Three days of tags, semantics and hyperlinks: Hypertext 2009

From a designer’s perspective, Hypertext – the annual conference of ACM’s Special Interest Group on Hypertext, Hypermedia and the Web (SigWeb) – is interesting for having its roots in computer science, but promoting a broad spectrum of topics with a deliberate multi-disciplinary approach. Hypertext 2009 was held at the Northern Italian city of Turin in the end of June.

While some of the presentations were about very detailed aspects of computing algorithms and the perceived majority of research was quantitative computer science, there was lots to take away for the social scientist and designer; not to mention numerous interesting conversations that revealed how even some of the most technical research presented had a strong social or philosophical background.

This report is the attempt to condense three intense days of workshops and presentation sessions. I herein summarize selected sessions and papers I perceived particularly inspiring, from my point of view and highlighting what I considered the most interesting. Naturally, I had pre-selected all sessions based on my fields of interest, so the main topics identified here are subjective and not necessarily in synch with the structure of the conference – the official division of the conference tracks can be reviewed from the program website. Also, a computer scientist would likely address very different content as the conference highlights and I want to explicitly suggest the advanced reader to refer to the original conference proceedings instead.

In retrospective, the themes of the three-day conference in Turin evolved around two major fields of research: (A) the role of folksonomies in the semantic social web for one and (B) the state, use and changing paradigms of hypertext-based information structures for the other. Naturally, these themes partially overlapped. For this report, I split my observations into four sections, grouped according to the core topics I identified:

1. SW² – the social web as a semantic web: searching for the semantic value of social web content, with a special focus on tags/folksonomies,
2. Social hyperlinks and live social semantics: exploring the “social semantics” of users in the social web,
3. Social search: looking for new ways to provide the search user with the right results by utilizing the semantics of both content and social relations, and
4. Users and usage of hypertext: insight into some of the applications of hypertext and related disputes.

As a trivial observation, one of the most obvious differences between Hypertext 2009 and the design conferences I attended within the last year was what one may call the “Mac credibility factor”: While it appears to be part of the unwritten rules of design conferences to carry an Apple computer, an iPhone and at least two more accessories with the distinctive logo, Mac users were a clear minority in Turin. An impressive share of attendants used Linux. Regardless of the operating system (which, as often pointed out, I consider the least important variable regarding productivity and creativity), we all shared the problem of the limited amount of power outlets around Villa Gualino, the otherwise great conference venue high above the river Po. The WLAN however was highly reliable, and so was the capacity of the catering to provide one culinary experience after the other.

1. SW² – the social web as a semantic web

A major topic of Hypertext 2009 was the utilization of implicit information contained in folksonomies/tags. About half of the presentations I attended discussed topics related to tagging in one way or the other. The debate at the conference was around the topics of how folksonomies emerge as well as what is their semantic value and meaning. Presenters discussed how this form of metadata can be analysed and applied for the benefit of semantic information structures (and the benefit of the users, respectively).

1.1 Tagging as a process

One topic was the process of tagging itself. The research presented at Hypertext analysed the process of how users add metadata to web resources and how this process affects the value of that data.

Tagging dynamics

In the opening presentation of the “Web 3.0″ workshop, “An introduction and an Overview: 3 years of tagging”, Vittorio Loreto (ISI, University “Sapienza” of Rome, Italy) outlined some of the underlying basics of tagging. Being the first session of the opening day, the overall message of this presentation was like a basic definition for lots of other topics to come. The use of folksonomies by users was described as “human computing”: the solving of a difficult problem with small cognitive overhead. By this definition, user-driven tagging is a collective activity; an uncoordinated distributed process generating meaning far beyond the process of attaching a keyword to a piece of content.

Collective dynamics of social annotation

Andrea Baldassarri gave a talk on folksonomies as complex web systems titled “Minimal Modelling of Folksonomies”, introducing the University of Rome’s framework for mathematically modelling aspects of social annotation. The paper defines social annotation as a “collaborative exploration of a semantic space, modelled as a graph” that “expose[s] aspects of semantics and of human dynamics”. He employed the “Flickr user model” visualization by Bryce Glass as an example of how complex the ecosystem of a folksonomy can be:

The “tripartite structure of folksonomy” referred to in Baldassarri’s presentation was another essential definition that reoccured throughout the conference. In their paper the authors describe it as “a central collective artefact of the information society” where folksonomy structures consist of three elements: users, metadata/tags and resources.

In the discussion, critique was formulated that tags might well be overestimated: how do we know that they are not just names with no semantic value? This question was part of many of the papers presented, and therefore returned into discussion repeatedly throughout the conference.

Why do users apply tags?

One poster that caught my attention on the first day already and was later rewarded the Student Research Competition award was “The Motivation behind Tagging” (poster in PDF format) by Christian Koerner, who identified two types of tagging behaviour and researches methods to discover the tagging type of a user on a website – ultimately an important asset in creating better engines to exploit folksonomies.

• Categorisers: Using tags to organise their resources for later findability. They use few tags, which are applied repeatedly.
• Describers: Establishing their own vocabulary by describing the content of the resource based on its content, not a limited amount of categories. These users are identified by a rapidly growing vocabulary with a low rate of re-use.

A related topic was covered by the presentation “Individual and social behaviour in tagging systems” (Santos-Neto, Condon, Andrade, Iamnitchi and Ripeanu), stating that tagging motivation is mainly personal information management, not collaboration and that user populations self-organize into clusters of interest. I believe this is a very important consideration, as to me the weakness of many Web 2.0 concepts appears to be the general assumption that users are tagging things for the purpose of collaboration while the real reason might be far less altruistic.

1.2 Extracting meaning from folksonomies

Based on the proposition that folksonomies include semantically valuable meaning – even though they are created by individuals for their own purposes – a lot of presentations discussed approaches and algorithms on the extraction of that data. The research questions mainly evolved around two major challenges; how to process data from the social web to turn it into valuable semantic data and how to connect the otherwise unrelated folksonomies of several web services for semantic exploitation.

Merging the Semantic Web and the Social Web

During the workshop “Merging Semantic Web & Social Web (SW²)”, both chair and audience pointed out that, interestingly, most of the papers discussed the same problem: how to turn folksonomies into ontologies. In the light of the research presented, “Web 3.0″ was eventually defined as:

• collaborative generation, “discovery”, management and population of ontology,
• automatic, user-friendly annotation of text,
• improvement of tagging with semantics, and
• use of tags to improve the intelligent behaviour of systems (recommendation)

Two of the papers in the workshop presented approaches to involving the user in self-organizing their own tags into ontologies. However, it has been discussed whether users are willing to do it, despite of the availability of the required tools? While adding a simple tag is a small investment, incentives have to be provided to do more. The power of the social web lies in leveraging the “wisdom of crowds” (plain text, tagging, folksonomies, collaborative filtering, social search), but its’ problems are that it is noisy, suffers from term ambiguity and lacks structural depth and reasoning capabilities.

In their paper “Working the Crowd: Design Principles and Early Lessons from the Social-Semantic Web”, Niepert, Buckner and Allen stated that the semantic web, based on formal conceptualizations of a domain (ontologies, taxonomies…) and applied successfully for instance in biomedical and business domains (e.g. the gene ontology), provides reasoning and matching capabilities, interoperability and enables semantic search. However, its’ problems are the need for double experts, lack of user participation and that it is too complex. Simply put: the more structured the content is, the lower the willingness to contribute.

Cross-Folksonomy Analysis

Martin Szomszor (University of Southampton, UK) addressed the aspect of user profiles and tagging in his paper “Modelling Users’ Profiles and Interests based on Cross-Folksonomy Analysis” (slides download). Quoting the Ofcom study “Social Networking – A quantitative and qualitative research report into attitudes, behaviours and use” he presented that UK adults have in average 1.6 profiles, in other words 39% of those having one profile have at least another one. In future, people will maintain an increasing number of online identities.

By tagging, people create an overall “profile of interests”:

• profiles could be exported to other sites to improve recommendation quality;
• profiles allow designing better user experience;
• profiles could be user to support personal searching.

By means of consolidation and integration, users can be provided with a coordinated view of all the tagged data. There are, however, challenges connected to this aspect. Tagging variations are the biggest problem found in investigations (as Szomszor presented in his Hypertext 2008 paper “Correlating User Profiles from Multiple Folksonomies”). Identical tags do not always mean the same, and different tags are used to describe the same concept. Therefore, there is a need to transform raw tags into filtered tags, eliminating spelling variations, typos etc. Disconnected identities are another challenge. The problem is ending up with a lot of isolated profiles.

Making sense of folksonomies is possible by creating a set of tagging semantics, in this approach by integrating contacts and tags. When looking for tag senses the main question is: what are the possible meanings for a tag? The research group at Southampton uses two reference sets: DBPedia for concepts and WordNet for synsets (TAGora Sense Repository TSR).

In the discussion, the question was posed how to mine data to match user profiles to one person. This obviously is a challenge, but can often be done by reverse analysing links from the profiles to other resources. Ultimately, cross-folksonomy allows building better profiles. This requires:

• knowledge on what tags correspond to interests (locations and topics are useful, but other terms are not),
• an approach that eliminates the obvious (it’s not that useful to find out all attendants of Hypertext 2009 are interested in HTML), and
• making use of the category hierarchy (as an example: the interests “facebook”, “flickr” and “lastfm” express the category interest “Online_Social_Networks”).

1.4 Folksonomies and their use

Some folksonomy-related papers were particularly interesting from a practitioner’s point of view, discussing the application of tags and its practical implications.

Tag recommendation

Andreas Hotho (University of Kassel, Germany) gave an introduction to tag recommendation techniques, “Tag Recommendation in Theory and Practice”. As a simple baseline for tag recommendation, he defined the use of an existing tag cloud related to the URL. In their research of collaborative filtering, datasets from delicious.com, last.fm and the social bookmark and publication sharing system bibsonomy.org (run by the University of Kassel) are used to calculate personalized tag recommendations.

In the discussion, a critical question was expressed whether it is even good or desirable to recommend? The commenter stated that users are conformist, therefore tag recommendation would lead to less variety in tagging. Summarizing the presentation and discussion, tag recommendations are an interesting and important topic for social bookmarking systems, though their use has to be carefully considered to avoid negative impact on the quality of the folksonomy.

Tagging in real life

Barteb Ochab (SONY-CSL) presented an approach called “the CSL approach”, applying the insight into online tagging systems to real-world environments. His presentation “Tagging Usages in the Reality” (slides) showed, how tagging principles as used on the web can be applied also to real-world context where they work in a similar way:

He presented Sony’s project “Noise tube” where mobile phones are used as environmental sensors. The idea is to overcome the challenge of collecting data in a real environment, currently relying on rather poor monitoring systems. The idea of the project is to monitor noise levels and tag the sources of exposure, as with current solutions it is difficult to find and separate the sources of noise from recorded sounds; modelled noise maps are limited to the assumed sources of noise. The proposed solution is to tag sources of exposure: this way people are utilized as sensors, adding layers of meaning to a measurement. This is based on the assumption that users are motivated to inform the community about sources of annoying noise.

Another project presented was “Linking Linke”, an interactive installation to explore tagging in the physical world. As a controlled experiment, exhibition visitors are asked to arrange high quality photo prints on a table that are then printed as a book. After picking 8 photos on a topic, the user has to arrange them and give the book a title, therefore “tagging” these images.

2. Social hyperlinks and social semantics

But Hypertext 2009 did cover the topic of semantics not only from a content perspective. Looking at social web services, this kind of applications is basically pure semantics: social semantics. The theme was omnipresent throughout the conference, thanks to a live experiment on tagging its participants. In addition, Lada Adamic’s keynote “The Social Hyperlink” was one of my personal highlights of Hypertext 2009.

2.1 Live social semantics

During registration for the conference in the morning of day one, we did not only receive the obligatory name tag and the usual bag filled with programme, sponsors’ ads and note-taking paper, but also a small RFID chip from sociopatterns.org to be carried in the name tag holder. A set of RFID receivers at the venue (in the conference rooms, the lobby and the cafeteria) could now track the attendees’ movements and face-to-face interactions. These movements were analysed in real time – allowing for interesting conversations when standing in front of the visualisation screens where proximity indicators almost allowed to identify the otherwise unknown strangers standing in the vicinity. The more hidden aspect of the experiment lies in the web profiles each of the users had previously filled into a web form – this is best explained by the following video clip:

In addition to this explanatory video, the research group also wrote a blog post on the experiment. It was interesting to experience live social semantics, and in addition we all contributed valuable data to the ongoing research “on the go”.

2.2 Social hyperlinks

“The Social Hyperlink” was the title of Lada Adamic’s keynote on day 2 (slides), an excursion into different fields of network analysis all based on data on social hyperlinks. She started with some anecdotes about differences in real-life social networking between different US universities, using examples to emphasize the difference between causation (similarities in behaviour based on a social relation) and correlation (similarities that are not related to a relationship).

The introduction led to the first out of three social hyperlinks presented, the “social influence hyperlink”. The talk referred to Second Life, where research on the exchange of virtual objects has revealed that transfers within friend networks are deeper (i.e. objects were handed on more often) but also that early adopters are not really influencers: they often have less contacts and are significantly less active in sharing than influencer type users.

The “knowledge exchange hyperlink”, the second link explored, refers to research carried out on the motivation and characteristics of people sharing knowledge online. The information exchange between individuals serves as the research object here, for example identifying that the number of answers from a particular user generally indicates better expertise, but also that incentives for answering ensures more answers but less attention (due to the fact that incentivised questions are often more difficult). The motivation to engage in online knowledge exchange can be manifold: for example a genuine aim to help, “learning by answering” or trying to promote the user’s business. Also, it has been researched that people appear to have a significant need to complete incomplete information which again leads to the fact that the last answer in a thread is often considered the best.

Thirdly, the talk referred to the global hospitality network Couchsurfing, talking about the “trust hyperlink”. Research shows that Couchsurfing is not based on direct reciprocity (hosting the same people who have hosted you before) but on “generalised reciprocity”, with people increasingly vouching for people they only know through Couchsurfing. While in the beginning, there is an almost 50/50 divide between people who join Couchsurfing to host and those who go “surfing”, over time both user groups tend to engage in both activities.

3. Social search

In the first two chapters of this report, I first summarized the discussion around general aspects of folksonomies and then referred to a keynote and an experiment around social semantics. These topics can be merged into the third “theme” I identified at Hypertext 2009: social search. The combination of semantic content and the semantic representation of social relationships allows for a broad variety of applications for search, discovery and analysis.

3.1 “Social” adds value to search

The paper “Social Search and Discovery using a Unified Approach” (by Einat Amitay, David Carmel, Nadav Har’El, Aya Soffer, Nadav Golbandi, Shila Ofek-Koifman and Sivan Yogev) presented an approach to social search as applied on the IBM intranet, to support employees in their daily work.

Among other services, the IBM intranet features these services:

• “Blue pages” people index: 475k profiles
• “Community map”: 28k communities
• “Blog central” internal weblog service: 40k weblogs by 70k people
• “Dogear”: Bookmarking system with 330k tagged links
• IBM’s internal social networks (“Fringe”, “Beehive”, “Sonar”) used by 100k employees

Unified search is conceptually based on a triangle of documents, people and metadata. The search space is extended to represent many information objects that may be related to each other. Some people like to find documents, others prefer finding people or a hierarchy of labels; still, most people are between extremes and search some kind of combination of those. The related work presented included some papers on penalizing for too frequent objects (in the IBM context for example the useless tag “ibm”) and the normalization of “serial sharers”.

The problem with the unified approach is that computation is expensive and not scalable; besides, the rate of creation and update is expected to be very high. The solution presented was “faceted search”, a commonly used technique for adding navigation to a search engine’s results. Hereby, any searchable entity is represented as a pseudo document with direct relations between entities and indirect relations computed during the request on the fly.

In the discussion, the question was raised how to handle the fact that some people only do real-world communication. We already have data on virtual relationships that can be weighted, but getting these non-virtual contacts would be important to get. Also, it was brought up that the complexity of social search is due to the fact that it is not a simple triangle, but several interconnected triangles of groups, persons, documents and tags. It has to be taken into account what entities this particular person has and how these entities relate to each other and other people.

3.2 Web usage as metadata for search

Ricardo Baeza-Yates’ from Yahoo gave an interesting keynote on “Relating content by web usage”. In his talk, the vice president of Yahoo! research explained how search is no longer about a user wanting to retrieve a document, but about task fulfilment. The challenge for delivering optimal search results (and search experience) is in identifying the user’s task and providing the means the user needs to complete it.

Besides explicit metadata that is assigned to content, a lot of implicit such as clickthrough patterns etc. can be gathered to gather a better picture on the nature of content. Even though this kind of user-generated metadata contains a lot of noise, it is a valuable source of data in its entirety. The presentation contained two example projects from Yahoo! that make use of metadata for search, TagExplorer and Correlator.

“When users use the web, they think” – data generated through web use (such as logs and UGC) is therefore a highly valuable source of information on content that should be used for achieving best search results. The same applies to social networks, not only those where users explicitly create links between each other, but also implicit social networks that can be discovered by analysing web use.

3.3 User content is not user interest

Sergej Sizov and Stefan Siersdorfer’s paper “Social recommender systems” presented methods for systems that make recommendations to users based on their social graph. The motivation for the research was that “web 2.0″ services have a huge amount of relatively sparse data that is often noisy, even spammy, and there are no natural mechanisms for reliability or trust. Thanks to the large pool of users there is however a huge amount of data that can be used for recommendation: favourites, groups, contact list and comments. The important learning is that the content of a user does not always correlate to her interests. In order to properly serve the interests of a user, the understanding should be based not on her own shared content but on her social environment and expressions of interest (see above, considering web usage data as metadata for search).

4. Users and usage of hypertext

The fourth topic of my report from Hypertext 2009 differs from the previous three chapters, that were all around the themes of folksonomies and semantics. This final section combines a series of papers whose focus was more on hypertext and its use as such. In other words, these presentations were not about tags and implicit metadata but about HTML (and its alternatives) and how they are used.

4.1 Hypertext as a tool for storytelling

Mark Bernstein presented a session on hypertext as a mean for narrative expression, titled “On Hypertext Narrative”. Narrative has three layers: story, plot and presentation. The story tells “what happened”, the plot describes the sequence in which we explain what happened and the presentation is what we see on the page on the screen.

For narrative in hypertext, the motivation is that “we want to do things we couldn’t do in print”. Bernstein presented two propositions related to this.

Hypertextuality is perceived through re-reading and reflection

From the three layers described above, plot, not story, is where we find meaning. The tone, the pacing, the point of view are what makes meaning. For instance, when thinking about the landing of the allied forces in the Normandy: numerous books have been written telling the story, but the plot is what makes them differ. However, plot is not a surface detail; it makes a major difference when do we tell the reader that the wolf has run ahead and eaten grandma?

• Telling it early: tragedy
• Telling it while in the wood: horror movie, melodrama
• Telling it at the last moment: comedy, romance
• Telling it afterwards: …

Bernstein referred to the paper “Nonce upon some times” by Michael Joyce where a narrative is told of a man and woman who fall in love. He is rich, she cheats him for his money and disappears. Some day, they meet again. Here we’d have four possibilities to continue the plot: recursion, time shift, renewal or annotation. The main message is: we want to vary plot.

Stretchtext to the rescue

The cure that Bernstein proposes for navigation is stretchtext. Ted Nelson has seen stretchtext as a (or “the”) natural form of hypertext from the beginning. The concept presented was called “generalized stretchtext”, applied in the software Tinderbox.

According to Bernstein, our “business” is varying plot, not story:

• text stays itself, electronic text replaces itself;
• stretchtext as traditionally applied keeps us from varying plot;
• general stretchtext lets us vary plot without arrivals and departures.

In the discussion, a comment was made that text changed dynamically at the end of a page (i.e. clicking a link on a page that affects a piece of text at a different position of the page) is confusing. Also, it was brought up that navigational hypertext is hard to read, to which Bernstein answered that this is mainly because it is work of the 80s and 90s – which is generally hard to read.

4.2 The personal territories of web users

The presentation “The Dynamics of Personal Territories on the Web” by Thomas Beauvisage (Orange Labs, SENSE lab), later awarded the Ted Nelson newcomer award of the conference, was about getting to know how people use the web. The research was based on a long-time data set about the web use behaviour of users and allowed some interesting insights to the evolution of browsing since 2001: the amount of time spent browsing has increased significantly, and a concentration and stabilisation of practices can be observed (routine and seasonal sites up, transient sites and ephemeral sites down).

The research divided browsing sessions into five groups of “session profiles” and showed how the last years have seen an increasingly clear divide between simple “routine” sessions (often very linear and with only very few sites) and long and complex sessions. For longer sessions, the complexity of the navigation around certain hub sites has increased.

Regarding personal territories, the web as it is used by a particular user, it can be observed how the majority of time is spent on a small amount of regularly visited pages whereas the majority of sites visited are one-time visits and usually for a very short time. Between 2001 and 2006, the number of routinely used pages had more than doubled and they took an even bigger share of the total web use time. However, also the number of “ephemeral sites” (aforementioned one-time visits) was quite stable, which shows that exploration is still an important part of user behaviour – regardless of the strong position of routine sites.

4.3 Weblogs

One session during the conference was about the topic of weblogs. An archive-based research investigated the emergence of the weblog as a writing space and two papers investigated use-related topics.

Blogs – where it all started

Rudolf Ammann reconstructed the emergence of the weblog community in his paper with the title “Jorn Barger, the Newspage Network and the Emergence of the Weblog Community” (weblog entry). It was in January 1997 that the company UserLand released what can be considered the first weblog software: “NewsPage”, a feature in a publishing platform that allowed publishing a news stream. Dave Winer, the owner of UserLand, updated the website scripting.com to run NewsPage and the first weblog was born. Other early adopters of NewsPage started to refer to Winer’s page and he linked back to them – and it was not long until Chris Gulker set up a “NewsPage network” (today we would call it a “blogroll”) to cross-refer to ever more NewsPage sites.

A little later, Jorn Barger joined the NewsPage Network and attempted to define the concept of the weblog as a space where content is vacuumed from the web and promoted in a clearly defined format, without extensive commentary. In Barger’s vision, weblogs were an access structure. The initially small “weblog community” eventually grew bigger, and at some point the concept emerged further into the writing space we know today – which led to Barger’s withdrawal.

The presenter discussed Barger’s role in the community and underlined the importance of the “weblog community” for the history of network theory on the web. The question what led to the transformation of the weblog into a writing space remains unanswered and, according to the author, should be investigated as well.

Blogging cultures

Thomas Mandl’s presentation “Comparing Chinese and German Blogs” opened with the question “Do you do research on blogs, social networks and social media or is your research on the fraction of English blogs?”. Despite its title, this presentation was more about culture than about weblogs, however the latter being the object of the research. Mandl took the audience through a set of theories on culture.

In result, Mandl concluded that Asian blogs are more versatile in graphic design, making use of many more design elements; this is supported by the blog platforms offering more layout options. In another finding, Germans assign more tags to a blog. Regarding the dimension of high vs. low context (as in the dimensions by Hall), German culture is more a culture of high context: Chinese blog texts tend to be shorter and Germans use more external links. In terms of reactions, blog posts are significantly more often reacted on after more than one week from the authoring date, and the culture of “keeping face” leads to less negative reactions in the Chinese set.

The diary of a researcher

The talk “Weblog as a personal thinking space” by Lilia Efimova was an autoethnographical study in which the researcher analysed her own blog. It was based on her observations while working on her Ph.D. thesis “Passion at work: blogging practices of knowledge workers”. She described the weblog as a knowledge space, like the papers on a person’s desk; the different functions of the weblog as a personal space are creating, organising, maintaining and retrieving knowledge. Her weblog was to her like a trusted environment to park random thoughts, drafts and notes.

4.4 Bringing dead links back to life

The research of Atsuyuki Morishima, Akiyoshi Nakamizo, Toshinari Iida, Shigeo Sugimoto and Hiroyuki Kitagawa (titled “Bringing Your Dead Links Back to Life: A Comprehensive Approach and Lessons Learned”) is motivated by the fact that meeting broken links is a disappointing experience. The presenter quoted GVU’s 10th WWW user survey from 1998, when lack of speed, broken links and slow ads were the three major issues for web users. While the speed issue has been solved today by broadband access and AdBlock and other solutions have been invented for the ad slowness issue, no toll has been invented to solve the problem of broken links.

Their research is an experimental study of the automatic correction of broken web links. It is focusing, in particular, on links broken by the relocation of web pages . For the solution, location factors play an important role, so it is possible to find new locations without an exhaustive search of the entire web. In the experiment, the researchers achieved a detection rate of more than 70%, much better than those of Google and other index servers.

4.5 HTML vs. Xanadu

The keynote by Ted Nelson was the opening to an entire afternoon dedicated to xanalogical data structures. As myself, quite a few people attended only the keynote and then moved to the other track’s session – most likely due to the fact that the topic of this track was otherwise very specific. But it was impressive to see that there is quite a community around Xanadu, and listening to IT pioneer Nelson’s speech was definitely another highlight of the conference for me.

According to Nelson, “the history of technology is a misunderstanding of human thought and life “, considering technology as a force to be obeyed:

1. Ted’s fight with technology began at age 9: in the 1940s his parents owned a pressure cooker that was designed to require releasing the steam before opening it. His father got seriously hurt because he forgot to open the valve. This is a misunderstanding of human thought: the design was based on the assumption that everybody always remembers to release the steam.
2. The hierarchical organization of information is an example for the misunderstanding of human life. Ted: “I am spending hours on the WWW every day and I hate it very much”. He considers social web sites “cattle pens driving animals down the shoot to slaughter”, for example the “perfectly ridiculous activity” in Facebook that one has to repeatedly answer the question “are you my friend or not?”.

The underlying misconception regarding today’s WWW – which Nelson compares to “paper under glass” – is that electronic documents represent paper documents and that all you can show is what can be shown on a paper. But the WWW is even less than paper as with real paper you can write on it or put sticky notes on it.

“Once you have a hammer, everything looks like a nail; once you have hierarchies and lump files, everything can be a file”, says Nelson and points out that the design of interactive software is not computer science, but movie making – a branch of presentational art. The computer, correspondingly, is to be considered a movie machine and a philosophy machine (defining philosophy as “designing an idea “).

Stating that “the web browser was explicitly designed to prevent everything i want to do”, Nelson presents the initial idea of xanalogical hypertext. The Xanagram shows a document as transclusions and links. Calling his concept “oceanic hypertext”, Nelson’s vision is a system that enables sharing from all over, with the right for re-use. The right to quote anything is a win-win environment for micro-sale. It also is not oppressive, but the way out to honestly participate rather than having little chunks, pointing in one direction .

What I took home from Hypertext 2009

When I signed up for Hypertext 2009, I knew this would be a very scientific conference in a field that is not my own. Still, I was attracted by the idea to go listen to what computer science is currently working on in the field of “the web”, and this ACM SigWeb conference has been a very positive experience. While the level and aim of some of the papers were slightly beyond my understanding and interest, the conference provided a lot of inspirational thoughts, insight into very specific fields of research and I had very good conversations with the other attendees.

While a global army of designers repeatedly praises tags and semantics as a power tool for many purposes, it was very eye-opening to dive into the full complexity of the topic of folksonomies and discover problems and opportunities that had never crossed my mind before. As a sociologist, I was particularly fascinated by the research related to network theory and network analysis – as well as the projects merging the social web with the semantic web. A series of sessions related to the application and use of hypertext and semantics was probably among the most fruitful ones for my work as a designer.

To sum it up: the biggest benefit of this conference for a social interaction designer was the possibility to inhale a lot of fact-based knowledge on hyperlinked information and communication structures, combined with a lot of historical background (Xanadu, blog networks etc.) and critical thoughts whether everything “semantic” and “social” is really such an easy and simple tool to make people (users?) happy. I will not be able to attend Hypertext 2010 in Toronto, but I for sure will closely follow the coverage on that conference.

Reports by other participants:

• Hypertext 2009: hypertext in the wild! by Markus Strohmaier