Everyware, Findability, and AI (Part 2)

Part 1 promised that Part 2 would discuss Greenfield’s Everyware. However, before we get to that discussion, a few considerations on Moreville’s Ambient Findability are needed. The discussion of Moreville’s book will make clear the contributions offered in Everyware.

Greenfield and Moreville express skepticism about the ability of artificial intelligence to solve basic problems related to ambient findability and Everyware, what Greenfield terms ambient informatics. As more and more ordinary devices are available for people to engage as they go about routine activities, the sheer challenge of finding the right device among those available to support an activity promises to develop into a significant hurdle. Both authors recognize the challenge. Yet, Greenfield and Moreville both fail to discuss straightforwardly the challenges faced by attempts to manage relationships between connected devices.

As I noted in a previous post:

Peter uses a “wayfinding” metaphor to develop observations on the relationship of ubiquitous computing and user experience. Peter admits to the limitations of the wayfinding, or navigational, metaphor, in the sense that the web is not in fact spatial. However, he contends that making things findable means classifying information using controlled vocabularies, and
developing ways to retrieve it. In other words, if there are no paths to retrieve what you want, or go where you want, then of course you cannot get lost.

Moreville is a strong proponent of using well-structured meta-data to classify objects, though the key focus is on documents. After all, he is a key author and visionary in the field called Information Architecture. Moreville observes that the development of ubiquitous computing (ubicomp) challenges the use of structured descriptions using taxonomies or ontologies due to the overwhelming scale of information sources. He notes,

This is the paradox of ambient findability. As information volume increases, our ability to find any particular item decreases. How will we Google our way through a trillion objects in motion? (page 86)

Indeed, as more and more objects are connected using near field communications (RFID, Bluetooth, etc.), WiFi, or GPS, the ability to locate objects depends less on taxonomies or the ontologies of the Semantic Web, and more on folksonomies. Folksonomies develop as people create tags for objects. The objects could be photos, online documents, academic journal articles, historical sites, coffee pots, refrigerators, or any number of things people want connected for Internet access or social networking. Therefore, the threat of information overload seems, well, overwhelming.

Moreville vacillates on how ambient findability is possible in a world where any object can offer itself for engagement with embedded RFID chips or other near field communications sensors, for tagging or reading tags to see what other people said about an experience with that object. Developers working on how to manage ubiquitous computing environments ultimately are forced to take a position on whether artificial intelligence can manage the threat of information overload portended by ubicomp environments, where many proximate and distant objects are tagged and available for engagement by people using mobile communications devices or standard devices like desktop computers. Moreville says at one point that we shouldn’t expect artificial intelligence to offer a solution noting,

Despite the hype surrounding artificial intelligence, Bayesian pattern matching, and information visualization aren’t even close to extracting or understanding or visualizing meaning. For as long as humans use language to communicate, information retrieval will remain a messy,
imperfect business (pages 53-54).

So, good enough, it looks like Moreville is realistic about the prospects for ambient findability. Yet, he then returns to his information architecture box, claiming that the combination of Semantic Web tools provide a foundation for taxonomies and ontologies to manage the semantic meaning of connected objects as an interface to infrastructure, i.e. ubicomp environments. Moreover, Moreville thinks folksonomies provide the flexibility for this architecture to provide for user feedback. But, the reader asks, how?

Moreville’s answer is to fall back on the same old source of magic that he earlier correctly construed as a dead end, i.e. artificial intelligence. He starts by recognizing that the cognitive model of AI starting with Herbert Simon is limited, but even in that part of his discussion absurdly states that, “naturally, the roots of AI, and the big wins in expert systems and game algorithms, are flush with decision trees” (pages 155 – 156). The reader asks, what big wins for expert systems? And, the only answer Moreville gives is in chess playing. Moreville knows more about the limitations of expert systems than his writing in Ambient Findability expresses. However, the coming tide of information overload from ubicomp environments pushes him to grasp at some straw that promises to organize the relationship between emerging semantic networks of meaning growing from object tagging, folksonomies, and the well-structured foundations of information architecture available from taxonomies and ontologies. He notes,

We’re starting to understand the pathology of information consumption and its long-term effects on decisions, thanks to novel insights flowing once again from AI (page 167).

Moreville falls back on the approach to intelligence by Jeff Hawkins as a promise for how to manage the information overload portended by ubicomp. Hawkins believes that if we can figure out how the brain itself works, then we can really understand intelligence and build machines that exhibit it in the same way as humans. The idea is that the key to intelligence lies in the way we retrieve answers to questions and solutions to problems from our memories stored in the brain. Well, duh…how can that approach come close to explaining where those memories come from? Memories and meaning come from human experience, and experience is the wall that artificial intelligence keeps bumping up against as its proponents continue milking those federal grants and stockholder dollars in their Sisyphean quest.

So, after all the twists and turns of Ambient Findability, the book ultimately folds up with little solid indication of how to proceed in designing for experience in ubicomp environments. Indeed, the book could probably have ended on page 70 with little lost in its contribution to people thinking about its topic. Moreville notes,

Visions of pervasive computing and ambient findability ignite our imaginations, but we’re a far cry from best practices for everyware, and the road ahead is neither straight nor narrow…As we wander the wilderness of ubicomp, our mobile devices will be our lifeline, connecting us as never before: indivisible and intertwingled (page 70).

Fortunately, Adam Greenfield’s Everyware provides us with some useful insights into criteria to consider using as best practices for ubicomp environments. Stay tuned for Part 3

Copyright © 2007 by Larry R. Irons

Share this post…

add to del.icio.us : Add to Blinkslist : add to furl : Digg it : Stumble It! : add to simpy : seed the vine : : : TailRank : post to facebook


Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s

%d bloggers like this: