The SocialDining app is available for Android phones through the Google Play store at bit.ly/1rEl1jx
Researchers at the University of Colorado are working on a formula that can make recommendations for groups of people, much like those used by Amazon and Netflix to recommend products and movies to individual users.
Many businesses use algorithms to determine which advertisements to send to consumers, which TV shows users might like to watch next, and which products online shoppers might be interested in purchasing.
But those recommendations, so far, have only been for individuals. The team in the computer science department on the Boulder campus is tackling how to make recommendations to a group of people, all of whom may have different preferences.
The CU researchers currently are gathering data on how groups make decisions through a mobile application they've created called SocialDining, which allows friends or coworkers to vote on a restaurant and when they want to meet.
They're hoping to learn a bit more about group decision-making processes: Does the chosen restaurant change based on where the users are located? What if there's a strong leader in the group or someone that the group views as an expert in food? Are they friends or family, or co-workers going out after work? Did one person in the group eat Italian food for lunch that same day?
"We get recommendations from all sorts of websites," said Richard Han, one of the lead investigators on the study. "Maybe they give you recommendations for a single person, but often times you're going out with groups of people."
Any user can create an invitation to dinner on SocialDining and add friends from Facebook. The host selects several restaurants, dates and times for the group to vote on.
Once everyone has voted, the host then finalizes the plan and the group goes out to eat, Han said.
The app is available only to Android users at this point, but the team is planning to release a version for iPhone later this summer. People who use the app must sign a consent form because they are part of an experiment, and they'll get paid a small amount each time they use SocialDining.
The researchers will, with consent, collect data on users' locations, Facebook profile data, restaurants they choose, how many people get invited to outings and any comments they leave on invitations.
Eventually they hope the app stands on its own as a useful tool. For now, it's helping their research efforts by providing them with data about groups.
"Our hope is that by understanding these factors, you can develop recommendation algorithms for groups of people," Han said. "That's a challenge because of all these different factors."
Individual recommendations like those made by Pandora, which calculates which song to play next, or Facebook, which decides which ads to display on your wall, are not necessarily easy to create.
But researchers understand these types of algorithms better because they've been studying them for the last 20 years, said Mike Gartrell, another researcher on the project.
Individual recommendations are pervasive, too, Gartrell adds. Even the advertisements sent by mail from companies such as Target are uniquely tailored based on past purchases and other data.
"Large-scale, publicly available datasets for individual recommendation system research have been available for some time," he said, "In contrast, until very recently, large-scale datasets for group recommendation were not available, and none of the large-scale group recommendation datasets that have been used for recent research publications are publicly available."
The team, which sees itself as somewhat of a pioneer in this area, eventually hopes to publish its findings. Then it's up to for-profit companies and startups to find ways to integrate group recommendations into their apps, websites and practices.
These algorithms, once developed, could go beyond restaurants and apply to group movie outings, music at social events and others.
"Once what we learn is publicly available, people can develop algorithms around this themselves," Han said. "Any of the current food apps like Yelp or Urbanspoon or Zagat, they could take a look and say, 'Hey maybe we need to add such a feature and can start from what these guys did in their study.'"