Ben Kraal

Larry Irons has been slowly reviewing Adam Greenfield’s Everyware and Peter Morville’s Ambient Findability and drawing links back to the long tradition of critiques of AI, right back to Lucy Suchman and Hubert and Stuart Dreufys.

He says that visions of ubicomp “where people, devices, and computation are seamlessly integrated into everyday life” are characteristic a utopian dream of a perfect seamless infrastructure. He writes:

The quest for seamless interfaces for ubiquitous applications trades on the same sleight of hand as many artificial intelligence applications in the past.

That “sleight of hand” often takes the form of the assumption (or promise, as Irons would have it) that agency can be imbued in devices that will substitute for people in certain situations, for example “I’m late for a dinner. My [ubicomp device] knows this (because it has my calendar, my location, and the traffic status). So, it tells me, and then it alerts the people who are waiting for me.” (example from Seth Godin’s Web4 post)

On the other hand, seamful design of ubicomp systems not only recognises the impossibility of a completely seamless, invisible, ubicomp infrastructure but embraces the messiness of everyday life.

Yvonne Rogers recently critiqued “Calm Computing” which is a flavour of seamless ubicomp and noted that ethnographic studies of how people live their lives have shown that day-to-day existence is so varied that it is difficult “if not impossible, [...] to make sensible predictions about what someone is feeling, wanting or needing at a given moment.” (p405)

Rogers goes on to suggest that while UbiComp can still retain Mark Weiser’s vision of “calm computing” it can also embrace “engaging user experiences” that recapture the “excitement of interaction”.

Rogers again:

To make this happen, however, requires moving from a mindset that wants to make the environment smart and proactive to one that enables people, themselves, to be smarter and proactive in their everyday and working practices.

Seamful computing and engaging user experiences offer two great advantages over seamless ubicomp. First, they don’t require solving AI-Hard problems related to the perfect capture of contextual information in situations involving human behaviour. This is a Good Thing(tm) as simpler techniques are more robust and transparent. Second, they are based on an understanding of existing human activity. Basing new designs for ubicomp on existing practices is not a Bad Thing(tm) in the slightest as new ways to use seamless ubicomp (or everyware, if you like) will arise spontaneously and with greater variety and creativity than can be baked in at the time of their creation. The myriad ways that web tools with open APIs are mashed up shows how this happens already.

Indeed, the greatest advantage of seamful, messy ubicomp, is that the creation of new and novel integrations of everyware technologies are more likely to arise than in the seamless “calm” version.