Shedding new light on mobile communitiesHP Labs debuts Gloe, a versatile new location-based mobile recommendation platform
By Simon Firth, June 2010
HP Labs Researcher Thomas
But what if you’re mobile and looking for a great spot to fish, a place to go birding, or to cycle off-road? Or maybe you’re in San Francisco’s Mission District searching for murals, or in Barcelona wanting an architectural tour of Gaudi’s home city.
Communities that could help you do any of these might already exist on HP Gloe – a new location-based mobile recommendation platform built by researchers at HP’s Social Computing Lab and which was publicly announced just recently.
And if they don’t exist, Gloe makes it simple to just go ahead and create them yourself.
Find and recommend anything
“The main purpose of HP Gloe is to make it really easy for people to find, create or recommend web content in the mobile setting,” says HP Labs researcher Thomas Sandholm, and Gloe project lead.
Gloe, he explains, is accessible via any mobile web browser as well as through dedicated applications for Android and Blackberry smart phones (with a WebOS prototype already developed).
It uses a phone’s GPS system to pinpoint exactly where you are and then lets you search for nearby categories of information. Those categories could be conventional items like restaurants, hotels or nearby gas stations. But Gloe also lets people create their own categories, so you might just as easily find nearby hiking trails, swimming holes or small urban parks.
Not all the information on Gloe is user-generated. The service aggregates over 7 million listings from established online services, as well as location-specific Wikipedia entries and geo-tagged photos from sites like panoramio.com.
To help you sift through all that content, Gloe turns to your friends. Any of your Gloe friends can recommend any item to any other member, as can you. The upshot: wherever you are, Gloe can offer you a list of whatever you’re looking for (lakes, hiking spots, hotels) sorted not only by proximity but also by ratings given by people you trust.
Customize your recommendations
Recommendations on Gloe are ranked by the number of ‘votes’ users have given a particular item, a figure that’s always visible next to the item when it appears as a result.
“A big part of the idea with this service is to make the search results completely transparent,” says Sandholm. That’s unlike results from search engines like Google, where the rules determining which results rank highest are at best opaque.
It’s also possible to select a subset of friends for any query on Gloe and see only what those particular friends voted for – choosing just your fishing pals, for example, when searching for a nearby place to cast a line.
The value of true and reliable opinions
The methodology Gloe uses to allocate votes is significant, Sandholm notes. Most ratings sites make no distinction between generous and spendthrift raters, for example. They also give users an unlimited number of votes, which offers little incentive to be really discerning about the number of votes they award each time.
“What we wanted to do was to create a voting mechanism that makes you vote in a way that corresponds to your true opinion,” he explains.
To that end, each Gloe user gets a limited ‘budget’ of votes, a strategy suggested by previous HP Labs research that has demonstrated how economic mechanisms can be used to incentivize users to reveal their true opinions.
Voters who introduce new items on to Gloe that are subsequently rated highly by others are also rewarded with extra voting points in their budgets, says Sandholm. That too, he notes, “helps us keep improving the quality of what Gloe covers.”
“A lot of people will go onto sites like Digg and YouTube and consume the content but not rate it,” Sandholm adds. “And that’s a huge untapped knowledgebase of people who actually have opinions but don’t put them in their system.”
Gloe, in contrast, aims to promote a virtuous cycle where true – and therefore reliable – opinions are both encouraged and maximized, which should lead others in turn to both use and contribute to the service themselves.
Refining the voting mechanism
The charter of HP’s Social Computing Lab is to investigate new ways of extracting knowledge from crowds of people and of aggregating that information into recommendations or concise pieces of information that are both simple to understand and easy to use.
In that context, says Sandholm, “Gloe is an example of a service that explores the ways in which people actually use social media, and asks how we can simplify that process by doing filtering based on your current context – the context here being geo-locational as well as social.”
The Gloe team plan to keep tweeking their voting mechanism so as to maximize its value to users, but they’ve already found that the service encourages people to rate items much more often than they rate videos at YouTube.com, for example, or news articles at Digg.
A versatile and open-ended platform
Much about Gloe has remained theoretical until recently when, after a year in incubation, the site was publicly announced.
“We’re now really looking to the crowd to see what content they recommend,” Sandholm says.
To that end he’s hoping that as many people as possible will try Gloe out. “More users would allow us to come up with and test more sophisticated mechanisms to identify and incentivize the users who make the most valuable contributions,” he suggests. “And those lessons could be applicable in a broader social media context and would be valuable both to our research and to HP in general.”
Unlike many such services, Sandholm points out, Gloe really can be anything you want it to be.
“Because Gloe’s an open-ended service, it’s really a platform for anything people want to be able to track with their friends,” he says. “We hope people will just start using and applying it to whatever it is they really enjoy doing out in the world.”