FOWA 07: Bradley Horowitz – Social Interaction – What the Future Holds


‘User’ is a bit of a derogatory term, and have to move away from using the term and turn ‘users’ into ‘people’. Came up with a pyramid:

1% – creators
10% – synthesisers
100% – consumers

Web 2.0 tries to make everyone a creator
100% – creators
100% synthesisers
100% – consumers

Try to make it easier for everyone to participate in all roles.

Anyone with an X is now a Y
keyboard –> author
camera –> photographer
iPod –> DJ
browser –> publisher

There is a dark side. If you factor out all the pirated content on YouTube you have chimps doing karate. Democratising publishing and the best they can come up with is this. Lots of UGC can be a sad experience.

But when you look at some of the most interesting photos in Flickr they are amazing. These are ‘user generated content’, all taken by ordinary people, not professional photographers. It’s amazing in quality, it’s emotionally moving, some of it’s funny.

Flickr’s ‘interestingness’ is an algorithm, done implicitly. The concept is stuff that is interesting to everyone, not just those who know the person involved. Look at how many times the photos was commented on, viewed, blogged about, linked to, etc. All that organic activity allows us to distil ‘interestingness’. When they launched it, there was no gaming of the system, no spam. Could use retroactive techniques to go back into the archives and see what the most interesting photo on the day they launched was.

Technorati Tags: ,

Every ‘user’ becomes an editor – their activity in the system can be reflected back. For any day you can see the most interesting photo.

Turning ‘users’ in to taggers. Applying metadata to content is not new, but tagging is very easy and flexible.

Digression: computer vision is hard, computers cannot understand images semantically, but people do it verbally. A banana-detector we could do, but identifying a church is tough.

Our tag for this conference is ‘fowalondon07’.

Those things can be hard to remember, so can use context to solve this problem. On Upcoming there’s an ‘add photos’ tag, and they use machine tags to link event and photo tag. But that tag is even harder to remember.

Camera phone knows a lot of context which is sometimes lost, e.g. who I am, time, location via GPS or cell tower beaconing. So given an ancillary database from Upcoming and geotagging, we can automatically tag the photo with the right event tag.

Emergent knowledge. Flickr launched geotagging so that one can take a photo, drag it onto a map and see the photo’s location. E.g. route 66 shows up on the Flickr map.

Tagmaps, looks at geotagging, with popular tags show up in bigger font in an attraction map. Photos then show in the sidebar.

Keep barriers to participation low. Use clustering. Librarians go pale at the idea of tagging, and invoke issues such as ‘Turkey’ the place and the bird. But clustering allows people to see photos grouped together, and it works on clusters based on which tags occur together.

Jaguar is interesting, but better is Love, romantic love, symbols, pets, family.

Lotsa tags. Night Explorer, same as the attraction map, but only photos taken at night, tends to show bars, night clubs and nicely lit architecture.

Turning users into neighbours. Only a couple of ways to develop a relationship with readers, via logs, but that’s not satisfying, and via comments, but that’s only a minority. So blogging is a bit like blogging into the void. MyBlogLog gives you a chance to see who’s coming to your blog.

Pipes: Rewire the web
Pushing HTML to the ‘users’, and the web is guilty of building web pages experienced through the browser.

RSS is mashup for the masses, so Pipes allows you to take any RSS feed and pull it into one page.

A real-world task to illustrate Pipe: find an apartment for rend within walking distance from a daycare centre. Would use Craigslist and Yahoo! Local listings to find the apartments and the daycare But would be hard to do that manually. Data is there in both services, but no easy way to bring them together.

Can reverse engineer the geocoding, and translation, for example.

Underestimated how many people wanted to play with a service like this. A few people will create value for others, and a small group of developers will do really cool stuff with it.

One applications is ‘Apartments Near Something’ so can find housing near any specified location type, e.g. parks, tube, daycare.

Pipes. Can vote on what features you want them to add next.