The creepiness of the Internet is all too familiar by now: Facebook ads cater to you on very personal levels, Amazon knows exactly what your next purchase should be, and Netflix knows more about what you love than your immediate family members do. Whether you think these predictive algorithms are amazing or monstrous, the invasiveness of the web will surely only grow as technology progresses. One of the latest developments in online data mining is eBay’s acquisition of Hunch.com, a website that specializes in Internet personalization.
Hunch states its mission is to build a ‘Taste Graph‘ of the entire web, using an affinity-based algorithm to predict what different kinds of people will like. If you’re looking at the website’s infographic right now, you might be feeling a bit frightened, overwhelmed, maybe you’ve even wet yourself a bit. But you shouldn’t be too stirred up, because as previously stated, recommendation technology is nothing new; whenever you rate something on Netflix or ‘like’ something on Facebook, that information is being tabled to create maps of similar items and similar users, in order to serve you up more potentially likeable things (and of course to sell to advertisers).
The Taste Graph however seems more powerful than typical recommendation systems, because it gathers information from a much more vast array of sources: ‘Teach Hunch About You (THAY)’ questions answered by users, item ratings, and data from users who have linked to Hunch through social media (‘likes’, Four Square check-in data, social connections from both Facebook and Twitter). The Taste Graph currently has a mind-blowing 30 billion connections representing the affinity between people and items. The 48 processor super-computer takes two whole days to refresh.
It’s easy to see why auction giant eBay would be interested in Hunch, as better product recommendations mean more purchases. eBay already had a product suggestion algorithm, but what Hunch co-founder Chris Dixon and eBay chief technology officer Mark Carges feel Hunch brings to the picture is an improvement in ‘long-tail’ shopping. Users of eBay should benefit from suggestions that aren’t obvious, as Dixon explains that if a collector is using eBay to search for coins, Hunch may be able to retrieve relevant products that are non-obvious, such as a microscope good for analyzing coins. The typical algorithms used for recommendations wouldn’t be able to make the same kind of connection. Dixon has said that the company’s relationship with Ebay began after Hunch started letting other companies use their Taste Graph. Yes, third-party sites and applications can tap into Hunch’s application programming interface (API) and use its data, much like the way Facebook apps can access user information.
But wait, it gets creepier! When I first learned about Hunch, my stalker-senses tingled with interest. Perhaps now I could finally woo Daniel Tosh by mailing him the perfect present after answering Hunch’s questions based on his own opinions and affinities, all of which I know by heart! Well, as it turns out, Hunch already has an application for gift solutions, which allows you to input anyone’s Twitter handle to generate product ideas. I toyed with the app, but it seems like it needs more developing; after inputting several of my friends’ handles, as well as those of celebrity figures ranging from Tosh to Herman Cain, particular products appear on almost all of the gift lists, most notably the ‘Ultimate Lego Building Set.’
In addition to Gift-o-matic, Hunch offers other apps including a correlation game, in which you are asked whether you think people who do/like a certain thing are more likely to do/like another kind of thing. According to the Taste Graph, people who use gel-pens are more likely to drive alone to work, as compared to pencil users who are apparently more likely to walk or bike. Another app shows you statistics of different celebrity-followers on Twitter. For example, followers of Karl Rove are more likely to view Rush Limbaugh as a voice of reason, have white hair, and dislike the art of Andy Warhol. No real surprises there. Followers of BreakThru Radio on the other hand are more likely to enjoy Scorsese films, buy from farmer’s markets, and be able to explain ‘Festivus’ (good going guys!).
While these apps, and the concept of Hunch itself, are framed as a fun and unique improvement on our experience of the Internet, I myself wonder about the potential nefarious uses they can be put to when placed in the wrong hands. Security analysts have already expressed concern for data mining techniques of sites like Facebook, Google, and now Hunch. Richer personal details are lucrative for advertisers, but also for identity thieves, cyberspies, and parties who may unfairly use the data against consumers, such as insurance companies, employers, and politicians. It remains to see whether the predictive powers of Hunch and eBay are worth the privacy risks, but one thing is certain: there’s no way Herman Cain would buy this shirt.