HyperHyper'89: Developments Across the Field of Hypermedia

BCS Workshop, London, U.K., 23 February 1989
by Jakob Nielsen

Hyper Hyper was a nice small meeting with about 70 participants, mostly from the U.K. but with a few participants from Germany, Denmark, and the Netherlands as well as some Americans currently living in England. One of the organizers, Dianne Murray from the National Physical Laboratory introduced the workshop by quoting an article from InfoWorld which complained that every new product these days claim to have "hypertext-like" features. Actually I use one such product and am reasonably satisfied with it (Data Desk Professional which has hyperviews to connect related statistical analyses). I don't mind integrating hypertext ideas with other computational ideas. In fact, more hybrid interaction styles may achieve a better match with the way we work (a concept called task integration).

Cognitive Ergonomics of Hypertext

Patricia Wright from the MRC Applied Psychology Unit in Cambridge started by giving her definition of hypertext: It was not enough to have just nodes and links, navigation support was crucial. On the other hand, she did not feel that nonlinearity was necessarily a prerequisite for hypertext since many nested structures could give the user a hypertext feel. Her experiments with hypertext went all the way back to 1980 and a study of the British Library experimental BLEND system (described in papers by Brian Shackel and others). One of the things they discovered then was that people had problems when they had to return to previously read information and that these problems could be reduced by using several windows. This would allow readers to refer back to earlier information without leaving their current context. On the other hand, multi-window systems such as NoteCards implied the need for users to manage their windows since they would otherwise end up in the "I know it is here somewhere" situation.

Currently, Wright was working on hypertext access to information graphics which is an application very similar to the Data Desk hyperviews mentioned above. The basic problem is that for any given set of data, there are several different possible graphical representations (bar charts, single line charts, combined line charts, etc.) and that countless studies by now have shown that no one graphical representation is optimal for all possible tasks. So we want to be able to work with multiple representations of the same data and allow users to easily switch between them. This is a typical job for hypertext.

Wright showed slides with examples of alternative graphics for some data sets she was testing and to a first analysis they were not very impressive. Similar results could be achieved from allowing users to play with a modern spreadsheet with integrated graphing facilities where new representations can be chosen from a gallery of styles. My second analysis was however that the spreadsheet approach might be OK for expert users but that it is really much too difficult for many more casual users. I use Excel which is one of the better graphical spreadsheets but where the finer points about graphing (especially in combining several data series) are almost impossible to penetrate. A third point is that integrating hypertext access to alternative graphics with an intelligent system for graphical layout (such as that by Jock Mackinlay from PARC) would result in a truly valuable decision support tool. The intelligent layout system would not have to divine the user's need but could construct several good graphics and provide links between them.

After these examples of her more concrete work, Wright discussed hypertext in the context of more general cognitive issues such as her model of how users interact with information. To provide a mnemonic for the participants in this British Computer Society meeting, she referred to it as the BCS-model for Before, Concurrently, and Subsequently: Before reading a hypertext node, users have to decide whether, how, and where to jump. Concurrently with being at the node, they have to be able to understand the information and where they are. And Subsequently, they must be able to remember where they have been. Wright compared many contemporary hypertexts to the funhouse at a fair: To open a cupboard to see what pops out of you is fun, but it is not funny that most hypertext jumps have the same ability to surprise you.

In her conclusion, Wright said that the cognitive costs of using hypertext are not yet well known with respect to non-hypertext solutions or with respect to which of several alternative navigation styles would match users' tasks. She also feared that the cognitive demands on authors could be excessive as long as we do not have better tools. I can testify to this from my own experience-even a simple thing such as the list of anchors without destinations in HyperTies would have prevented a bug my system had for several months.

Lessons from Old Books

Approximately once a year we technofreaks deserve to hear somebody give a historical perspective on what could be achieved without computers. Robert Waller from the Open University did just that in his presentation on document design. He blasted the naive theories of traditional text which can be found in many papers on hypertext such that traditional text is purely linear and that only hypertext would allow links from graphics to text while traditional text can only use references in the direction from the text to the graphics. As a counter example he showed a double page spread from a manual on sailing which had information about related items scattered around the spread in a non-linear fashion and had call-outs from graphics to text.

Waller reminded us that many popular books from the so-called book packaging industry were planned as double-spreads and were in fact designed and layouted before they were written. These books are on many subjects such as cooking, gardening, etc., and even we computer people should own a few examples.

The history of writing was summarized by Waller as having had a trend towards accessibility and greater control by readers. Originally, text was just a stream of characters without spaces between the words and without punctuation (somewhat like written Japanese today). Typographical differentiation such as italics was invented in the 15th century and punctuation was codified in the 17th-18th century. A simple thing such as page numbers were introduced in the 16th century and were originally seen as aids to the printers even though they later enabled inventions like the table of contents and index. Many of these developments seem obvious to us now but took a long time to develop.

Waller warned that new media introduces cultural problems to people who are not used to them. As an example, in early cinema, people were used to having a stationary viewpoint from sitting in a theater audience so it took some time to develop the moving camera viewpoint and to understand how to utilize it.

Glasgow on 20 Disks a Day

Pat Baird from the University of Strathclyde reported on the development of the Glasgow Online hypertext with tourist information about Glasgow. They had had 31 people working on gathering all the information and converting it into HyperCard. The result was publicly available and free but it took two full boxes of double sided diskettes so the distribution will be quite limited in practice. Since the exploding disk requirement is a typical problem with hypermedia, I could wish that somebody would produce a CD-ROM with the demonstration projects currently being done around the world.

Baird stressed the need for continuous testing of the user interface and reported some results from a test at the Glasgow Garden Festival. Currently they are testing the system at the British Airways airport shuttle lounge and they also conducted a test during an international conference of mathematicians since most of the participants in that conference did not have English as their first language. The most general test result was that the system was easy to use. At the Garden Festival, many people who had observed the system without using it had been older people (51 years plus) while the group of people who had actually used the system had had a lot of young children.

Another test had been in a hotel where they found that 70% of the time of some of the reception staff was spent on answering questions where the answers could be found in the Glasgow Online information base.

Finally, Baird gave a demonstration of using the system to find a hotel to stay. At one point she said that "I always choose the Albany hotel when giving these demos - not because I have any special preference but because it is first on the list." This is exactly one of the major problems with this kind of information system which does not have an editorial content in the form of ratings of the services discussed. I can trust the Michelin guide to list the best hotel in town first, but Glasgow Online lists hotels starting with 'A' first.

Don't go into "Question Answering" Mode

The director of the University of London Audio-Visual Centre, David Clark, started by justifying his interest in videodisks. He showed a graph of the access time versus information capacity of various media, including books (do medium well), micro computers (good on access time, poor on capacity), and videodisks (good on both scales). In passing, Clark said that it was no wonder that Prestel (U.K. videotex system) had mostly been a failure since it "took forever and gave you nothing."

He said that most people who think that they have a wonderful picture database really mostly only have a database of words about the pictures. Only in very few cases do we have pictures entirely encoded such that users can explore and poke them in a hypertext way. As an example, Clark showed a medical system using a German video disc where users could click as various parts of the picture and get response from the system. The most important part of using such system for teaching was not to use multi-choice questions such as "Is the infected area A, B, or C?" since that will put students in a question answering mode where they study a closed world of pre-defined areas on the picture. Instead one should ask students to point to the part of the liver which seems to be infected so that they can explore an open-ended world.

MUCH Collaborative Writing

As the final scheduled lecture, Roy Rada and Judith Barlow of the University of Liverpool gave a collaborative presentation of project MUCH (Multiple Users Creating Hyperdocuments). This project was fairly complex and had ten graduate students working on different kinds of support for collaborative writing based on a semantic nets model. As Barlow said, if all of them carry through, the resulting system will be really great. They wanted to provide support for different kinds of authors who work in different ways and maybe even write their documents in different orders. Some work top-down and others write some of the bottom-level sections before starting on the outline. It would be a challenge for the project to provide support for collaborative writing even in the case where the authors who want to work together have different working styles.

Demo or Die

After the lectures, several people gave demonstrations of their hypertext systems. This included the managing director of OWL, Ian Ritchie showing Guide, myself showing the third release of my hypertext system built on top of HyperCard, and several demos based on the lectures. Nick Hammond from the University of York showed their home made guided tour system which now had slightly better human factored colors than when he presented it at CHI'88.

As part of the workshop proceedings, participants were given a disk with summaries of the lectures and demonstrations in both Guide and HyperCard format. Unfortunately the HyperCard version had lost some of its hypertext qualities because the link between the stacks with presentation abstracts and speaker addresses was just a generic link back and forth between two stacks without going to the specific card with info related to the departure point of the jump. Of course having so many hypertext fanatics present in the same room generated a lot of discussion, including the traditional question of which of the commercial systems is "best." This last discussion continued over Australian wine in a wine bar next to the Charing Cross Station. I don't think we reached an agreement.