Is Google Instant a Spivey experiment?

Google have added a feature which allows search results to be returned as you're typing a query. I'm not sure that this is immensely useful, although it's an impressive technology demo showcasing how fast their system runs.

One thing which occurred to me is the question of what, if any, additional information this sends to Google and how they might try to monetize it (attempting to adopt the mindset of someone trying to sell advertising). In this instance each keystroke probably is invoking a query, or maybe the system just polls the query box every few hundred milliseconds - probably the keystroke event would be more efficient in terms of coding and processor usage than continual polling. This could convey - directly or indirectly - time stamped information about when keys are being hit, in addition to the partially typed search query itself. I think something similar occurred with the now defunct Google Wave, and possibly they might have reused the same code.

Does the temporal sequence of keypresses convey any information about the user or their intent? According to a company called Digital Proctor, it does. I'm not sufficiently expert in this area of psychology to know whether Digital Proctor represents real innovation, or is merely snake oil/pseudoscience, as many psychometric type tests are, but the notion that you might be able to identify someone from their typing style doesn't seem completely unreasonable. Both handwriting and speech are really just motor patterns, and those are relatively unique, so typing patterns might be also.

Here's the elevator pitch:



If their claim is true then possibly Google may be able to identify users, even if they're not logged in, via their typing style. A more likely scenario is that typing style is not completely unambiguous but can be used along with other weak classifiers such as browser info to make a strong one.

But even if typing style only provides such a weak signal about the identity of the user that it's not worth considering, other information may still be latent in the semantic-temporal pattern. Are there systematic relationships between the semantic content of the query, or the intent which the user has, and their physical movements? According to the psychologist Michael Spivey, there might be.



Spivey describes his ideas in a book called The Continuity of Mind. His basic idea is that there are multiple cognitive processes going on in parallel which are in some sense vying for control of your motor output, and that this dynamic struggle is to some degree apparent in your movements. In his view the decision of which action to take isn't the final output of a sausage machine but is created and modulated in flight by multiple other processes occurring simultaneously. This is quite consistent with a Brooksian notion of cognition as multiple parallel streams rather than a strictly sequential or "club sandwich" cognitive architecture.

So maybe with enough mining of a large semantic-temporal database it might be possible to extract any systematic correlations between search queries or particular words and temporal typing patterns, and this could be saleable to advertisers or useful for displaying contextual adverts. Whether Google realize this or not I don't know. Maybe they just wanted to show off the speed of their search engine, and hadn't considered that keystroke sequences might contain additional information.