-gathering practices may be violating some countriesı privacy laws (Kobsa, 2002 May).
For the information architect who must develop a personalized site to best suit clientsı needs, the issues regarding privacy are far-reaching. First, it is important that the local and national laws regarding online information privacy are understood and followed. This means that any information collection mechanisms must adhere to these laws, and also that once information is collected it is handled in such a way so as not to violate any laws. In addition, it may be necessary to ask the consent of the user to collect certain information, and often it is required to notify the user that information is being collected and how that information will be used. Kobsa (2002) provides a list of guidelines that he suggests will help comply with many privacy laws as well as help address most usersı privacy concerns. He also suggests that, instead of the current methods of collecting certain user information, information architects refer to the new alternative proposed by Schreck (2001), which allows users to maintain ³pseudonymity² but also allows full personalization.
Assuming that privacy issues can be addressed in an appropriate manner, there is still the issue of providing appropriate personalization features. This can be an especially daunting task for the information architect if users are unwilling to provide the information necessary to make a web site appropriately personalized for them. If information cannot be obtained explicitly from users (that is, by directly soliciting information from them), then intelligent agents and user modeling tools like GroupLens, LikeMinds, Personalization Server, and Learn Sesame (Schreck, 2001) may be used instead to collect information implicitly (that is, by tracking aspects of user behavior), and that information can then be used to represent users (Instone, 2000 December).
The problem with any user model is that, even if it is relatively accurate initially, user needs change over time, and a personalization system must take this into account to accommodate ever-changing user needs. As suggested by Instone (2000), amazon.com addresses this problem by using both explicit and implicit user information to support lifecycle personalization. Although Amazon tracks a clientıs purchases and makes recommendations based on those purchases (implicit information), Amazon also allows clients to determine which of their purchases is used to determine recommendations (explicit information). For example, if a client typically purchases books on information architecture and human-computer interaction, Amazonıs recommendations will reflect the clientıs buying habits. However, if that client purchases a number of childrenıs books as Christmas gifts for various nieces and nephews, Amazon will start recommending childrenıs books to that client. Since the client is not really interested in childrenıs books, he or she can remove the childrenıs books from the list of previous purchases that are used to determine their personal list of recommendations. Although clients may not take the time to provide the explicit information needed to make good recommendations, amazon.com does permit clients to have an active role in personalization by allowing them to control exactly what information is used.
Providing appropriate personalization features is a careful blending of user needs (which are defined by a user model) and content, governed by what Instone (2000 December) refers to as ³personalization rules². These rules are typically based on both a business model as well as ³the quality of the user experience² (Instone, 2000 December). The personalization rules are then used to match the user model with the appropriate content, thereby providing users with a personalized web site. To use this framework of user needs, content, and personalization rules, the information architect must define all of these components appropriately. User needs are defined by a userıs interests and behaviors, and could be determined by doing user studies, focus groups, or by developing and acting out user scenarios. Content can be defined by a companyıs business model, by looking at existing content, and can also by determining what information is necessary to meet the userıs needs. Perhaps the most difficult component to define in this framework is the personalization rules. However, applications are available that enable a company to input user and content information and generate personalization rules (Hall, 2001). As an alternative approach, template-based applications exist that allow developers to define different categories of information, then tag content accordingly (Hall, 2001). Regardless of the method used, it is important for the information architect to make sure that the personalization rules or template serve up appropriate content and structure based on the user.
Another aspect of personalization that must be considered is the fact that, once a web site or application is personalized to a user, the information that users see may be limited by the personalization scheme. Thus, it is possible that users might miss information that is important to them, but that has been deemed unimportant by the user model that determines the information displayed. In addition, unless the user can control some aspects of personalization, the user may not even know that a site has been personalized, or exactly what aspects of the site are personalized! For example, it is obvious that the modules available at my.yahoo.com are customized, since the user selects the modules they prefer, but is the content of those modules personalized? If the user wants the news module, do they only get local news based on their zip code or city of residence, or do they get national and international news as well? If so, what sources do those news stories come from? Is it just the Washington Postıs headlines, or does my.yahoo.com gather headlines from CNN, the Associated Press, and other sources as well? It may be impossible for the user to tell.
Because the goals of personalization include building the userıs trust (Fink, Koenemann, Noller, Schwab, 2002), and this includes building trust in the sources for the personalized content provided, as well as trust in the fact that users can control the information they receive. Thus, the information architect must make sure that users know that information is personalized, and in what way. If information is available (or can be made available) that users might not normally see because of personalization, they should be able to access that information and control it if desired. Currently, most sites seem to hide information about personalization features deep inside the ³legalese² section of the site; however as personalization becomes more prevalent and users are more aware of it, they may demand that this information be made more readily available.
As personalization and customization become common e-business models, information architects, UI designers, content providers, developers and others must be aware of both the benefits and the limitations of personalization and customization. Issues such as users not wanting to provide the information necessary for appropriate personalization, privacy, and the difficulty in developing and maintaining appropriately personalized content and structure must be addressed before creating a personalized site that clients will visit repeatedly. The issue of privacy is most prominent, and will probably continue to be so until more countries take an active role in defining usersı rights to personal digital information, regardless of whether that information is gathered explicitly or implicitly. Of course, the lack of boundaries on the Internet make enforcing any laws all the more difficult, but even if laws are not enacted, those sites that users feel violate their privacy will not be visited, and thus will be weeded out. As time goes by, the users themselves may end up defining what constitutes acceptable information gathering and distributing practices by their use (or lack of use) of certain sites. For those developing personalized sites, information on user-accepted types of information gathering would be helpful.
Providing appropriately personalized content without restricting information access, as well as ensuring that personalization features evolve with the userıs needs, are also key issues facing personalized web site developers. Although there are methods, based in the realm of usability, for defining user models and categorizing content, developing personalization rules that bring user models and content together is a newer area. It seems that most of the personalization tools available provide developers with little insight as to what exactly is going into their personalization rules, which for some information architects may be a problem. In addition, there does not seem to be a good framework or set of guidelines available on developing personalization rules. All the examples found during the course of this study were too vague for practical application; thus it seems that a set of guidelines on creating personalization rules would be useful for the advancement of the field of Internet personalization.
It will be interesting to see what the future holds for personalization. Because e-business and the Internet are by nature consumer driven, the evolution of personalization in these areas will almost certainly be determined in a large part by users. Their perceptions of digital information privacy, their willingness to provide personal information, as well as their idea of what constitutes useful personalization may help decide which personalized sites will succeed and which sites will fail. Information architects and others in the realm of web site development do not as yet seem to have available clearly defined guidelines and processes that enable them to consistently successfully cater to usersı personalization desires, but the work that has been done in the field and techniques that are continually developing will hopefully ensure that future developers can create personalized sites that consistently fulfill both usersı needs and their companyıs business goals.