I get asked what the next big thing is a lot. I haven’t had a good answer in a while. So much of what I see in technology feels iterative, or worse, derivative, especially in the social Web. All the interesting niches have been mapped out.
Lately, though, there’s one big concept that seems really exciting, and that’s anticipatory systems.
We’re starting to see glimmerings of these new, smarter systems in everything from check-in services like Foursquare to calendar apps, advertising and even online-personals services. Increasingly, rather than waiting for us to tell them what we want, in the form of a search query or command, they’ll prompt us with suggestions.
What Is An Anticipatory System?
Here’s a simple definition of anticipatory systems. Think of them as artificially intelligent services that are aware of external context — including ambient inputs like time of day, social connections, upcoming meetings, local weather, traffic and more. Taking all of that into account comes naturally to humans. But for computers, it’s hard.
The big challenge in artificial intelligence isn’t that computers are stupid. It’s that they’re ignorant. We haven’t given them enough data, nor the tools and rules to process it all. But that’s rapidly changing.
The notion of anticipatory systems in computing dates back at least to the late 1990s. Daniel Dubois, a professor at the University of Liège in Belgium, defined an anticipatory system as one “that computes its current states [by] taking into account its past and present states but also its potential future states.”
That’s a bit vague, and the practical application of anticipatory systems has proven accordingly tricky. But all of the trends we’re kind of bored with now — social, local, mobile, big data — have laid the groundwork for the realization of anticipatory systems’ promise.
Foursquare, for example, has been collecting years of data about where people are and what places they’re interested in — not just their explicit check-ins, but their local searches, tips and likes. So far, that’s allowed Foursquare to offer personalized recommendations. But now the company is taking the next step into anticipating users’ needs, Foursquare’s head of search, Andrew Hogue, told Fast Company. Hogue gave the example of giving users recommendations for lunch spots at 11 a.m., rather than requiring users to type “lunch” into a search.
That kind of ambient awareness is at the center of the latest version of a mobile dining guide made by Ness Computing. Older versions of Ness sucked in data from Facebook, Foursquare, Twitter and other sources to offer personalized dining recommendations based on friends’ tastes. The next step Ness is taking is to tailor those recommendations based on context — time of day and location. Currently in beta, the new version should come out later this month.
Merely analyzing social data isn’t enough, says Ness CEO Corey Reese: “Just because a computer is aware of what you’re doing doesn’t mean it will add value to your life.”
Anticipating Your Schedule
Schedule-management apps are another field getting reinvented by anticipatory computing, as startup consultant Semil Shah recently noted in TechCrunch. Apps like Twist and Leave Now alert people we’re meeting with to our real arrival times. That’s a welcome, computer-assisted acknowledgement of the reality that calendars are a perpetual act of optimism, subject to real-time revision by factors we can manage — like self-discipline — and factors we can’t, like traffic and transit delays.
Even our social lives are getting transformed. Consider Facebook’s “People You May Know” feature, which draws on both its own social graph of our connections and external cues like our email inboxes to recommend friends. That’s perhaps the most widely distributed and used anticipatory system in the world. Dating sites are getting smarter, too, relying on the implicit cues of self-presentation as well as explicit data in user’s searches to match up people. That’s what online daters are already doing, more or less manually as they sort through profiles — the trick is for personals sites to start doing the work for them.
The biggest bet on anticipatory computing at present is Google Now, Google’s intelligent mobile assistant that’s built into Android. Drawing on all the data Google has, from flight confirmations in your Gmail to upcoming events in Google Calendar to your history of Web searches, Google Now attempts to give you what you might search for without making you search.
Apple’s Siri, though more of a voice-command system, also has anticipatory elements. But it is hobbled by the thinness of the data Apple has on tap. If it wants Siri to anticipate our needs, Apple will have to partner more deeply with Facebook, Yelp and a host of other services so it knows more about us.
The true challenge for Apple, Google and Facebook is how to design a great anticipatory service around a specific need — without feeling creepy or, worse, clumsy. So much of what makes an anticipatory system great lies in the nuances of the service. Written prompts and design cues will play a huge role in getting people comfortable with computers that know a lot about us and make eerily accurate guesses.
But if people can get it right and design anticipatory systems that feel human and respond to our needs — well, I can only shiver with anticipation.