A New Way to Google: the Prompted Assistive Search

#rightproblem #constraints #analogies #meansends #gapanalysis #abstraction #heuristics #dataviz

So I got to visit Google in SF a few days ago. Indeed I ate their deliciously free food, marveled at their hip office and gorgeous view of the bay, and chatted product ideas with a Google employee. But I’m not here to talk about that, I’d much rather tell about an odd concept I came up with, a new way to search the web, the Prompted Assistive Search (PAS).

Google, like any advertising revenue funded company, is always looking for new users. Thanks to cheaper and better technology giving internet access to those who never had it before, billions more users will soon be joining the masses on the world wide web. This opens up the market again for larger companies who want to get these users to use their products. They will use the internet for the same core purposes we do: information, social connections, and entertainment.

However, these upcoming users are also quite different from us. They don’t have the luxury of 4G LTE data speeds or shiny iPhones, in fact, they don’t necessarily even have reliable phone coverage and for the most part will be using cheap and old smartphones, ones we might consider quite worthless. Biggest of all, most of them don’t speak English which most of the internet is composed of.

Languages Used on the Internet:

These constraints make developing a product for this new market a bit harder, and they demand that a new and original solution be made, a totally different product designed for a totally new group of users. But what’s the main issue, what do people need?

They need answers, and the internet can give these for virtually any question, but the problem is being able to parse through the mass amount of data amounted from just one search, much less be able to read and understand the answer once it is found. That search bar everyone is so familiar with hasn’t changed since the very beginning. Upon first sight, it is a bit intimidating, acting as a window to an infinite world, and you have only a couple words to narrow trillions of bytes of information. What if this was different? Could we reimagine the searching process to not be so ambiguous and open?

Ever played the game 20 Questions? The idea is that one person thinks of a object, and the guesser has 20 yes or no questions to ask in order to deduce the answer. The key to the game is to ask the right questions, not to be too broad so as to eliminate no choices or too narrow as to only eliminate one thing at a time. Computers are great at this. They can use learning algorithms to get it right almost every time, after they’ve had a couple thousand people go through and “teach” them the paths to the correct answer. They do this by asking easy questions, each answer acting as a step that gets just a bit closer to the solution by eliminating other choices. When the computer gets something right, it gives that answer a higher weight during that situation. If it gets something wrong, it decreases the weight of an answer for the situation. The computer is more likely to pick solutions with higher weights.

A new solution that can identify what users are really trying to find would help alleviate the overwhelmingness of searching and would be similar to the game of 20 Questions. However, unlike the game, the users would not know the exact answer either, only what the search results give them. Therefore, the algorithm would have to really leverage data collected by asking(current emotions) with data collected without asking (location) in order to find the solution that would most satisfy the user.

Learning algorithms are key here, and Google has a lot of data at its disposal to identify what users like, that’s how it tailors ads to people. Google Translate is a similarly great advantage, allowing this prompted search process to work across languages and integrate nicely across the web. Let me be clear though, this isn’t the new Siri and so it won’t be colorfully animated nor will it boast a sexy voice. PAS’s purpose is simply to guide those who aren’t yet experts at searching towards the information they want, the information they need. The real work would be done on outside servers, and the user’s experience would be optimized for available bandwidth. If successful though, PAS wouldn’t just help developing communities, but might just be able to teach anyone what a search needs to be successful and even provide truly invaluable data about how we learn as humans.

If there’s one company with both the resources and ambition to reimagine the search (again) it would have to be Google. The question then is, do they have a reason to?