Saturday, March 05, 2005

Sorting out Social Classification (Peter Merholz, Peter Morville, Thomas Vander Wal, Gene Smith)

Session description

Huh. Looks like Stewart Butterfield bolted at the last minute. Now we'll have a bunch of IA gurus nattering on about social classification, while the one guy who's actually made a wildly successful implementation of social classification is absent. Hmph.

Raw notes to be polished later:

- long tail, emergent trends/vocabs, wisdom of crowds

- broad = many users tagging one reource
- narrow = few users tagging one resource

- tags (public - private) vs ownership (my stuff - others' stuff)

various opinions re: experts vs amateurs
- Clay Shirky, Rosenfeld, Weinberger: The old way creates a tree, the new way rakes leaves together

- retrieval, quality, authority, economics, scalability, usability

- Revenge of librarians - "Internet will make everyone a librarian"
- metadata is sexy today
- took a provocative stance: Folksonomies - Better than Nothing?
- Apophenia: mistaking noise for signal
- Weinberger is right! Leaves rot and provide food for trees
*- content owners have lost control of aboutness (e.g., Google "miserable failure" = "GW Bush biography")
*- folksonomy ties tightly together people and content
- data and metadata relationship is changing (data was suitcase, metadata was the name tag on it)
- child tracker watch (findability and breadcrumbing), RFID chip implant makes Mexico Attorney General a smart object
*- 5 lessons of folksonomy:
1. leverage what already exists
2. tap wisdom of crowds (and users)
3. tap the compulsion to share (don't ask too much, pennies vs dollars)
4. context counts (avoid generalizations)
5. never underestimate people's thirst for anarchy
- persistent disequilibrium (Kevin Kelly): how we define things continues to change

- tags vs metadata
*- harvesting outlier tags (long tail) may be lucrative opportunity (e.g., Carhart farm wear is cool wear to German skateboarders, Carhart has new market)
- identify trends in realtime

- metadata for the masses
- issues of practicality
- traditional wine classification (country, year, grape, etc.) vs "goes good w/ steak" etc.
- metadata problem: content creators tagging their content
*- folksonomies aid rapid development of categorization schemes
- folksonomy problems: how to relate synonyms (e.g., nyc, newyork, newyorkcity)
- multiple meanings (e.g., flow, flowchart)
- misnomers, bad data (e.g., people think archaeology = paleontology)
- desire lines: use folksonomies to develop more robust classification
- examples of hybrid classification systems: Getty
*- folksonomy allows for poetry in classification ("me" tag in Flickr is more personal, visceral than "self-portrait")
- Platonic ideal of classification (perfect, robust) vs personal, human approaches (compelling, inspiring)
- findability vs discoverability
*- metadata creating community (squaredcircle in Flickr - people taking pictures so they can use a specific tag)

- combining taxonomy and folksonomy
- folksonomy as research tool
*- linking disparate domains of expertise that use different terms for the exact same thing
- how to make everyday people want to tag their content? ... increase ROI of tagging
- Morville: folksonomy as socially acceptable form of spyware
- are critiques of folksonomy more implementation issues (e.g., tool maturity)?
- the power of naming things: neologisms, coinages (e.g., AJAX)
- JJG: Web development as history of people running away from the hard work of IA
- Lou R: are we getting too excited about another new scheme or technology? Shouldn't we be more skeptical? Are we missing opportunity to demonstrate IA strengths are in synthesis and skepticism
- example: IAs going in to sort out a Wiki
*- feedback loops in tagging tools to help maturity
*- freelisting: search tags and related tags and related related tags, then use those tags for cardsort
- what is the tipping point for a tag? Where does the power curve, net effect begin?