How Websites Learn Information Architecture that Adapts to

How Websites Learn: Information Architecture that Adapts to Use Peter Merholz Work: http: //epinions. com Play: http: //peterme. com/ia 2000/

Little Architects

A problem: Increasingly, websites corral massive amounts of information In fact, a strength of the Web is access to unlimited information But how to present the information meaningfully?

A problem: Providing a singular, top-down editorial structure isn’t feasible But nor can you allow a morass like USENET or an unstructured Web

Another problem Web sites don’t respond to use They’re static, and assume all information is of equal importance

Case in point Productopia, R. I. P. Dozens or hundreds of employed people are costly

A potential solution Adaptive information architectures Bottom-up organization based on qualities of the information and how the information is used Information spaces that metamorphose based on use Rich information spaces are complex systems, the study of which can inform our designs

Structure based on the qualities of the information Linguistic processing and categorization schemes like Autonomy and Northern Light Display systems like Self-Organizing Maps (http: //websom. hut. fi/websom/), Thinkmap (http: //www. thinkmap. com/), Cartia (http: //www. cartia. com/)

The use is more important than inherent qualities While the Thing qua Thing is important, It’s more important how people relate to and interact with that thing Cognitive scientists have found that people categorize the world not on the inherent qualities of things, but on how they interact with those things. (Women, Fire, and Dangerous Things, Lakoff)

So, what does all this mean? What is an information space that adapts to use? An obvious and popular example: The Bestseller List - People are interested in what’s popular

Models for social effects Self-policing Word of mouth Footpaths The Agora

Self-policing – Epinions. com Nearly 700, 000 opinions—no editing Community can - Rate content - Trust each other

Self-policing—Epinions. com Generic

Word of mouth – Information structured through people You tell people about a movie you saw, a book you read, a car you drive Known experts are sought out for their input

Word of mouth – aligning yourself with others’ tastes Epinions. com – Other products written about, other opinions worth reading Launch. com – Collaborative filtering Napster – Hot lists

Word of Mouth—Epinions. com Generic Personalized

Word of Mouth – Google. com That HREF tag carries a lot of meaning Page. Rank – measures importance through linking

Footpaths How is the information traversed and used? Take advantage of what people do

Footpaths Designers begin with a form… http: //fury. com/berkeleypaths

Footpaths But people will make it their own…

Footpaths – Swiki : Icons representing use http: //swiki. sics. se/ Footprints and Dinosaurs Explicates what’s there, doesn’t make new connections

The King – Amazon. com People who bought X also bought… Creates meaningful relationships that can break typical taxonomic bounds Purchase Circles Localized Bestseller lists The Page You Made Personalized clickstream analysis

Amazon’s biggest lesson Track use passively—don’t expect people will do the work for you

The Agora People in the same place at the same time likely have something in common Purple-moon. com H 2 G 2. com

A Meta-Web – Everything 2. com A bottom-up network of thought connected by Hard links – explicit Soft links – implicit Ranked by voting Tiered community ‘Who’s online’ …. They’ve got it all!

Except a sense of purpose, structure, meaning… Difficult to figure out Community is double-edged In an email from Will Sargent: The downfall of Everything 2 (the site, not the architecture) is sadly linked into its quality control. Editors are picked from users who have made significant contributions to the site and have been there for a while. These editors/gods have free license to downvote and nuke as they see fit. So far so good. The problem is that the editors are not impartial observers. They have been known to downvote nodes which use language which they don't like, political opinions they don't agree with, and nuke any nodes critical of E 2 editorial policy itself. In one case, they outright banned a user who had extreme right wing opinions nobody liked very much. Of course, the guy who writes thestileproject. com got banned in seconds flat. There is no appeal, and if a node gets nuked there's no way for the writer to retrieve that content.

So, should you fire your IA’s and let your users structure your content?

Yes!

I mean, No! Obviously, there must be an initial organization Create an organization that is flexible, not rigid But make sure what changes are the connections, not the addresses

Those %$&$#! Three Circles Business Context Content Users

Further Concerns Data analysis – sooper-dooper important Make sure your project has a real mission Close watch—site structures based on use require a community… And online communities require gardeners to nurture them

A Quick and Dirty Algorithm http: //pespmc 1. vub. ac. be/INFOJB. html Frequency—reinforce a link (A B) that a person traverses Symmetry – reinforce reverse link (B A), a little less Transivity – If a person goes A B and B C, reinforce A C This is what leads to true restructuring

Future Uses

Peer to peer Think beyond music Think of an ever-shifting stream of information “Architect” that

Personal Categorization If I’m involved in SO/HO, I’m not interested in ‘electronics’ or ‘computers, ’ ’I’m interested in what can help me

Evolutionary Model Again and again it’s been shown that attempts at God-like ‘organized’ top-down design typically fail Too many contingencies, too chaotic Bottom-up, rules-based structures adopt, adapt, and improve Organization without explicit design is very powerful

From http: //pespmc 1. vub. ac. be/papers/namurart. html “We believe the laws of evolution, in which natural selection guarantees the survival of the fit and the extinction of the unfit, apply in all cases whether living beings, dead matter or knowledge is concerned. Ideas, chunks of knowledge, can be considered specific entities that rely on human or other carriers to multiply, mutate, adapt and survive. The human population and the technology devoted to communication can likewise be regarded as a huge ecology populated by ideas, theories or knowledge in general. The Internet has in the most recent years been becoming an integral part of this so-called ecology of knowledge…”

Sites need to evolve structures Evolution requires An Environment In which elements are Selected Based on their Fitness --Natural Selection

Natural selection White and black moths

Sites need to evolve structures Structures require Principles In which Meaningful organization Arises from a Significantly complex system --Self organization

Self organization Astronomical phenomena, crystal structures

Resources Principia Cybernetica Web - http: //pespmc 1. vub. ac. be/SUPORGLI. html Self-Organizing Systems FAQ - http: //www. calresco. org/sosfaq. htm The Symbiotic Intelligence Project - http: //ishi. lanl. gov/symintel. html Cosma Shalizi’s Notebooks on Self-Organization http: //www. santafe. edu/~shalizi/notebooks/self-organization. html Social Computing Program - http: //www. sics. se/humle/socialcomputing/ Referral. Web http: //www. cs. washington. edu/homes/kautz/referralweb/index. html The Origins of Order and At Home in the Universe, Stuart Kauffman Women, Fire, and Dangerous Things, George Lakoff How Buildings Learn, Stewart Brand