An interesting thought... but you've got to elaborate on that :-) - Ankush Narula

People-centric web? I don't trust people. I trust their "interests". People love shiny and trendy things. I love people who love what I love. those people often love trendy things but that's okay, since I love them :) — In other words: an attention-centric web is better, IMHO - directeur

Humm, not sure how you'd go about a people-centric web. You'd have to first get past the integration level (duplicate name spaces, different personas, householding) and then deal with the age/time variable (who you are today isn't who you were a year ago, or 10). How would we determine what a person stood for? You'd be back to on-site analysis of who THEY say they are and what other people say they are which is done via ... links. Google has an algorithm that tries to emulate how people would interpret and process a page or site - it's not very smart right now, but the task is enormous and is essentially the construction of artificial intelligence. - AJ Kohn

AJ, what "I" look like, my skin color, my name, my religion, my language... are NOT important unless someone wants to "sell" me something. The online "Me" is the set of "things" I'm interested in and the set of "people" I like to interact with. Google algorithm is based on a simple fact: You have content about a "subject", and this content is linked by content that is already "popular" in that area, then your content is worth interest. It's like being recommended by Einstein when you postulate for a "physics" job. The same recommendation shouldn't be interesting if you're working on "crocodiles nutrition" - directeur

To extend your analogy... shouldn't Einstein's recommendation carry more weight since he's a known authority in physics? I think the Google algorithm might not take this factor into account unless the algorithm looks at the popularity of the linking page/site as well. - Ankush Narula

Ankush, and that's how it works in fact. A link by Dmoz directory is better than a link by well, britney spears' fan geocities page :) — their algorithm iirc looks at the semantic subject of the content + the rank of other content that links to it - directeur

Hadn't realized that... now back to what a holistic, people-centric web is all about... - Ankush Narula

@directeur: I completely agree, which is why a people-centric web seems ... improbable (impossible?) and likely unsatisfactory. @Ankush: directeur's right, Google gives sites more authority based on the number and quality of links it gets - and not just overall but from the same topic/category/niche. It's still not very smart in the scheme of things (hence my blog's title) but it's getting better and better. - AJ Kohn

This is not a positive and in a way quite false assumption. Google's algorithm has not done any damage, if you remember where the state of research were before Google came. Google enhanced in some ways the Web at this time. But if you have an idea for a people centric Web search, then go for it ! ;-) - Thierry Lhôte

And also the approach of Google search IS holistic even i it is not people oriented. it is perhaps the first real try of Web-mapping. - Thierry Lhôte

Agree Thierry. Also, even if google is not perfect (no AI will be perfect) *people* are responsible for its imperfections. People love to game the search engine for fun and more often for profit. Those "gurus" that other people often call "authority" ask their subscribers/readers/"managed community"/horde to "bookmark" (not fave/vote/share) their entries are corrupting the value of "value". This is another kind of spam, it's a taboo, but it's a vicious kind of spam. - directeur

I think the problem is that there's potentially great content in the long tail that won't surface in a Google search because Google focuses on popularity + authority. This is just what makes the Google search useful, so I wouldn't expect a change. However it would be good to have alternatives - search that finds quality in obscurity, the sort of stuff that we now find through curated filters like boing boing. I met with a South Korean now based in Austin who's developing an alternative search engine with this in mind. (I don't think I can say much more, though - but the project's very interesting.) - Jon Lebkowsky

I saw a paper a while ago (can't find the reference now, sorry) that found that Google's existence increased the popularity of the “tail” pages, because people no longer link to the big popular sites anymore, since they can find them on Google. They link to new, interesting, and hard-to-find things, which over time gives those things more PageRank. - Amit Patel

So long as people, not bots, are the ones making the links I don't see the problem. - Garin Kilpatrick

I think my point was lost. Generally speaking, the focus on links as the atomic unit of the web has done several useful things — but in terms of being able to ask: "Who is this person that wrote this content? And what else have they done? And who do they know? And what is the complete picture of the self-conception?" I think that focusing so much on *document* links has inhibited the development of the social web. - Chris Messina

and 4chan would never have existed :) - Jérôme Flipo

Chris, I see what you mean. Look, one thing I'm humbly sure of is: trends and hype will always win over personal relevancy. That's sad I know, but I'm almost resigned. I've humbly created socialwhois http://socialwhois.com which btw uses google SGN, supports microformats, apml, xfn but most people don't know about it because I'm just directeur, not say... Kevin Rose who afterwards launched wefollow. I'm not sad or bitter about it, not at all. Just more sure about how this world works :) - directeur

Yes Alex, exactly, Web was not designed to be social-media, usenet, IRC and the like were. The future paradigm of the universal social-media tool should evolve around Google Wave concept and XMPP. - Thierry Lhôte

Thierry and Alex, it was not designed for this in fact. But efforts like microformats make this possible. Data about people that is machine-readable using POSH. And Google who Chris is blaming here actually made a big and useful piece: SGN. Questions asked by Chris above "who wrote this, what did he wrote..." are ligitime, but if I had to chose between content and persons, I'd go with content. in other words: I would like google to give me more similar content (semantically) than all content by someone. For I'm not tracking people, but interests. - directeur

i use the web from a greedy interest of problem solving. thats why my needs are problem/solution centric. the best solution is the one that fits to my situation and worked for the most people in similar situations. therefore i love a net that i feed with my data (location, situation) and i get socially rated customized solutions. perfect :) and since i have lots of time im interested in popular problems and popular solutions beside my own ones.(based on the "maybe this knowledge will help me one day" spirit)...i also like lolcats - Chris Hofmann

@Alex: that's kind of my point. What if it had started with proper representations of people, like <person> tags, etc? The reason why we rely on centralized services like Twitter and Facebook is because they've added a "people" layer on top of the "document web". And, if I search for someone in Google, I'll get many links to their profiles — but not a complete view of them. Hence my original tweet. - Chris Messina

"I think that focusing so much on *document* links has inhibited the development of the social web." Well now, there's a whole body of work in the semantic web directed at that problem. So rather than complain about Google's statistical model, what will you do get the diso, activity streams (and the html5) crowd to use or learn from it? For example activity streams is still in need of a way to represent actions. - Bill de hÓra

I think sites like BackType and Disqus remedy the "what else has this person said" problem to some extent. Google Friend Connect + Google Profiles tries to solve this problem also by becoming a distributed social network. Over time, I think similar competing protocols will help the social web will integrate to the point where aggregating contributions from and information about a person will be rudimentary. We just have to make sure the protocols stay open and and the technologies proliferate. - Ankush Narula

Let me add that I think OpenID needs more support. - Ankush Narula

@Bill: Hrm, the semantic web is still rather reductionist and mathematical, rather than being expressive or generative. In terms of the effort it takes for my mom to benefit from what I understand as being the "semantic web", I don't think she'll see it in her lifetime. - Chris Messina

OK Chris, so it is 1997, and Altavista is the king of search, with a bunch of others coming up behind. Everyone's search results are getting crapped up with SEO-driven spam, such that meta tags are now being ignored by all major search engines, and various forms of 'invisible text' on the page are state-of-the-art. What would you like to have happened differently from the approach Google took for increasing search relevance? - Michael R. Bernstein

@Ankush: BackType is a great service — Disqus, CoComment, Co.mments and IntenseDebate are all also interesting solutions — but require you to re-establish a profile. Why can't one just pick their definitive profile and federate that? These services hint at what I'm getting at, but still leave much to be desired. - Chris Messina

@Michael: Good question. I guess one possibility for routing spam and whatnot is to become more people oriented — as Facebook is. If you source content to particular profiles, then you can build up what Facebook now calls the "social graph", to help inform reputability. Back in 1997, there wasn't enough connectivity or enough people with online presences to make such a situation really work — but I think that situation has or is changing. - Chris Messina

@Chris: Well, you're just looking for the overlay of the semantic web then, correct? Whether it's self-tagged by the person, or derived through an algorithm, that's how you'd map document to author, and author to author universe. Self-tagging is the only way to ensure this happens correctly - algorithms can be developed to determine if it's the 'same' ... AJ Kohn or not. So, Google *should* be able to disambiguate AJ Kohn the skateboarder from AJ Kohn the marketer/SEO evangelist. However, I don't think they're very far down that road yet - the disambiguation algorithm is quite complex and can't determine when someone is using a pseudonym or simply a different persona. - AJ Kohn

AJ, relying on self-tagging has some serious problems. See the aforementioned SEO tactics that rendered HTML meta-tags useless. This is known as the 'metacrap' problem. To use your scenario, what happens if you actually self-tag as AJ Kohn the skateboarder? - Michael R. Bernstein

@Michael: True enough. Self-tagging can be exploited in nefarious ways as well. Which brings us back again to the links caring some sort of semantic meta data - so that it is others who help define the who - and can actually be a safeguard against 'semantic-theft'. But, that too can be abused in a Google Bomb sort of way if everyone agrees to associate someone as the author of some content. That's why the 'who' is such a difficult nut to crack. You'd need OpenID to help you - and I'm unsure a lot of folks are willing to provide all those associations for public use. - AJ Kohn

Well, there's a re-balancing that's needed between social knowledge (i.e. who knows who — vis a vis Twitter/Facebook graphs) and the links that are about, related, or created by people. If profiles were first class citizens, and everyone was incentivized to have one or two EACH (rather than many), perhaps the social web would be further along than it is. - Chris Messina

What is a "first class citizen", Chris? - directeur

Documents are currently first class citizens on the web. As are images. Videos are not. A basic test is to consider whether an object class shows up in the source code of the web. We have <img> and <object>, but no <person> tags. It's simple, but it'd be much easier to create social websites if designers could just use something like <person class="viewer"><avatar /></person> to show the active viewer's face in their HTML! - Chris Messina

Chris. Why? I mean.,microformats do this already. why add a new tag? We just can say to the designer use this class name please. - directeur

Sure, I support microformats, but they don't "do" anything right now — not without custom JS. In this case, I'm talking about tags (or microformats — I don't really care) that are supported by browser rendering engines — that could reveal social information based on the viewer and his/her friends. This is similar to OpenSocial's model, so I guess there is *some* progress being made my Google! ;) - Chris Messina

Chris, I think you're about 5 years behind Google on this. Google in 2000 only knew about documents. Google now knows about (and ranks) entities such as businesses, products, and people. Now it has a variety of ways of linking those entities to documents (submitted data feeds, microformats, and algorithmic extraction, among others). Authorship is another question, and you have a good point about that -- there are no good solutions there. But railing against link analysis seems like saying, "It's bad people have half a loaf, because if they had no bread at all they'd be so hungry that they'd work harder to get a full loaf." - Daniel Dulitz

Hrm, I wouldn't say I'm railing on Google at all. I'm only contemplating the effects of decisions — which, in hindsight, prove them to be rather wise. Instead I'm asking: what did we not focus on because we chose instead to focus on links? It's great that we have half a loaf — but what about the other half? - Chris Messina

Well, microformats and RDF are being used by Google and others in a number of ways. However, I think we'd all like to develop a way where the user didn't have to do the work and instead the relationships could be discovered and learned by the algorithm. @Chris: I think the one thing you're assuming is that people would want just a few profiles, or that it would be the optimal way to see the 'who'. The Internet has been good at allowing people to involve themselves in various areas, without being shunned by others or simply being 'outed' - it allows for compartmentalized exploration. So, a person who enjoys death metal and knitting might have a very different profile on Last.fm and a Knitting Forum. I think having disparate profiles is a source of rich data to generate the 'who' - both in diversity and in time (e.g. - an old, outdated profile could reveal how a person has changed - MySpace profile versus Facebook profile perhaps?) - AJ Kohn

What about => FluidDB » Blog Archive » Truly social data http://blogs.fluidinfo.com/fluidDB... - Desirade

Damage isn't quite the right word. - Cliff Gerrish from iPhone

@Cliff: "inhibit"? - Chris Messina

@Aj: I support the idea of multiple profiles — and of sharding identity, if you *chose* to. That it's the default baseline means that people who don't want to fragment their digital identity are forced to. Put another way: if BOTH sharding AND holistic identities were the norm today, I wouldn't likely have raised this point! ;) - Chris Messina

@Chris: Got it. If anything, a holistic identify - a primary or default identity *would* be an interesting idea. It could serve the purpose for those who *do* want a holistic identity and can be a sign-post for those who want fragmented identities. [edit] I appreciate and thank you for the dialog. - AJ Kohn

Was this really some sort of decision Google made, or was it just a natural evolution of the "one page = one URL = one object" paradigm we've been living with since day 1 of the web? Blogs, search engines, URL shorteners... everything's perpetuating the idea that you don't link to *stuff* or people per se, you link to pages. - Ken Sheppardson

And of course everyone wants to own that page, or your views of that page, like Twitter, Facebook, etc. Everyone wants you to use their UI. We can't just link to a person or an an article or a comment, we have to link to that object with all sorts of UI overhead, ad banner wrapping, etc. We're stuck in a site/page centric world, vs a user/content centric one. See also http://friendfeed.com/vanelsa... - Ken Sheppardson

I don't think I would be too worried about self-tagging. It's just one piece of the puzzle. There are other forces available, such as all my friends that know I'm an extremely lousy skateboarder. Also, I would like to think that I'm just one person on the web, but that I can choose what aspects of me are visible in a given context. - Alexander van Elsas

Totally agree we should have a whole loaf. I think there are a lot of folks who want to solve the identity-on-the-web problem, including the JS_Kit guys, Facebook, etc. This discussion here is very interesting, thanks. - Daniel Dulitz

"the semantic web is still rather reductionist and mathematical" @Chris I can see why you'd say that, and I don't go for the high formalism much myself. I think it helps to consider something like RDF as a medium, and one that arguably has more degrees of freedom for expression that say relational technology, or the current web2.0 argots. In comic book terms, RDF and Linked Data is for penclilers, XML and JSON is for inkers. The story is another matter ;) - Bill de hÓra

"However, I think we'd all like to develop a way where the user didn't have to do the work and instead the relationships could be discovered and learned by the algorithm." @AjKohn back in the day, this was called a hybrid architecture - lower layers dealt with statistical, telemetry and event data and higher layers dealt with representation, concepts and logic. A good example from robotics is InteRRaP http://www.dfki.uni-sb.de/mas... - Bill de hÓra