social metadata, relevance revolution

I may be somewhat distracted. I have my phone in my pocket on vibro mode, waiting to hear from C when he gets back from the beach. (per the whole glasses thing) And if this doesn’t grab me, I’m going to bail.

harness user actions to make site more relevant. he’s from the middle of nowhere, aka Edmonton Alberta Canada.

blah blah blah about his company. he does IA, mostly. has been a metadata nerd. take what users are doing to create architecture of sites.

emergent information architecture -> collective intelligence -> information architecture 2.0 🙂 or social IA.

get a better understanding of social systems
practical ideas for tweaking your own systems
feedback loops as part of design process

he’s tweaked to fit with some similar presentations.

structural design of shared information environments “fancy term for websites” interested in systems where users are co-creating the environment. less structured, more organic.

social IA: user actions create some or all of the environment. use wisdom of crowds to solve problems of IA.

examples: amazon, starting with “customers also bought”, highlighting listmania, combining algorythms with user-generated data. wikipedia, not just creating the articles, but also the connections between them, created by contributors. flickr. not just tagging, but contacts. (I rarely view by tags, but I like the contacts stuff.) delicious. “canonical services” digg.

range of uses, from augmentation (amazon, ebay) to co-creation (delicious, wikipedia).

“contrails in the great database in the sky” and nobody seems to mind. ah, yes, how the information we throw out into the universe can be mined to serve us more and better ads.

“whoever has the most friends wins” (re: linkedin)

the new yorker cartoon, nobody knows you’re a dog. vs. a new cartoon “Ogle Earth” — the shift in expectations. web as part of our social infrastructure. (hi, guys!) people expect to be part of the conversation, and to get the most relevant stuff. (thinking again about personalized home pages?)

3 ingredients for social IA: way to capture user actions; way to aggregate & display; feedback. today less concerned about why (from our POV or users) than how.

user actions: the things they do that we can track. understand popularity, community, reputation, etc. for the moment, put aside higher goals & motivations.

speculative graph…low to high engagement vs. social intent (personal to participatory). automatic, personal, low intent: pageviews, clickthrus, downloads (server log stuff). personal but more engaged: purchases, tags (boundary-crossing), bookmarks, linking to something. more participatory: posts (flickr, youtube, blog), ratings, buddylists, comments/reviews, wiki-ing (nice coinage). trackback is off in its own little corner: low engagement but more participatory. (I wonder if that is part of why trackback is broken. it’s really easy to fsck with other people.) plus mentioned not on his chart:

as moving to the right & up, identity becomes more important, including its persistance over time. how many of these actions involve an “object” (a virtual or real).

topographies of services. delicious (blob in middle right), amazon (a line of sort heading up and to the right), youtube (a blob similar to delicious, maybe a little higher up, with the anomolous dot of pageviews).

on to aggregation & display. bringing together user actions in a relevant way, and then displaying them. coming up with a set of rules for both.

lots of kinds of aggregation (5 (of N), he’ll talk about 4): listing, ranking, clustering, collaborative filtering, and other stuff.

ebay example of listing and prototagging. men’s apparel category; used to shop there before there was a banana republic in edmonton. 🙂 new pants or used pants? evolved conventions in the subject line. NWOT: new without tags; NWT: new with tags.

listing friends.

youtube and ranking. counting & ordering actions. the essence of popularity. a bunch of different axes of popularity. worst rated: “the cesspool of youtube, but that might be redundant” — getting different videos with different qualities.

yahoo most recommended photos. girls in bikinis: never show up in most recommended, but always in most viewed. reveals something. need multiple rankings to call everything that’s interesting about a particular set.

collaborative filtering: comparing your history with those of others to find stuff you might rate highly. amazon, also netflix (yay!). open source libraries for collaborative filtering (?! in php?!) — kewl.

examples from amazon: new U2 album links to a variety of stuff, but Joshua Tree only to other U2 albums.

digg shows up under “other algorthyms” 🙂 a whole bunch of factors are included. number of votes, source of the story (original or repost), history of submitting user, traffic of category, reports on the user.

feedback experiments. “okay, that was not very successful” “I tricked you into [the wave]”

feedback loops in social software. positive feedback in Digg. influence wanes over time. plus the feedback loop on the individual users and their ratings. “Top 100 Digg users control 56% of Digg’s home page content” how do those other users feel? do you really want to participate? is it really democratic? “but we’re just going to talk about the hows”

5 (of N) places to intervene: introduce delays (moderation of blog comments), modify strength of feedback loops (imagine adjusting the rules), access to information (cliques on digg: what if you couldn’t see who had posted a story until after you voted for it?), adjusting incentives/punishments, change the system altogether.

examples: the emergence of tagging with flickr & delicious. google’s aggregation of links as votes.

mefi cover charge: another intervention point, through limiting membership. or gang initiation. recommended by a member, etc.

danella meadows, list of 12, based on big systems: economies, but they work great for smaller systems. look on wikipedia.

challenges. spam, gaming the system, achieving balance, relevance (do your users find it valuable? talk to real people!), unintended consequences.

his design principles. allow for different levels of engagement (that connects to something earlier today, about the 90%), monitor & tweak feedback loops, participate in the larger ecosystem (ala Powazek & the company town), design new actions, aggregators, displays.

q: when tweaking, don’t you run the risk of alienating the user? yes, absolutely. (I think I remember some discussion on this topic in the new web apps panel at sxsw.)

q: ???? basic thing is figuring out what people are doing that you can do something with and doing that. wary of talking about systems that need high level of participation to get off the ground.

q: inside the enterprise. scuttle (what is that?) something at IBM.

q: what did you not talk about? myspace, consumating.

q: is there a site that lists them all? no, maybe on wikipedia? questioner has found all of them really boring.