Towards continuous usability evaluation of web documents

Towards continuous usability evaluation of web documents

Yin Leng Theng, Gil Marsden and Harold Thimbleby

School of Computing Science
Middlesex University
Bounds Green Road
London N11 2NQ
{y.theng, g.e.marsden, h.thimbleby}@mdx.ac.uk

ABSTRACT
To ensure continuous usability evaluation of web documents, this paper proposes integrating evaluation methods and techniques into practical authoring tools.

KEYWORDS:

INTRODUCTION
As the use of the Internet and the web grows, scalability not only refers to the handling of the increased number of servers, but also of handling the increased number of end-users. Because the web was not designed to handle so many and such large applications with more and more people using them, there are potential problems associated with the web, of which navigation is one of the more pressing problems. This view was supported by the results of the 8th Web User Survey by the Graphic, Visualisation and Usability Center conducted over October/November 1997 (Pitkow, Kehoe and Morton, 1997). The report showed that navigation is still a problem (16.7%) despite much research effort being invested to address it, as opposed to the top two problems of data privacy (30.5%) and censorship (24.2%). Often the development process of the web site follows the six stages described in the well-accepted iterative development lifecycle: (1) feasibility study; (2) conceptual design; (3) building; (4) implementation; (5) integration; and (6) maintenance. Unfortunately iterative design, which usually helps improve systems, has problems (Theng, 1997): (i) there is a lack of a disciplined and systematic approach to designing well-structured web documents to meet end-users' behaviour and navigation needs; (ii) prototypes are not thoroughly tested before being developed into the final system; and (iii) end-users involved in the process experience too little of a system to help make significant design contributions.
If the conventional iterative development process is found lacking, better ways to ensure that good web documents are produced are needed. This could be achieved by involving end-users and taking into account their needs throughout the design process. To ensure continuous usability evaluation of web documents, we propose integrating evaluation methods and techniques into practical authoring tools. These evaluation techniques are categorised under real user testing and non-human user testing (Theng, 1997). Real user testing includes observations, surveys, expert evaluation and experiments, and should be carried out before the system is ready for implementation so that qualitative results and impressions can be obtained. Non-human user testing methods are encouraged as a means to perform evaluation early enough to influence design while it can still change direction. Analytic and heuristic evaluation methods, and executable user models are some ways of evaluating without requiring the attendance of real users. We believe that if designers were to apply these methods and techniques religiously to the development lifecycle, quality web documents could be achieved.

WORK ON CONTINUOUS EVALUATION OF WEB DOCUMENTS Various authoring tools have been developed at the School of Computing Science (Middlesex University) as illustrations of our approach to provide better support and more systematic usability evaluation on web documents.
HyperAT: Tool supporting the different modes of usability testing Apart from the basic editing facilities of creating, editing and saving, embodied within HyperAT is an experimental, authoring testbed which allows hypertext designers to carry out different modes of usability testing on the web documents created by HyperAT, all within the authoring environment of HyperAT: (i) structural analysis which formally analyses the structure of the web documents; and (ii) real user evaluation which analyses end-users' browsing behaviour based on real users' transaction logs. The ability to toggle between different modes makes testing less cumbersome, and hence more convenient for designers, thereby increasing the chance of creating more usable web documents. The first and second modes of usability testing have been implemented in HyperAT. As future work, we also recommend exploring the potential of using executable user modelling or non-human user testing as the third mode of usability testing.
Tools for site authoring and maintenance of the RSA web site
Three tools have been developed as part of current research into distributed web authoring, whilst building a web site for the Royal Society for the encouragement of Arts, Manufactures and Commerce (RSA). These tools allow the production of sets of pages that have a consistent and easily changeable design/layout, achieved by separating design from content. Gentler (Thimbleby, 1997a) was written in HyperCard and was used to create the first 'automatic' RSA web site, using a database of information provided by the RSA. Gentler managed the site from 1995 to 1996, when it was superseded. Siteview (Thimbleby, 1997b) was written in Java, built on the ideas of Gentler, using content or 'source' files whose presentation was controlled by a number of design files - thereby allowing quick and reliable alterations to the visual appearance of whole subsections of a web site. Siteview proved useful as it provided semi-automatic generation of certain site elements, such as consistent navigation bars, link checking and a graphical representation of the site structure. Building on the ideas introduced by Gentler and Siteview, we have more recently developed a third tool known as StyleGeezer. This tool concentrates on delivering the core features of the earlier tools in a simpler and more usable way. Using these tools, designers can easily recreate the site, making global changes to the design - for example, changing background colour across the whole site without editing each individual page. This ensures consistency, and makes the maintenance task simpler, thus helping designers to manage the complexity of the design and maintenance processes. Future work includes validating these tools with types of designers (e.g. novice, intermediate, experienced).
REFERENCES
Pitkow, J., Kehoe, C. and Morton, K. GVU's WWW User Surveys (1997),
http://www.cc.gatech.edu/gvu/user_surveys/survey_1997_10 Theng, Y.L (1997), "Addressing the 'lost in hyperspace' problem in hypertext,"PhD
Thesis.Middlesex University.
Thimbleby, H. (1997a), "Gentler: A tool for systematic web authoring," in
Buckingham Shum, S. and McKnight, C. (eds.), Special issue of International
Journal of Human-Computer Studies on 'Web Usability'.
Thimbleby, H. (1997b), "Distributed web authoring," WebNet'97 World Conference
of the WWW, Internet and Intranet, Canada.