Facebook: Town Hall | 2018

How might we give people more insight into and control over their civic life on Facebook?

There's already a lot of civic content on Facebook in groups, events, discussions, content, etc. Town Hall is a hub for civic discovery on Facebook with a path to action to help people engage in a more intentional way.


Project length

4 months


1 designer (hello!), 1 product manager, 2 engineers, 1 researcher, and 1 content strategist


Foundational Research

Existing surface

Town Hall is a surface that already existed when we started, but wasn’t doing anything we needed it to for people.

We looked at research on representatives using social media, people who had used town hall before, and interested bystanders.

It was built for the 2016 US election in the US and is just a list of representatives served based on your location, and their contact info. It also has a bunch of posts the these representatives make so people can stay up to date.


The problems

  • No clear value prop or path to use this  - isn't even clear what the surface does

  • No explanation of what we're going to do with this data

  • Available actions are scary and mysterious

  • People didn't remember visiting this surface

  • People expected more opportunities to inform themselves or engage

We needed to offer value to people trying to engage with their civic lives.


The Audience: Interested Bystanders

50% of US is interested bystanders. They are interested in their civic life, but not enough to be active in an intentional way.

Interested bystanders:

  • Believe power comes from having a voice but don't share opinions

  • Believe change happens at the local level but only vote in national elections

  • Take action when the public interest aligns with self-interest

So how do we help them be more aware of their civic life and give them opportunities to get involve that appeal to them?

Interested bystander efficacy has a steep drop-off when taking action. People are

  • Blocked by non-response

  • Have difficulty finding information

  • Discouraged by high activation energy

So the more important problem is how to lower the barriers to action when we don’t even know what actions matter to each person. The first step in solving this is figuring out what matters to people, and so we chose to focus on signal-gathering.



Our tools

I did an audit of how various products were surfacing useful content. There are four signals that Facebook employs to figure out what is meaningful to people, some of which the company is better at using for people than others.

For example, Facebook is pretty good at social and local signals, but not necessarily at timeliness or true personalization.


Our goal was to double monthly local engagers.

  • “Monthly” is important because it means they are coming back

  • “Local” is important because local is more meaningful than national for every day civics.

  • “Engagers” find enough meaning to interact on the surface



What’s going on in my community?



What relates to my life right now?



What are friends and influencers doing?



What are my passions?



Knowing the signals we wanted to use to surface the right kind of information, I started wireframing how to show people the right information for them based on the signal they could share with us.

Explorations ended up falling into one of three different categories, each with their own pros and cons.

Facebook has pivoted heavily on full automation of gathering various signals and giving the viewer content it thinks they will like. While this has been a successful strategy for a long time, on civic and political topics and in an era where trust and control are very important, it’s interesting how we can think of giving people more control over what they see.



(+) Gives person exactly what they're seeking

(-) Not helpful: doesn't guide people to a solution

🚦Directory Search Bar Copy 10.png

Mixed Controls

(+) Empowers people to find their own way with some nudges

(-) Creates work for a person who may not be that interested

Full Automation

(+) Consistent with FB products

(-) Signals won't always be right and there's no easy way to correct this = drop-off


A Model for the Best Model

The second option allowed us to test out giving people control over things they’d want insight into, but it wasn’t clear what to give people more control over versus what to automate.

Knowing we wanted to minimize Facebook creepiness while still giving people a helping hand when they needed it, I plotted the signals we were looking at on a set of axes based on how creepy it is when Facebook knows something about you versus what signals each person is an expert in.


Areas of Control

Then, I thought about how much control a person would want over those signals based on the parameters. For example, if it’s something a person feels is very personal and also something they know a lot about, like their own political preferences, then it’s better to give a person full manual control over giving those signals.

For things like social signal - what a person’s friends and influencers care about - Facebook is an expert and that is one of the reasons people use the platform. So automatically surfacing things one’s friends care about would probably be helpful.


First Solution and Research

The first solution was focused on giving people control over what they actively could make changes on and making automatic what I hypothesized they cared less about.


  • I put location in a very prominent place so that people would understand it is a setting they control and feel empowered to change it at any time

  • We also added location prediction down to city level but asked fo input for anything more specific than that


The freshest information is presented first, as is information from government services and representatives a person has interacted with before.


To remove the stigma of civic/political involvement for interested bystanders, we tell the person which of their friends has gotten in touch with representatives and how to do it themselves.


I added a manual control for people to filter posts and contact information for the representatives and services. The issues we picked to pilot this filtering system was a set of 20 non-partisan categories News Feed developed in partnership with the Legal team for their political content taxonomy.

These issues were also going to influence the post feed experience behind the scenes.

The new layout relied on tabs to separate the content into particular categories, which was a precedent carried over from the old Town Hall. Surfacing a control like issues filtering was enough of an experiment that the team was hesitant to make any more changes.

Research1_Philly_Alt.2019-02-18 20_57_50.gif


At this point, I suggested eliminating the tabs for a more direct issues control experience, but received pushback on diluting the core value of the product, which was following and contacting representatives.

I was also getting pushback on introducing more content types (other than posts) for the same reason. It was a compromise for the team to add issues, but since control and transparency is what we most wanted to give people, it was worth the trade offs.



Working with a researcher, I went on the road with a prototype of this update to figure out what was working and what wasn’t for answering the question of giving people insight into and control over their civic lives

  • What did people like?

    • Controls were close to what they were trying to do, so people understood the cause and effect of using filters

    • Being able to see content by issue made them feel safe to play around and empowered to not be force-fed content, potentially fitting someone else’s agenda

  • What did they have trouble with?

    • People were excited about the concept of the surface, but the value wasn’t obvious on landing

    • The issues experience seemed isolated and fragmented not being in each tab. They felt it should influence more of the experience and that it should be personal to them

    • The contents of each tab felt static and so that people wouldn’t see the need to revisit

    • They couldn’t find how to change their location because to them location felt like a setting


Second Solution and Research

In response to the findings from the first round of research, I made some edits to the design system in between days of research to get more feedback. The biggest changes were in making information more personal and controllable.

  • I added a screen in onboarding to set the expectation for issues as a sticky setting unique to a person

  • Filtering was added to all tabs, although because of the tabbed structure and nature of the content, the mechanism was a filter for representative/service contact info (Directory) and a setting for the post feed (Explore)

  • The settings for what a person would see in the post feed was added to Settings

  • Location was also added to Settings

The new onboarding flow included a screen to choose issues the viewer cares about, which set the expectation of preferences as a manual setting that personalized all content within Town Hall.

USTH_PhillyRound2_092018.2019-02-22 08_45_26.gif


  • The idea of issues preferences as a setting resonated with people

  • The value of this surface still needed a little explaining, as the content was separated and static-feeling and people did not see it as the portal into civic life we intended it to be

  • The tabs still prevented people from using issue filtering as effectively as they would have liked but they understood where we were going with it

Another thing we noticed was that people expected to get informed about the issues they chose through content - not necessarily to see information they agreed with. Preventing a filter bubble and making sure the content was balanced and truly informative was going to need to be an important feature of this surface.


The Solution

The remaining problems

  • People feeling the content was too static

  • The tabs separating content too much and providing a fragmented experience that made having insight into one’s civic life difficult

  • There weren’t enough ways for a person to engage in this content so it didn’t feel useful to them


What We Did

This was an opportunity to de-fragment the experience by removing the tabs and making the experience modular. Modularity would allow us to order content by the stated and assumed preferences of the person viewing it, and create a much more cohesive solution.

Having put together a framework of control and signal-gathering that made sense, we were able to start thinking about what “taking action” meant to people. Action had previously meant contacting or following a representative. We considered other types of content people could engage with that could be considered a step up the ladder of engagement, and started running experiments with different content types.

We also introduced an area with better hierarchy to articulate the value of engaging with civic content and help people understand what they were looking at.

After launching this experiment, we doubled monthly local engagers, meeting our goal.


Next steps

  • Because of the ambiguity around expectations for content people would see with such high-level categories, content strategy, legal, and the News Feed team kicked off a process to develop and train more specific taxonomy of categories that better fit the type of content people will want in this space

  • We will continue developing our thoughts on what it means to take meaningful action and how to give people opportunities to take it based on where they are in their civic journey

  • We will explore a more distributed set of of interventions to find people where they are already engaging with this content and think about transparency, control, and new tools to further give both representatives and community members transparency and control

Town Hall at SXSW 2019 at a Facebook-hosted event for representatives

Town Hall at SXSW 2019 at a Facebook-hosted event for representatives