Fourteen Users in Search of a Newspaper: the Effect of Expectation on Online Behaviour Ann Light School of Cognitive and Computing Sciences University of Sussex, Falmer, Brighton BN1 9QH, UK February 1999 CSRP 507 Fourteen Users in Search of a Newspaper: the Effect of Expectation on Online Behaviour ABSTRACT People expect the friendly familiar paradigm of media to guide them through uncharted territories of information. [1 p38] The potential differences between newspapers and online news services are still being explored. Producers have come a long way from reproducing newspapers verbatim on the Web, but have readers made a comparable leap in their expectations? This paper reports the reactions of a sample of first-time visitors to The Guardian's Election 97 website, offered by the newspaper during the run-up to the British general election. Visitors were asked to evaluate this news site alongside election sites created by a broadcaster and a new media producer. Findings suggest there are major obstacles to innovative news practice on the Web, especially where there is identity with a paper product. There was a tendency for the evaluators to overlook or reject interactive elements on the Guardian site. The paper concludes with some recommendations for site design which could help in integrating interactive aspects of the new environment. Part One: Introduction 1.1 Introduction The aim of this paper is to investigate the potential of online news services. The paper is based on findings from a project exploring the use and usability of The Guardian's Election 97 website. The usage project was undertaken at the request of the newspaper as preparation for the development of the Guardian website into a more comprehensive news service. The paper sets the Guardian project in the context of developing online news media and then summarises the findings and conclusions that have significance for other news media producers. It deals with the expectations and first-time impressions of a sample of visitors to the election site during the run-up to the British general election of 1997. It concludes that people's expectations of newspapers did influence the way that they regarded the news site and also, to some extent, how they moved about it and where they went. This finding has implications for how other sites of this kind are developed. The paper includes some initial recommendations about how these findings might be incorporated into successful designs of the future. 1.2 Paper vs Online transmission Clearly, some aspects of newspapers cannot, at present, be transferred to the digital medium. For instance, the physical process of news consumption alters dramatically: the see-at-a-glance big pages that can be spread on the table or folded on the train - factors that influence people to choose newspapers rather than watch or listen to the news - are not reproducible on the 1990's computer screen. Other aspects of paper reading, such as the opportunity to browse news at a chosen time and speed, read analysis and study graphically presented information at leisure, can be carried over to the Web. But there is evidence that reading on the screen is more demanding than reading on paper [2]. On the other hand, compensations include some broadcast conveniences: in theory, news can be updated whenever an event requires it. Additionally, there are the factors that are exclusive to the digital medium. The Web's potential for mediating interaction between people (and between people and machine) has only been touched on in development work so far. For many producers, interactivity means putting paper contents on webpages connected by forward and backward buttons. This offers little more than teletext. 1.3 Newspapers' response British newspaper publishers have generally been slow to respond to the challenges of the new medium. Hammond [3 p61] talks about the `poorly thought-out attempts' of newspaper publishers to transfer their goods. And things have only improved slowly. Delano attacks `the process by which data assembled for one medium is converted for use in another. This is already evident in the on-screen Web sites occupied by any number of publications - and dull and dreary places they are, too' [4 p54]. Much the same is true in other parts of Europe: `There is a contradiction between the declared intention of providing Web services as supplements, and the current practice of making Web services very similar to the newspapers themselves' [5]. But it is not just conservatism that keeps newspaper producers from innovating. Hypertext does not have boundaries in the same way that paper products do. A visitor may specify a site by location, search for it or stumble on it. Regardless of the amount of promotional work, there is no way to guarantee that one's visitors have planned to be there. This adds fuel to concerns that publishers have about recognisability. Some providers have deliberately minimised the differences between their online and paper products. For instance, the Telegraph publisher committed itself to carrying most of the parent paper word for word in the Electronic Telegraph and did just that for two years. Despite being the first British online national paper, it was slow to add original material to bring it into step with the publishers that followed [6]. But its goal was to recruit a new generation of readers for the paper product and in this it has been successful [7]. The issue of identity is critical here as the commercial reality for most news websites, like others, is that they lose money, justifying their existence to the company accountants by promoting the parent product. The close linking can also work to benefit the site. Newspapers' sites have, through their brandname, credibility as information providers in the glut of publishing on the Web. And credibility is a rare commodity: because distribution charges are nominal, anyone with access can be a publisher. Hence `none of the standards of authenticity journalists set for themselves, nor the limitations they have come to accept, apply. Information can be invented, sources misrepresented, images manipulated and falsified'. (Delano p55). However, a distinction must be made between brand image components, such as credibility, and the product image - its look and feel - although the two closely reinforce each other. Advertising strategies often seek to blur this distinction to the consumer: creating a certain set of brand expectations; which may relate directly to, or work as associations with, the product [8]. And information producers, unlike most developers, find themselves in the interesting position of building sites that both are the product and exist to promote it. 1.4 The Guardian's New Media Strategy The Guardian site was for a long time a composite of supplements: linked by a Web strategy, rather than a single conceptual entity. The election site was within a notional Guardian site, linked to other sections through a mini search engine. This atypical presence for a newspaper developed from the themed nature of the parent's daily supplements, starting with the transfer of science coverage in early 1996 (most newspapers concentrated on transferring the main contents of the paper to their site). But its design also sprang from a specific view of news and the Web and a desire to experiment. The sites' coverage have been based round significant events, rather than the contents of the newspaper - using the currency of the event to attract visitors, instead of relying exclusively on the concept of the newspaper. This model attracts the occasional reader, just as they are drawn to buy a paper by a front-page story on a royal funeral, though they do not routinely buy one. Events were chosen in areas which would appeal to the paper's target market, such as the Ashes cricket series, the run-up to the Orange literary prize, the 1997 general election campaign. The episodic nature of the sites worked to enable the team to experiment without endangering the consistency of one particular site by 'fiddling'. The election site went further than other Guardian sites, with its large and changing news content and its political aspirations, to take the Guardian's corporate identity to the Web. Its development was conducted by a group comprising a navigation expert and representatives from the design, editorial and technical units. 1.5 The Guardian and Interactivity Although all the Guardian newspaper's election news was also published on the site, there was a commitment to various kinds of interactivity as part of an experiment into news presentation on the Web. These included: * a database of constituency information searchable by candidate, location and party, which included the voting record of the present MP * several discussion forums, collected together as the Debating Chamber * some lighthearted devices, such as a quiz, a straw poll and a vote-o-meter (which allowed visitors to measure their policy choices against the three main political parties' manifestos and then gave them a score indicating their leanings). There was, in part, a political motivation for taking this step. As well as meeting the need to stay abreast of developments and improve its background coverage of the election, Guardian management saw the website as offering the higher goal of extending political education and debate by supplying information about parties, policies and elections - and a place to discuss them. This idealistic agenda had the commercial benefit of being in accord with Guardian readership aspirations [9]. There is as yet limited analysis of how electronic media might affect the political process. `There can be a shift in balance from passive reception to active search, which can tell in favour of an improved political communication situation', [10 p146], though, at present, the advantage of gaining this valuable political and economic information belongs to the elite with access. Discussion groups can be regarded as another politically enabling element of on-line service. McQuail sees the tendency to open newspaper columns to `alternative, sometimes conflicting points of view within broad spectrum consensual newspapers' (p146) as promoting political understanding. The parallel with providing online forums is obvious. In supporting less sensational information than the main paper, such as census data, changes to constituency boundaries and party manifesto details; in making these details available to search; in updating the existing store throughout the campaign and in providing room for the readership also to discuss the content and other related matters, the presentation is less partial. It can offer substantial background information to inform, rather than influence. It can challenge the tendency to soundbite politics by providing users with access to the whole story in the abundant space on the Web. However, people may not want to interact with their news. `While interaction allows a remarkable degree of user control and independence, it also demands thought...there are signs that interactivity can be too demanding for some people's taste.' [11 p9]. Television news viewing is essentially passive. News reading, which involves more activity, has a history of being conducted in short, superficial bursts - reflected in the shape of news stories, with a summary for a first paragraph. Interaction may only be attractive at the level of selection of stories, rather than 'participative interactivity' like discussion [12]. Part Two: Users and their Expectations 2.1 The Web User According to the 1996 GVU survey [13], which is a voluntary survey and reflects this in its findings, web users have an average age of 35 years and 31% are women. The median income is $50 000 and many users work in computing or education, perhaps giving rise to the high access figures: about 70% view the Web daily. The Guardian newspaper readership tends to be professional men and women, working in the media, voluntary sector, in social care, education and computing. Their politics are centre-left. The Guardian site straw poll and other Web-based daily polls during the election showed a marked centre-left tendency, suggesting that Web visitors to British election sites - or at least those prepared to interact with them - generally share this trait with Guardian readers. There are no figures for how many Guardian newspaper readers visited the Guardian site, though there is evidence that people who like the paper but have no access (such as those abroad) did visit. One certain qualification is that any Guardian readers who did so were part of the subgroup with full internet access. The only way to sure that the site's visitors were typical in other ways of Guardian readers or of Web users would have been to require them to fill in a registration form before entering the site. Registration was only required to access the debates. To conclude, the typical Web user and the typical Guardian reader seem not dissimilar, though readers from areas such as social services are less likely to be represented on the Web. 2.2 The Visitor's Purpose Dozier and Rice produce a good summary of why and how people read news, applying it to early electronic newspapers [14]. They identify two main modes of reading news: * to accomplish the task of finding information that is of benefit, such as sales figures; * to pass the time entertainingly, by reading what first catches the eye and then sustains the interest. Of course, choice of mode hinges, to some extent, on whether the individual likes reading or not, but also on the amount of time available and other local factors. The parallel with how preliminary research suggests the Web is used, is clear. Schneiderman contrasts browsing on the Web - for pleasure - and searching for information [15] in a similar way. 2.3 Expectation However it is not sufficient to identify that similar purposes exist in media consumption both on and off the Web and that these generate parallel behaviours. Unless these parallels exist in visitors' minds, they remain no more than a reflection of potential. The common model of the Web is a seamless connection of pages across an arbitrary number of sites [16], which is easy to use because the visitor's browser gives a semblance of consistency. `Users don't want to read any manuals or help information for individual sites but demand the ability to use a site on the basis of the Web conventions they have picked up as an aggregate of their experience using other sites'. Set against this will be expectations of particular sites, especially if the brand is familiar in another context. Some conflict might be predicted if the models require different approaches. So producers must find models that interpret their product appropriately for the new environment, while maintaining an awareness that the new environment itself brings expectations with it. At present, they have less control over how their efforts are perceived than they would like. The example below demonstrates one of the possible frustrations in experimenting: L'eggs tights site, in 1996, was attempting to present a general lifestyle chat site, which was the perceived wisdom in site building at the time to attract a following. `Irrespective of the assigned topic, participants - many of them men - were talking about tights... The deeper lesson: no matter how hard brands try to become something more, folk seek them out for what they actually are.' [17 pp36-7] Or, looked at another way, if producers are relying on people to recognise product branding in a buying context, it would seem reasonable to expect people to be guided by it in other contexts too, particularly unfamiliar ones. There is considerable literature on mental models and the effect they have on the execution of tasks [18]. That people compose notions based on models and thus carry expectations about purpose and function is relevant to their ability to interpret what they find. A good conceptual model allows us to predict the effects of our actions...When the model presented to us is inadequate or wrong (or, worse, nonexistent), we can have difficulties. [19] But expectations can affect judgement more profoundly, through selective distortion: set attitudes lead to expectations about what one will see or hear. People perceive what fits into their belief system, adding things to the message that are not there (amplification) and not noticing other things that are there (leveling). This level of preconception has implications when one considers the differences between newspapers and the online news services modelled on them. Will visitors' expectations blind them to certain aspects of sites? 2.4 Metaphor There is good reason for perceiving these newspaper-branded news services as online or electronic newspapers and they are often called by these terms. But what is the trade-off in presenting them like this? The concept has obvious limitations, even if no additional interactivity such as a search mechanism is included in the site. The term `newspaper' in this context is a metaphor for an assembly of updateable databases, distributed by a network of computers and viewed by pressing buttons. It is unlikely to pose problems that factors such as portability are outside the scope of the metaphor. Call something an online newspaper and very few people who understand the term `online' will expect to be able to fold it. But, depending on their model of the Web, they may expect to be able to contribute to it, or they may see it as the essentially one-way information provider that paper-based products are. And more subtle suggestions may be incorporated which emphasise some aspects and play down others. Most concepts are partially understood in terms of other concepts [20], but the use of `newspaper' at this point still has a freshness not attributed to the majority of metaphoric terms in our vocabulary. It has not yet gained a fully distinct meaning, set apart from its metaphoric allusions. Does this freshness cause problems? Part Three: The Project 3.1 Aims and Objectives The project with The Guardian had two linked aims. The first goal was to explore how people interact with and through a site based upon a newspaper, with a view to learning more about how traditional and new media differ. The other was to find out more about the use of the Guardian's election site in order to provide information to the newspaper, which could guide the designing of future sites. This paper focuses on findings which relate to the former. The relevant objectives were: * to compare response to the Guardian's site with other prominent sites carrying similar material, * to investigate which aspects of the Guardian newspaper transferred well to the new medium, * to explore reaction to the interactive sections of the site. The decision was made to keep the emphasis at all times on internal factors, relating to structure and content, rather than external issues, such as speed of transmission and the effect of different browsers on what is seen. 3.2 Timescale The date of the general election was announced on 17th March 1997, on which day the Guardian began promoting its election site. The new British government was elected on 1st May, with all but a couple of seats outside Northern Ireland decided by the afternoon following. The new media team then withdrew the resources that had been allocated to the site and, having written up the results, kept just a limited political news service running. Visitor observation was carried out in the final week of the election build-up. After the election, data from the site's log for the whole election period and the days following were processed. 3.3 Existing Knowledge of Site Use The Guardian uses a freeware package, Analog [21], to process the logfiles of requests to its sites. So the team already possessed figures for numbers of visitors to the election site and the number of times individual pages had been requested. The team knew which pages were most popularly summoned, but not by whom, or why, or for how long they were looked at and why. 3.4 Adopted strategy The data sources available for this research were the site's log files, the Analog-processed statistics about requests, the registration data and the site itself. An overview was provided by analysing sections of the site's request log. This was made to yield approximate figures2 for the shape and duration of visits to the site. This provided a check on and a context for the predominately qualitative methods used elsewhere. The log was analysed to see whether any generalisations could be made about visitor's visits, concentrating on the following questions: * Is there a difference between one-time visitors and multiple-time visitors? * Where did these people go within the site? * Did interactivity play a significant role here? This analysis also revealed one of the weaknesses of relying on page request data to assess page popularity. The log showed that only 10% of page requests were for the day's news, which the team found a surprising figure. An examination of a significant sample of visits showed that more than 50% of the time spent on the site was used in visiting pages displaying the day's news. As part of the project, visitors to the Debating Chambers were sent a questionnaire asking about their behaviour, perceptions and view on participation, but findings from this are not included here. The third and most revealing section of the study, was the observation of visitors as they evaluated a variety of election sites. This was intended to yield detailed information about behaviour as it related to opinion and expectations. A description of the methodology and findings is given below. 3.5 Observing visitor evaluations By videoing visitors as they conducted a `think-aloud' evaluation of the site [22,23], an observer could employ the dual technique of listening to what they said while watching what they did. Not only did this allow access at a time when opinions were being formed, it also showed the kind of experiences that led to their formation. A transcription that recorded the evaluators' locations, actions and comments formed the basis of the analysis; this supplies all the quoted material in the following sections. While there are generally drawbacks with demanding concurrent verbal protocols [24], in this case talking did not greatly compete with conducting the task as most descriptions took place in the delays between page request and page arrival. The log of comments and actions was expected to show both a text of intentions and expectations and a subtext of these affecting passage through the sites. Particular interest in the Guardian site was concealed from the evaluators. The methodology was chosen to answer questions about visitors' perceptions and their use of the site, such as: * How do people greet the site on first visit? * What would visitors expect to find? * Is this determined by the newspaper-ness of it? * Are there difficulties understanding its design? * How drawn are visitors to the interactive components? * What is perceived as acceptable? * Is it seen as doing as good a job as its peers? 3.6 The task The task given to the evaluators was as follows: The task is to visit three sites: the BBC election site, the Guardian election site and the Online Magic GE97 election site (OM) and compare them - all the time speaking your thoughts aloud. You have freedom to look at any aspect and spend as much or as little time as you want with each. Feel free to read things if they catch your interest, or move on if you are bored. You may start with any of the three. The only thing I would ask you to do is to say what comes into your head when you see or do anything. I may ask you questions, especially if you go quiet. The general task was followed by a structured verbal questionnaire which required the evaluators to summarise their reactions and say whether they would visit any of the sites again. It also collected information about media habits and personal details. 3.7 Conducting the evaluations The setting up, briefing and debriefing followed the model in Monk et al (pp73-80), employing the techniques of co-operative evaluation. Where the technique deviated from theirs, it did so by involving a less defined task than is recommended, accompanied by a greater acceptance of evaluator choice in what to do. This freedom, it was made clear to evaluators, was a permissible component and the reason for very little guidance. 3.8 Evaluators' characteristics Fourteen people, eight men and six women undertook the task. Their ages ranged from 21 to 41, with nine under 30. They all had first degrees, except one and came from a variety of backgrounds including media, computing, law and education. Two were not eligible to vote because they were foreign nationals. 11 of them would expect to read a copy of the Guardian at least once weekly, though only two read it daily. All were interested in the election and all but one followed news of it through other media. One had glimpsed the Guardian site before; another had used it for a constituency search. Three others had visited other election sites. All were well acquainted with the Web. All of them knew the equipment with which they were working. Regular visitors to the site would have been inappropriate here as one purpose was to get initial impressions. There is no way of confirming that this was a good cross-section of potential visitors, but it looked plausible, corresponding well with the characteristics mentioned in section 2.1 Part Four: Findings 4.1 Observations relating to the Task People coped well with the degree of freedom given them. They spent, on average, about an hour across the three sites, but it varied greatly. Comparing this to typical visits from the log, it can be seen as unusually long. Having opted to conduct the evaluation, most visitors approached it in a leisurely fashion, indulging themselves by looking at what they fancied - with neither phone bill nor work constraining them. No one had a problem with the task, though certain aspects of using the sites caused difficulties. There were two predominant modes of activity: evaluating (which could be equated with searching, as being task-orientated) and following an interest (browsing) and people moved effortlessly and frequently between the two behaviours. No two people approached the task similarly. There was noticeable orientating behaviour from a couple of evaluators. Both these people found maps where they existed and seemed more `at home' afterwards. This contrasted with a plunging in approach, which involved finding something that looked of interest by scanning the home page and starting with this to see if it was all it appeared to be. One of the most intriguing behaviours on display was some evaluators' tendency to describe the expected contents behind links differently depending on which site they were looking at. This would be influenced to some extent by local context, such as the look of links, but in listening to their justifications, seemed to emanate largely from what they thought the particular producer would provide. For instance, the quiz on the Guardian site was very similar to that presented by OM in content, but the OM quiz was considered more `fun', even when it was merely noted, rather than visited. This section of OM was also more likely to be recollected in summing up. In fact, often, the expression of what was expected to be behind a link was accompanied by the action of going somewhere else, so that the accuracy of their prediction was never tested. It formed part of the decision making process of choosing a route. 4.2 Observations relating to the Site Evaluators were confident in setting off from the Guardian home page, even though some expressed disapproval at the length of the lead story - a regular newspaper page-lead length - and lack of hyperlinks in it. A praised feature was the arrangement of news and issues summaries, which led on to fuller information through a neat and consistent `more' button. This arrangement promoted confident reading in two ways: firstly, the small units were easy to digest, and, secondly, visitors could see what the competing sections held, which reassured them that they were reading the most interesting content available. 4.2.1 Benefits of the medium There was a positive response to the amount of information that the site was able to offer and the fact that the information could persist till the evaluator wanted to see it, rather than depending on topicality for publication: [Summary] ...One thing the internet can do better is give you access to information not everybody wants, which a paper wouldn't be able to print... and: mGuardian: That's good, you've actually got the facts here. Blimey, it's got everything - well, that's useful... Presumably this information will exist post-election so you can look it up. 4.2.2 Interactive frustrations A few evaluators alighted on the button to the Debating Chamber to look at the debates. This brought up a registration form and did not allow access without completion. Only one evaluator even considered filling it in, abandoning it part way through. Curiosity was less compelling than dislike of registering. Evaluators began filling in the quiz or vote-o-meter questions they could see on the screen and then paused to scroll to the end of the section and back. Having, seemingly, reassured themselves about the length of the task, they continued to choose answers. This behaviour happened despite an introduction telling them the number of questions, which most people did not read. The behaviour contrasted with that during news reading, where some long articles were either simply abandoned, or abandoned after the length had been checked. People wanted to select multiple options as answers in the vote-o-meter but the system didn't support it. There was also indignation with the available options and a desire to enter comments. The reason for the choice of options was given in the introduction - they were derived from the three main party manifestos - but no one read this. 4.3 Observations relating to Preferences The transcriptions show a spread of preferences across the three sites, with the Guardian and BBC sites appearing more popular. The following list refers particularly to the Guardian site: Appearance: A third of comments about the appearance of the Guardian site describe it as a newspaper: seven of the 14 comments are positive (1 newspaper comment), five are negative (2), two are neutral (2). Descriptions included 'cleaner', 'that nice white space' and 'all very smart'. Five people mention the bright colours of the OM site; all but one preferring the more subdued tones of the Guardian and the BBC. Content: A quarter of comments about content relate the site to the newspaper; the rest compare it with the other sites: 13 (3) comments are positive, 3 (1) are negative. Mention was made four times of the strong editorial line that the newspaper site presented, all favourable. The Guardian coverage of background information was praised by several people. There was enthusiasm from some evaluators for the voting records of the sitting MPs included as part of constituency data. Structure: The Guardian site structure was mentioned six times positively or as preferable to that of the other sites and once as less desirable than a newspaper: 'Now this looks to me, reading it, it's written for reading on paper, it's not easy to read online - online I'd rather see bullet points.' 4.4 Observations relating to Expectations There were 24 specific references to what was expected on the three different sites and further allusions, with most people expressing clear expectations of each site, or the absence of any expectation. In deciding where to begin the evaluation task, all but one person used their expectations of the parent product to chose an order for examining the sites. Answers to the observer's question: 'Why did you begin with this site?' included such responses as: 'the bastion of independence [BBC]' or: 'I like to have my prejudices reinforced [Guardian]'. Others went to the OM site first specifically because they had no expectations of it. (Once it was seen to be sponsored by Yahoo, a whole new set of judgements appeared - see below). Their evaluations showed a similar domination. The products were not so much being compared with each other, but each with the expectation created by the original thought of it: [BBC:] As expected, it's vaguely amusing and highly educational. The Guardian? Like a newspaper. And Online Magic? Like a Yahoo site, bright and colourful and not an amazing amount to it. Most references to Yahoo were negative: 'all the characteristics of Yahoo, a bit of a mess' and 'I wasn't expecting much of Yahoo', being typical. Implicit in this was higher expectations for the two others, although there was an exception to this: OM: ...too much text and not enough headings - from internet experts I would expect a lot better. [[Would you give allowance to the others for not being `experts'?]] To some extent: the Guardian being made up like a newspaper, marketing it under the same name and people will expect that. Having said that, they have been fairly innovative and made it very clear. Personally, I think that's better, you can find what you want, than the Yahoo site which should be internet friendly. Here, her perception that it was a new media company behind the site influenced her judgement. Two summaries are worth citing in full to illustrate the strength of the prejudging that was going on. The first is that of someone who had difficulty with the interactivity of the site because his perception of how a newspaper works is so strong (see below). His is aware of the role of expectation in guiding, but only expectation of content, not of form: I suppose expectation is the key thing here. I started out by going for the Guardian, I want to go to a site which discusses my views as sympathetic. I'm a regular reader of the Economist [featured on the OM site] so I go for them. I more or less know how to find my way around what they think. I hadn't realised how far this conditions how I go around the Web. I'm importing my reading habits - except when I'm searching when I expect to get information - otherwise when browsing I let my prejudices have full sway. The second is that of someone who moved swiftly through all three sites and then based his analysis on what he thought was there: I quite like Yahoo - lighthearted. I wouldn't expect lighthearted from the BBC or Guardian. [[Is that what you got?]] Yeah, pretty much, could tell from the sections that I flicked through, though didn't read it in detail. Suppose I got what I would have expected. The most dramatic instance of expectation affecting outcome was this struggle with the vote-o-meter: [Vote-o-meter] (reads the introduction, goes through changing weights of importance for each question but only reads policy options, doesn't tick any) It's not entirely clear what I'm doing but I'll do it and we'll see... These statements do not reflect... ...[[is it fun to do?]] Well, I'm doing it, but I'm more interested to see what they think are the important issues, it's a long time since I thought my views were representative. What is interesting is an insight into the editorial policy of the Guardian/Observer, sort of traditional leftie-liberal. The same thing put together by the Telegraph would be completely different...I've said that everything is important, let's see what happens (submits, pauses, surprised, and reads) Oh...oh, that's nice - I haven't got a score! (quite upset, but diverts himself with another section of the site: Issues) [Issues] Can't be bothered to compare all the policies, don't want to look at manifestos (moves off Guardian) [[Summary]]...The only [interactive component] impressed me was the one that allowed me to check out about the constituency. All this stuff about how important are issues and then didn't give me a score, fairly pointless and seemed rigged anyway as various things that they seemed to think I'd find important, I wouldn't. I preferred Yahoo, having got started - questions did appear to be specifically answered. What seems at first sight to be a simple oversight in filling in the vote-o-meter form, looks more complex if the previous sequence of comments is examined. The vote-o-meter had sections on major policy issues such as transport. Each section required the user to select a preferred policy from three shown, with the additional chance to weight the importance of the issue relative to the other issues featured. The system could not evaluate someone's political persuasion unless they chose at least one policy option - he left them all blank. But, in fact, it would appear that he is not expecting a score. When he reaches the next page, he expects to see some further information based on the policy areas he has marked as important. Hearing his thoughts as he goes through the questions suggests that he thinks he has been requesting the Guardian's perspective - not his own. His model of a newspaper seems to be that of an information provider and does not include self-assessment quizzes. As a result of this experience, this evaluator became confused and then hostile. He went from expecting to prefer the Guardian site, to being angry with it. Part Five: Analysis By comparing a 'newspaper' election website with a 'television' election site and a 'new media' election site, evaluators were given a chance to respond differently according to the media genre of the parent producer. This difference between sites was commented upon by a couple of the evaluators and may have influenced others in their identification of the `newspaper' site with a newspaper. There were a number of instances indicating that the evaluators were looking for different things from the sites, based on their knowledge of the producer's dominant medium. An important finding from the study was the limited usefulness of explanations, especially about interactive devices - certainly as the Guardian presented them. Evaluators found themselves mystified by the length of the quizzes and the reason for the policy choices in the vote-o-meter, having ignored the introductions. However, people were more likely to read the search instructions on the archive and constituency pages. Here, the visitor is behaving in a more task-orientated fashion, knowing the purpose of the search, while the goal is placed on a subsequent page - either or both factors could make the difference. This has implications for site design. Where there is the visible distraction of buttons to press, this pulls visitors into engaging with the functionality of the site rather than reading. If the new environment makes visitors even less likely to read than normal, then it is here that intuitive design is needed So, how well did the paper's identity transfer to the online medium? The look of the paper was recognised by several of the evaluators, who found the space (the `white space' design of the paper) and simple colour scheme, easy to orientate through. By contrast, the length of the front page story was off-putting and one evaluator suggested bullet points would be more appropriate. Observation suggested that reading is less pleasurable on screen than on paper, but also suggested that it happens less as there are so many other distractions. Pressing buttons is more fun than reading and there is an awareness on the Web that behind every link might be something even more interesting to read, not a suggestion that presents itself as persuasively from paper. The Guardian's presentation style, which gave a page of news-in-brief summaries leading to each story or issue in more depth, is comparable to the way that the newspaper can show a spread of stories. Both allow the reader to choose their own hierarchy of interest. Observation showed evaluators' recognition of the editorial stance of the paper and most were happy to see it reproduced in the site, feeling it made for less bland copy. They talked with respect of the Guardian's values and political aims, having no problem with the subjective stance of some sections compared to factual elements such as constituency information. They were grateful for the additional information provided. However none of them felt motivated to participate in the creation of opinion by posting to the debates. In fact, with less than 3% of logged visitors viewing the forums, the presence of the debates was a marginal feature to most visitors, likely influenced by the need to register even to read them. The interactive sections of the site sparked some of the evaluators' strongest reactions, but got little acknowledgment in recalling the site in summary. However, usage patterns extracted from the log suggests that many people spent the time necessary to play with the quizzes and polls. This required far less commitment than getting involved in the debates. Evaluators' initial, but faltering, enthusiasm for pressing buttons suggests that presenting a number of simple choices to visitors is good business. The Guardian offered 10 at a time, but less might be more successful, especially if all could be seen on a typical screen at once. The vote-o-meter - with its dual controls for choosing policy and its relative importance - proved too complicated. However, there was evidence from evaluators' behaviour that some people were using a model conflicting with the nature of what they found; with the confusion created by the vote-o-meter epitomising this. There was also a tendency for evaluators to ignore aspects of each of the sites that did not meet their expectations and it seems likely the `newspaper' site was recalled so little for its interactive devices - compared with recall of the two other sites - because of the dominance of the news' presence in subjects' minds. There was one striking use that occurred throughout the election run up that was particular to the electronic product. There is indication in the log that many visitors who only visited once - estimated at 80% - came on a fact-finding mission. It might be inferred from the brief time that they spent reading news, that they had low interest/tolerance for newsreading on the Web. News was the key component that changed and therefore would bring a majority of returners back. More intriguingly, the amount of time spent in constituency and historical information suggests that many one-timers visited the site to see their constituency details and, possibly, to sort out what they needed to know about voting - especially in view of the significance tactical voting played in this election, and particularly among those of typical Guardian political orientation. They would not need to return. All evaluators confirmed this behaviour by visiting their constituency page and reading it. As a couple of the evaluators pointed out, the newspaper could not have carried this amount of information daily, if at all. Part Six: Conclusion It is evident that visitors' expectations affect what they want to find, what they notice and how they approach tasks found. This made the Guardian election site popular in its coverage of news and issues. But these findings highlight awkward tensions for any newspaper producer in the Guardian's position of wishing to develop greater visitor involvement: strong branding identifies the election site as a `newspaper' site, despite the innovations that have been introduced in products for the new medium. This branding strengthens the model that visitors have been shown to carry, that any product of the Guardian will be a kind of newspaper. The absence of this branding might be predicted to cause just as much confusion and dissatisfaction. Besides, branding is a critical marketing strategy, showing that the Guardian corporate concept is comfortably on the technological bandwagon. Conveying the absence of expected objects or actions is a design challenge, (Schneiderman), and this study's findings suggest that conveying the presence of unexpected objects and actions is similarly demanding. The medium's additional potential, including all customised interactive devices, may be overlooked, ignored or used incorrectly, leading to hostility or the impression that the site is barren. If a producer chooses to break with visitors' models, it must, in the face of possibility for misunderstanding or dissatisfaction, re-educate. However, the nature of the new medium does not lend itself to lessons. As Nielsen says, and these findings indicate, Web users do not want to read help manuals. Time may show that newspaper-branded sites should behave like newspapers; success hinging on a regular diet of interpreted news and well chosen links. This would severely limit the democratisation of news that the Guardian was exploring. Alternatively, perhaps, as visitors become more sophisticated Web users their attitude to the Web will overshadow their expectations of particular sites. This might increase demand for interactive and participative sections. A stream of new Web users will continue to appear - many of whom will lack the explorative nature of early users and may seek the safety of familiar brands as the Web grows increasingly overwhelming. The Guardian will have to deal with prejudice from visitors' expectations for several years while the medium is being appropriated and incorporated into everyday life [25]. In the meantime, certain measures might minimise the distance between visitor expectation and producers' innovation. The results of this study suggest some design principles might be extracted from observation and questioning, to ease the transition from novelty to acceptance. These would begin the assimilation of interactive elements into people's expectations - reining in the `newspaper' metaphor until it becomes a full concept - while keeping the trade-off to a minimum, between catering for novices and developing the site to be as interesting as possible. Site Addresses BBC election site: www.bbc.co.uk/election97/frameset.htm Guardian: www.guardian.co.uk Online Magic GE97: www.ge97.com References [1] Ray Kristof and Amy Satran, Interactivity by Design - Creating and Communicating with New Media (Mountain View: Adobe Press, 1995) [2] Kenton O'Hara and Abigail Sellen `A Comparison of Reading Paper and Online Documents' (Rank Xerox Research Centre Research Paper 1997) [3] Ray Hammond, Digital Business: Surviving and Thriving in an On-line World (London: Hodder and Stoughton, 1996) [4] Anthony Delano `Prepare for Do-it-Yourself News' in British Journalism Review, 7 no4 (1996) pp 53-56 [5] Lars Bo Eriksen and Pal S(rgaard, `Organisational Implementation of World Wide Web in Scandinavian Newspapers: Tradition Based Approaches Dominate' presented at the Second IFIP International Working Conference on Diffusion, Transfer and Implementation of Information Technology, Cumbria, June 1997. http://internet.adb.gu.se [6] Steve Outing, `NetMedia Brings UK Internet Press into Focus' (July 10 1997) [7] Hugo Drayton, `Beyond Paper: the Electronic Telegraph' presented at the Second Annual BSME/PTC Conference, Reaching the Reader, London, November 1995 [8] Sean Brierley, The Advertising Handbook (London: Routledge, 1995) [9] D. A. N. Jones, `What's News?' in The Press We Deserve, ed Richard Boston (London: Routledge and Kegan Paul,1970) pp27-39 [10] D McQuail, `Diversity in Political Communication: its Sources, Forms and Future' in Communicating Politics: Mass Communications and the Political Process eds Paul Golding, George Murdock and Philip Schlesinger (Leicester: Leicester University Press, 1986) pp130-149 [11] Tony Feldman, Multimedia ( London: Blueprint, 1994) [12] Ann Light, http://www.cogs.susx.ac.uk/users/annl/tax.html (1998) as at 12/4/98 [13] James E. Pitkow and Colleen M Kehoe, GVU's WWW User Surveys, (Dec 1996) Georgia Institute of Technology, Atlanta [14] David M. Dozier and Ronald E. Rice, `Rival Theories of Electronic Newsreading' in The New Media: Communication, Research and Technology by Rice E R. and Associates (Beverley Hills: Sage Publications, 1984), pp103-128 [15] Ben Schneiderman,`Designing Information-abundant Web Sites: Issues and Recommendations' in International Journal of Human-Computer Studies, (July 1997, Special Issue on World Wide Web Usability): http://www.hbuk.co.uk/ap/ijhcs/webusability as at 12/4/98 [16] Jakob Nielsen, `The difference between Web Design and GUI Design' Alertbox, (May 1 1997) http://www.useit.com/alertbox/9705a.html as at 12/4/98 [17] H. Madsen, `Reclaim the Deadzone' Wired UK, 3, no1 (1997) pp 34-37 [18] Donald A. Norman, `Cognitive Engineering - Cognitive Science' in Interfacing Thought: Cognitive Aspects of Human-Computer Interaction ed John M. Carroll, (Cambridge, MA: MIT Press 1988), pp 325-336 [19] Donald A. Norman, `The Psychopathology of Everyday Things' in The Psychology of Everyday Things (New York: Basic Books, 1988) [20] G Lakoff and M Johnson, Metaphors We Live By (Chicago: University of Chicago Press, 1980) [21] Steve Turner, Analog, http://www..statslab.cam.ac.uk/~sret/analog.html (1997) [22] Alan Dix, Janet Finlay, Gregory Abowd, and Russell Beale, Human-Computer Interaction (New York: Prentice Hall, 1993) [23] Andrew Monk, Paul Wright, JeanneHaber, and Lora Davenport, Improving Your Human-Computer Interface: A Practical Technique (Hemel Hempsted: Prentice Hall, 1993) [24] Dan Diaper, 'Task Observation for Human-Computer Interaction' in Task Analysis for Human-Computer Interaction, ed Dan Diaper (Chichester: Ellis Horwood, 1989), pp210-236 [25] Roger Silverstone, Television and Everyday Life (London: Routledge, 1994)