The Beginning of Anticipatory Interfaces?

Ready for a new buzzword? How about anticipatory interfaces?

I’ve seen a really interesting feature now in two different applications—systems analyzing your text as it is being typed. Let me repeat that: as I am typing information, things are happening—on the fly—that might be of use to me.

In the case of SystemOne, adding to a wiki entry serves up relevant content—without ever hitting a submit or save button. Here’s how TechCrunch describes SystemOne:

Essentially, it’s a wiki that analyzes what you are writing in real time and offers up related search results from other pages in that wiki, the web in general, your uploaded OPML file of RSS feeds, your emails and any files the system is given access to.

The second example is Stikkit, which I just stumbled across. Whereas SystemOne scans keywords to find relevant content, Stikkit tries to use natural language recognition to ‘guess’ what type of content you are adding. So, as I add text—things like ‘meet Erin for lunch’ or ‘pick up dry cleaning’—Stickkit is parsing this information and trying to classify the text as an event, to do, contact, etc. Check out the Stickkit demo movie to see what I mean.

The potential uses of this type of ‘interaction’ are…intoxicating. These are the beginnings of systems that can in effect ‘listen’ as you’re typing and anticipate your next action.

I’m having fun imagining how this might be used in the travel industry—or any industry for that matter. Imagine entering data for a client (CRM, client intake forms and such) and having this sort of ‘real-time conversation’. No submit. no save buttons. No searching for relevant information. All the information you might want to pull for a client is there, coming to you. And the next steps that follow the human conversation? Booking travel. Canceling an order. Looking up tracking information. These actions—in the form of links, buttons, steps, workflow, whatever—are anticipated and brought into focus, filtered out as choices if they don’t apply, or submitted (if that looks to be the next likely course of action).

Sound too far fetched? Having worked on a project that used natural language recognition to interpret dates, I will admit that language is a quirky, complex thing, fraught with unexpected frustrations. Are we going to meet ‘this Monday’, ‘next Monday’, just ‘Monday’, or ‘the Monday after last’? And this doesn’t even consider global systems that support multiple languages! Put that aside though. Combine natural language recognition, complex search/filter algorithms, and eventually human intervention, yes, human intervention—and you have an even more promising technology.

I mention this last bit about human intervention as there is a popular article on Web3.0 making the rounds right now, describing a sort of ‘pseudo AI’ that emerges when people augment systems, as is the case with Amazon’s Mechanical Turk and Google Image Labeler. See, had I been using an anticipatory interface to type that last sentence, I wouldn’t have needed to search for those links!

Good interface design is like a conversation. I guess these kinds of interfaces actually listen to what you’re saying typing!