Time for Transparent Algorithms in Civic Technologies

Facebook strengthens our filter bubbles. So does Twitter. So does cable news. This has serious civic consequences. All three of these organizations are civic technologies regardless of whether they put profit or public interest first. The algorithms and business priorities of private institutions have shaped our civic reality and will continue to shape the future of democracy.

However, these organizations’ algorithmic authority does not have to facilitate authoritarianism.

As technologists we have a civic and human responsibility to assure the algorithms we create place the public interest before personal or corporate financial gain. Not every entrepreneur or technologist has to wear the label of civic technologist or social entrepreneur, but every technologist and entrepreneur who facilitates public debate has a societal responsibility to put the public before corporate profit.

What could this look like? For Facebook, this might mean solving the heavily debated fake news problem at the expense of getting fewer clicks. For Fox News or the New York Times, it might mean more transparency around their biases or the business goals that drive their reporting decisions.

As a technologist who proudly wears the civic technologist and social entrepreneur label, I have been sensitive to the effects of the public engagement rules we are creating for a while now. SeeClickFix itself was created as an antithetical reaction to a design decision by government to make service request communication opaque by default. A number of times, we’ve made decisions at SeeClickFix that favored the public interest over near-term financial gain. An obvious example is the choice to make SeeClickFix issues publicly viewable where some local officials might have preferred opacity.

A less obvious — but perhaps more interesting — example is SeeClickFix’s use of POI’s or points of interest as the primary tool for re-engaging users. In the spirit of transparency, this feels like the right moment to share the motivation behind POI’s and the algorithm itself.

The public interest decision behind POI’s

When you report a pothole on SeeClickFix, we create a small POI around that issue. The POI will then opt you into emails when other issues reported in the near future are reported very close to yours. An example POI email looks like this:

POI’s have been successful at creating the following benefits:

  1. Sense of neighborliness, connectivity and de-isolation: When you report an issue, you are reporting to your neighbors and your city.
  2. Reduction in duplicate issues reported: You don’t report something your neighbor has already reported. Doing so would create extra work for the city and for you.
  3. Encouraging volunteerism for certain types of activities like snow shoveling: If your neighbor asks for help, there’s a good chance someone will want to jump in if they’re nearby. It’s human nature.

Here’s where things get interesting. The history of POI’s is rooted in another SCF feature: watch areas. The watch area enables a user to define exactly what areas they want receive alerts about by drawing polygons on a map. A later version of the watch area allows a user to follow existing boundaries created by neighbors, geo-political boundaries, or neighborhood boundaries drawn by Zillow.

This feature is still available and used today, but SeeClickFix does not opt the user into it or prioritize it when it comes to engaging users. When it comes to opting users into reengagement algorithmically we chose the POI over the watch area. We made this choice in an attempt to create more diverse conversations and promote higher levels of local empathy. It was a small way we tried to place the public interest over our own goals of engaging more users using easier, pre-existing neighborhood boundaries.

Let me explain.

The POI algorithm assumes ‘your neighborhood’ is the place YOU care about because you live in it, walk through it, or interact with it on a daily basis. In contrast, the watch area assumes your neighborhood is defined by housing economics as strengthened by Zillow or political boundaries as defined by gerrymandering and politicians. How you define your neighborhood — and as a result your neighbors — is of the utmost importance to society because the neighborhood is the original filter bubble. Who you are physically near or far from has historically shaped our perspective and our world view. Prior to the internet, physical infrastructure was the primary way of strengthening or penetrating those bubbles. A highway or train tracks have the ability to separate two neighborhoods within 300 feet of each other, while connecting neighborhoods 300 miles away. On the flip side strong public transportation penetrates filter bubbles by forcing disparate populations to stand inches from each other.

With the advent of social media, algorithms, and other technologies, neighborhood social networks should push you to engage with the people you are near on a daily basis, and not just the people who own a home next door or live in the Zillow shapefile created for your neighborhood. Aside from being more contextually relevant and more interesting to most people, it is the socially responsible thing to do. Using a contiguous group of radii based on where you have expressed geo interest pushes you to interact with others who share similar geo interests but don’t necessarily live in your cul de sac. (Side note: the cul de sac is another great example of public infrastructure that creates physical filter bubbles.)

I walk 2 miles to work each day through multiple neighborhoods traditionally delineated by politicians and realtors but I personally think of all of them as my one, singular neighborhood. This is the neighborhood where I document things I’d like to see improved on SeeClickFix (my POI’s). As a result, I consider all the people who live on that 2 mile stretch my neighbors. You can imagine that I am having much more digital heterogenous social interactions than if I were only communicating with the people live in the upper middle class neighborhood where I live, whose boundaries have been defined by politics and city planning, and reinforced by realty.

As a vocal civic technologist, I’ve ensured that my company has the support of investors, employees, and users in making choices that place the public interest before near term profit. Wearing the civic technologist badge helps hold us accountable and set expectations with affected parties up front. But a company does not not need to wear that badge to make the right decisions for the public interest.

Lastly a call for transparency in algorithms

If we are going to put the public first, our algorithms should not be proprietary. As a first step for all companies facilitating public dialogue, I would suggest that we make our algorithms and our biases public. Users of information platforms deserve to know the decisions that are being made under the hood to keep all of us informed. I know that others have called for this recently and I’d like to take this first step by sharing our POI algorithm.

max_distance = 0.003
 Poi.where(Scf::Geo.dwithin_arel(arel_table[:point], issue.point, max_distance)).
 where(“issue_id <> ? OR issue_id IS NULL”, issue.id).
 where(“created_at > ?”, 1.year.ago).
 order(“id DESC”).

In English: “Send an email to any user who has commented, reported or voted on an issue within the last year within 375 Meters”

It’s nothing particularly brilliant or exciting, but there it is. Please don’t hesitate to to tell us how you think it could create more public benefit.

This article was originally published on Medium.