A Recommendation Engine for Politics

The ElectNext recommender takes on politics just in time for the November Elections.

3 min read
A Recommendation Engine for Politics

Like so many others in the U.S., I was glued to the TV screen during this week’s presidential debate, even though I already have decided whom I’m voting for. My time, frankly, might have been better spent reading my ballot packet. Because I have yet to figure out how I’m going to vote in some of the local races. I’ve got the school board figured out, but haven’t tackled the governing board of the local community college district yet. Or chosen my candidates for my water district; I’m not entirely sure what a water district is. Selecting local candidates, these days, means pawing through the mailers arriving at my house, reading the local papers, and looking around at the yard signs of people I know I agree with on most issues to see how they are voting.

And indeed, looking at ads, reading reviews, and checking with friends is a reasonable way to make decisions—it’s how we decide what books to read and movies to watch. But today, we have another way to help us choose books and movies—recommenders, those computer programs, on websites like Amazon and Netflix, that watch our behavior and make suggestions based on their understanding of what we like and what people similar to us like.

The technology behind recommenders has been evolving quickly; the October issue of IEEE Spectrum reports on that evolution extensively in “Deconstructing Recommender Systems” by Joseph Konstan and John Riedl. And recommenders have starting being applied in areas beyond online shopping—helping university students identify courses, for example, and helping cell phone providers figure out which customers they are in danger of losing.

And, it turns out, as startup company ElectNext reported at DemoFall earlier this month; they are also useful in helping people make political choices.

OK, that sounds a little creepy at first glance—computer software that tells us how to vote? And it could be creepy if the company behind the software has an agenda. But ElectNext promises to stay non-partisan; it will never take advertising money or sell user data. Instead, it plans to sell itself as a service to news and other web sites interested in drawing traffic.

It starts by asking users to take a stand on ten issues and rate each one on whether or not it is important. Users don’t all see the same ten questions. After each question is answered, ElectNext’s recommenders determine which question to ask next. Then, based on the users' answers, the answers of people like them, and many of the other factors that book and movie recommenders consider, ElectNext will “suggest” candidates. ElectNext, like many good recommenders, is transparent; if a user wonders why a particular candidate is recommended, he can click through to an analysis of issues on which the user and the candidate are aligned and issues on which they differ.

In my test, the first ten issues didn’t get me any useful recommendations; my answers, it seemed, were all over the place. I took the option to look at more issues; going through four more seemed enough to allow ElectNext to find its recommendations. As ElectNext builds a wider and deeper database of user ratings, this process will likely become more efficient. The website has a few other glitches; for one, the only way to get back to the home page is to log out. But it has potential.

Between elections, ElectNext could track candidates and issues for users, sending an alert when a vote is coming up on an issue they care about, letting them know that it might be a good time to write some letters to their representatives.

Right now, ElectNext is focusing on national races; after November, the company plans to build its database of local candidates, using interest group ratings, campaign finance records to infer positions based on where a candidate’s support comes from, data points gathered by analyzing interviews, and data coming in from politicians themselves and ElectNext users. So for now, I'm going to have to figure out that water district election on my own.

To try out ElectNext, go to www.ElectNext.com.

The Conversation (0)

The Future of Deep Learning Is Photonic

Computing with light could slash the energy needs of neural networks

10 min read

This computer rendering depicts the pattern on a photonic chip that the author and his colleagues have devised for performing neural-network calculations using light.

Alexander Sludds

Think of the many tasks to which computers are being applied that in the not-so-distant past required human intuition. Computers routinely identify objects in images, transcribe speech, translate between languages, diagnose medical conditions, play complex games, and drive cars.

The technique that has empowered these stunning developments is called deep learning, a term that refers to mathematical models known as artificial neural networks. Deep learning is a subfield of machine learning, a branch of computer science based on fitting complex models to data.

Keep Reading ↓ Show less