Passing around a copy of the New York Times the class highlights what in the paper is metadata. David Weinberger, who is leading the class, wonders why no one chose to highlight the headline as metadata, which boldly proclaims Spitzer’s indiscretions (John Palfrey suggests it was perhaps too seedy for us!). The problem is that the headline can be data itself, we even have headine news. But the headline is also imparting information about the information in the article which makes is a borderline case, it tells you about the article only if you choose to read it. Is the font size information? Metadata? Yes, it tells you in all caps this guy screwed up big. Placement on the page is also metadata. So the newspaper itself is metadata, even the difference between NYT online and the paper version in terms of space. The fact that something appears in the print version gives us information about the article because there is limited space (claims of all the news that’s fit to print notwithstanding). Does the space between words tell us something? (other than the dominance of oppressive mainstream gramatical structures!) Spaces are metadata because they show you the end of the information you care about, you are told this is the end of the word.

Existential Crisis Alert after the jump

It was suggested that words themselves can be seen as metadata, they are conceptual tools that tell us about something else as does our choice in using them. Could you argue that everything encoded with words is metadata? Yes, but for now we want a class of information that is metadata so we wont take the bait. The problem is framed as: does one think everything is information and relates to everything else, which means it becomes impossible to differentiate information from information about information.

Organising information and metadata pre-web

Melvil Dewey – the Dewey Decimal system inventor, also founded societies to promote simple spelling, the metric system, and shorthand. Came from a tiny school in Massachusetts, had limited experience of the world. Decided as a Senior to organise the world’s knowledge, using a conception of information/knowledge in the tradition of Hagel etc. Went with 10 top categories, then 10 within each, then 10 more. See the classification system here

What guides the order of the categories? Vaguely from something conceptual/ethereal to more practical. Follows philosophers ordering of knowledge, which means philosophy is the top rank. The entire system is an ordered list, the top always indicating the most importance.

The third most important topic is paranormal in Philospophy and psychology! Religon is all Christian! It has been updated, Islam now includes Bahaism and more. Where is Budism? 294.6, didn’t make it to the left of the period, only a billion budists, but Dewey didn’t have many books about budism, he lead a cloistered life. Why hasn’t this been fixed? Mainly because it would be just too embarrassing. Did fix the computer stuff by putting it into the zeros. If they did make an update, think of all the problems: the ordering of Shiite and Sunni, where do Jews for Jesus go, is Palestine as a country, gender stuff (women’s education but not men’s). There doesn’t appear to be a way to fix Dewey. This comes up everytime you have a taxonomy.When you decide how to divide things up your hand is forced.

Organising information on the Web

Amazon gives you an unbelievable set of information about a book. Even SIPs, Statistically Improbable Phrases, that you can search by other books for. User generated metadata. Amazon has multiple categories. An endless amount of metadata. What does it tell us? Sometimes it helps us decide whether we want the book, a SIP might give us a reason to buy from Amazon rather than someone else, maybe it helps us to look for other books. The metadata ties things together that otherwise we never would.

JP poses the question:

If the idea is that by using metadata we put down markers that might be useful, might we go to far and have too much information

Compare Amazon to an old style library card system. Differences with a library card: less space, limited time, rules of how to structure, static rather than dynamic. Amazon can change in time, anybody can change parts of it. Searchability suffers. Ability to sort suffers. Can’t have multiple copies really. No structure to adding metadata. Social life of information says dog eared is interesting. Most effort given over to excluding information.

Recap: systems that rely on taxonomy are fixed systems and tend to be very rigid. This is in a sense mainly because when we work from paper we inevitably lose flexibility.

Taging

Go to Flickr, where you can search on tags provided by other people. There are also groups that a picture can be added to, such as one for pictures of noses. People can even tag other people’s photos depending on settings. What about when there is too much metadata broken out of a taxonomy. Compare Flickr with Corbis which has a 70,000 word taxonomy that has synonyms, a controlled vocabulary. Flickr has an open ended system whereby any tag can be thought up by a user on the spot, no matter how idiosyncratic. Flickr can create clusters of images, automatically generated based only on the tags. This is an analysis of multiple tags. With enough tags you can create order.

Wikia – the search engine launched by the founder of Wikipedia which is a combination of human and machine. The web difference is that you can combine the human inputs with the technology on top. The web means you can update taxonomy much more quickly: the Library of Congress has a flexible taxonomy and will make extra categories when it needs them. When you move online there is no limit to the number of categories that you can create, because you can use faceted structures to organise data it doesn’t matter how many new categories you add to the taxonomy. The LoC has more stuff than it can categorise, 150 million objects, training people in a taxonomy doesn’t scale. LoC put 3000 photos from the 40s into Flickr in order to jump start the metadata. They put the metadata they already had in and within days users had filled up the tag list. Ran out of tags on Flickr at 75, people hacked it. Can put a tag on the picture annotating it. The comments section has lots of comments.

What do we gain and what to we lose from this system?

In Google Books we can search by “call me ishmael”. We get the book back, is it metadata? Someone argues that it is not, the search that leads to it is the metadata. Would Herman Melvil be metadata? Depends on the context – when it gives you the information that he is the author. DW says everything is, is it the information age that makes it so or is it the frequency of use that makes it so? DW suggests it has been useful to have a strict distinction between data and metadata, because we have been stuck organising the real things, when we get too many things we need to separate the data and metadata. We had to reduce the amount of information. When we went into the digital world we started to replicate with databases, but now we realise that anything can become metadata – anything can functionally become metadata. The importance of this is that we just got much better and finding things and will be much better at it – a huge web difference.

Today’s topic was “Knowing on the Web.” For the first part of class, we delved into the murky world of philosophy, talking about the history of knowledge and Descartes’ Meditations. We finished by examining several controversial Wikipedia entries and the accompanying discussion threads.

Perfection and the Web
David Weinberger (DW) began class by bringing up the perfection-related idea we talked about last time: that the Web is always a little “broken.” DW thinks this is important because we tend to think systems “work” when they approach perfection: the more perfect they are, the better they are. But this is actually a trick we play on ourselves by defining what is “broken.” We could interpret a busy signal or 404 page as a sign that the system is broken, but we don’t do that. DW suggests that much of what’s good about the Web comes from the fact that it doesn’t even pretend to aim for “perfection.” We could have easily built into the architecture of the Web a mechanism by which broken links don’t show up as hyperlinks. But the cost of reworking that architecture would be so high that, in attaining a more perfect Web, we would end up with a very different Web.

DW then asked us to come up with other examples of things that can be taken as “imperfections” on the Web — things that are not controlled, but could have been. Tom pointed to errors in newspaper articles that we know are mistaken, but we leave up as a historical record of the error. This led us to wonder: how often do bloggers take down or edit posts they later learn are mistaken?

In an informal poll of bloggers in the class, we found out that sometimes bloggers go back and change mistaken posts, sometimes not. It may depend on whether the error is content-related, or grammatical. Things like getting names or genders wrong can be embarrassing, and DW and others often just fix those. But the students (i.e. Kevin) who don’t strike mistaken posts said they don’t do it to make sure readers don’t “think they’re going crazy.” DW said he sometimes leaves visible errors in case people try to link to it, and cited the craziness factor (i.e. “I could have sworn Chris was a guy!”). This is certainly an example of leaving the Web imperfect because it’s better that way. Another example of that, as Justin reminded us, was the “flaw” in MySpace code that allowed MySpacers to design their own pages in html.

Cutting against this idea that leaving “mistakes” online is a good thing, Dorcas noted that most people don’t bother to remove the mistaken or outdated parts. This can mislead people who get to that page thinking it’s fresh. DW pointed out that there would be ways to minimize this, but that in doing so we’d lose a lot of valuable info.

Another student questioned whether this “mistakes” thing is really a web-specific difference, and argued that these things happen with print media too. I argued the other side – I do think there’s a web difference here. Because it’s easier to make corrections, people online have a greater expectation that corrections will be made. (Although perhaps people often expect mistakes online in a way they don’t with traditional media – as DW pointed out, making it easier to publish also makes it easier to publish mistakes.)

This part of the discussion concluded with most of us recognizing that systems make decisions about the proper balance of control and accuracy. CNN makes these kinds of decisions all the time: they may “get it wrong,” but they can go back later and correct. Typically, though, CNN will want more control than, say, DW wants to have over his blog.

Knowledge: history and meaning
We then took up the question of knowledge: what is it really, and where does it come from?

(DW got pretty excited about using his “visual aid,” which led to a brief but animated discussion about various outdated media like overhead projectors.)

He showed us a USA Today crossword puzzle that he himself (and not a research assistant!) had completed – much to John Palfrey’s surprise. (Perhaps law school professors have become too reliant on RAs?) Anyway, the right answer to the “have no doubts” clue was “know.” But DW pointed out that knowledge didn’t start out that way – as being about certainty. It began as a practical distinction ancient Greeks had to make in everyday political life. In Athens, citizens (at least rich, white, land-owning ones) could get up and argue their views to the people. But some way of sorting out the stupid opinions from the interesting ones was needed. This is how philosophers started out investigating the meaning of “knowledge.” It wasn’t enough that what the person said was TRUE – the person had to be “justified” in believing it. Most of the subsequent philosophical efforts centered on the meaning of justification.

DW then asked us what “true” means. Is it “true” that John Palfrey has a Thinkpad? Yes. We could see “true” as meaning “everyone can/ would/does agree on it.” But we could also see it as being independent from what people think, and being purely about existence (Conor’s view). Yelena thinks it includes both objective and subjective components.

DW drew some interesting figures on the board, one representing a person’s head (the “knower”), and another representing a laptop (the world). He suggests that if the statement represents or matches the world, then it’s “true.” Kevin points out that it’s not the thought that’s true; it’s more like running an experiment – determining truth doesn’t require people. The experiment verifies the claim, and the process is what makes something true. DW said that there’s lots of debate about this stuff, but reiterated that truth requires some type of correspondence between the statement and the world. This is called the “correspondence theory of truth,” and it’s dominated Western culture for many centuries.

Descartes and his Meditations
We then turned to discussion of Descartes’ Meditations. Damien outlined Descartes’ basic idea: he wants to question the veracity of everything that isn’t verifiably true. To that end, he starts on a project: can he tell whether he’s awake versus dreaming?

Conor elaborated on how Descartes goes about this project in the Meditations. Descartes eliminated all thoughts or beliefs that were based on potentially fallible sources. For example, anything he gets through his senses is potentially faulty because his senses have deceived him in the past.

So, in Meditations, Descartes goes through a process thinking he can examine all his beliefs. Senses are out because they’ve fooled him before, and could fool him again. DW asked: why don’t we doubt our senses the way Descartes did? JP suggested that it’s because we’re not philosophers… Richard said it’s because we don’t care whether what we’re getting from our senses is “fundamentally true” – we are pragmatists and the senses work for us most of the time. Yelena added that we trust our senses out of necessity – there are no alternatives, except paralysis. Finally, another student noted that we’re usually right about our senses, and when we’re wrong, we learn from it – so it’s a workable system of reliance. There’s a difference between mistakes in our senses and random hallucinations.

DW then asked: how can Descartes doubt that 2+2=4? Doesn’t he “know” that for sure? Damien responds by saying that this part depends on Descartes’ belief in a deity – someone who could arrange the world in a way that would completely convince him of its truth.

Then, after a very confusing moment about whether 2+2 really does equal 4, DW brought up the idea of “the malignant demon.” (Apparently Descartes had to talk about this as a demon because of the religious constraints of the time – one couldn’t talk about God as being deceptive.) DW asks why Descartes engaged in this extreme experiment (so extreme that he was doubting whether he has hands!). (Ultimately Descartes goes on to discover “I think, therefore I am,” and on the basis of that, he’s able to build back everything he reasonably knew before. DW admits that it doesn’t hold together very well.)

So again — why do this crazy experiment? One student used the analogy of creating art on a blank canvas – you need to start with a clean slate. Richard added that Descartes seems to be cutting it all down to the most basic thing that can’t be disproven – he’s creating a super-strong foundation for the “building,” which makes it harder to refute.

On a side note, DW asked: how helpful is Descartes’ method for making political decisions? The answer seems obvious: not very. We tried to envision two politicians debating, and then one asking, “How do we know we’re all here?” Suffice it to say that we don’t need that kind of certainty in the political realm.

Stepping back, DW concluded that Descartes was aiming for the perfection of knowledge. But the problem is that, in seeking such perfection, you end up with basically nothing. And the only way Descartes is able to build it back up is by saying that God wouldn’t allow him to be wrong about various things (his senses, for example). Descartes has tried to say the only things we can put in the “knowledge” category are those which pass the really strict test. And over time, the bar of certainty has been continually raised. We circled back to the crossword puzzle, which defined knowledge as that which we know for sure – it’s about certainty, not justification.

Phew. After all this profound talk of philosophy and Descartes, we turned to Wikipedia.

Wikipedia and “knowledge”
We started with Shakespeare’s sonnets. The Encyclopedia article is mostly about the sexuality of the object of the sonnets. DW was surprised that the article might as well have been called, “Just how gay are the sonnets?” The Wikipedia article is much different. It lists a number of “contenders” for the mysterious W.H., for example, instead of simply saying who it was.

Then we talked about the discussion section attached to the Wikipedia article. Apparently most of the discussion centers on the identity of W.H., but also on the sexuality of Shakespeare himself (and what to say about it). There’s a question of whether or not the W.H. question should even be talked about at all in an article about the Sonnets. So even settling the fact – who is W.H.? – wouldn’t necessarily settle the issues. (Christina was initially impressed that the Wikipedia community was keeping all these comments up, but it turns out that you can’t delete discussions…who knew?)

The discussion pages raise the broader issue of the purpose of Wikipedia entries. Is it just about presenting competing views (what other people have said is true)? Or is it about presenting “truth”? DW made the point that there are decisions being made about what’s a mainstream dispute and what’s not. (For example, there’s very little discussion about Shakespeare’s death date.)

We then moved from Elizabethan England to Colonial America to talk about Sally Hemings. The controversial issue was whether she was the mother of Jefferson’s child. The Wikipedia discussion was interesting in that there was no real consensus or closure about this stuff. There was also lots of talk about the DNA evidence – and that there were four Jefferson males who could have been the father. DW thought it was interesting that this is all back in the discussion page and not on the entry page. Others commented on the debate over the more stylistic aspects of the piece (opening with “Sally Hemmings was the chambermaid of Thomas Jefferson” etc.).

Chiming in appropriately, JP talked about his interviews w/ young people in which he and Dana ask how students start research projects these days. Most said they start with Google and then Wikipedia, linking to external sources for verification. But, interestingly enough, they almost never edit the entries, even when they find mistakes. He wonders why more people don’t edit (other than the alleged group of drunk German grad students at the helm).

DW said that in all of these controversial articles, there are statements put forward as the clear truth; and then there are statements qualified by phrases like “some believe that.” We looked at the Swiftboating entry for signs of these “weasel words” (which Kevin argued were not as commonplace as DW suggested). “Were criticized” was one example. Conor gave us an overview of the Swiftboating discussion pages: most of the argument was about the appropriateness of the introductory paragraph. As for the article, DW thought that it wasn’t an unreasonable presentation of the basic issue.

Finally, the JFK assassination page. Students noted that the tone of the discussion is pretty contentious, especially with regard to the authority of the Warren Commission report. Evan suggested that this article is an outlier. He would compare it to the 9/11 conspiracy theories and the corresponding Wikipedia pages. (DW explained that a separate assassination page was created because the JFK page was getting overridden with conspiracy theories.)

Class ended with DW posing a broader “web difference” question: what is the analog in the real world for what we’re seeing here in the discussion pages of Wikipedia? Justin suggested that it is talking to friends and reading books. But DW said he doesn’t know anyone with the depth of knowledge and obsession demonstrated in these Wikipedia discussions. Another analogy was made to peer review in academia.

Ultimately we were all left wondering whether Wikipedia has any kind of analog in the non-Internet world. Is knowledge all about certainty? Do you only “know” something if you can logically justify why? Are traditional notions of perfection simply inapplicable to the internet – and is that a good thing? These were all questions we addressed in class, and ones we’ll no doubt revisit throughout the remainder of the course.