FormaCities ask residents about a street, then go quiet.

Forma is a participation tool for changes to streets and public spaces. A resident opens a link, drags a handle to see the place today and as proposed, and pins a comment on the corner they mean. Later, the same link shows the theme their comment joined, the team's answer, and what changed in the design and what didn't.

Behind it is a staff workspace where planners group comments, answer each theme and publish a revision, so residents and staff read the same record.

I built it as a working prototype around a real project: SE Hawthorne Pave & Paint, PBOT's 2020 redesign of 28 blocks of one of Portland's High Crash Corridors, with four alternatives on the table. The street, alternatives, survey counts and decision come from PBOT's reports. The resident comments in the product are samples, and it says so on every screen.

In a within-subjects test with 8 residents, 6 could find out what happened to their concern in Forma. None could with the city's survey and PDF.

The people who use a corner every day know things no traffic study does. Yet the usual way to ask them is a weeknight meeting or an online survey with a PDF attached.

People are asked to judge what they can't picture

Proposals arrive as engineering drawings, and surveys turn the confusion into a score from one to five.

The loop never closes

Comments go into a report and a decision comes out months later, with nothing connecting the two. That silence teaches people not to bother next time.

Before the interviews I reviewed what planning research already says about participation. Four sources shaped what I asked.

I wrote one observation per sticky, one colour per person, sorted all 41 by hand into findings, then ran a cluster pass over the sort to group them into themes.

• Silence reads as a no → a receipt link with no account, and every theme marked Addressed, Partly addressed, Still open or Not changed.
• One concern per person → comments start from a spot on the map, and the comparison filters by topic.
• Pictures get misread → named features on the drawing, and a text version that matches it.
• Fear of exposure → a preview of exactly what becomes public, and a team-only option.
• Replies are the bottleneck → the planner described 2,000 comments and three weeks to report. After the readout I moved the staff workspace from “later” into scope.

A resident only sees an answer if a planner writes one, so I designed for both around one shared object: a theme, a group of comments with a public response linked to a design decision.

Resident: “Show me what this means for my corner, let me say what happens to me there, and tell me what came of it.”

Planner: “Help me keep small but serious concerns visible, and publish an answer I can stand behind.”

Who sees what

Interviews made privacy the deciding factor, and a neighbourhood tool has a small-group problem: if three people near one corner use a wheelchair, “anonymous” stops meaning much. I wrote the privacy model before any screen.

I tested Forma against what the city uses today: an online survey with the proposal as a PDF. Eight residents did the same four tasks in both, half starting with each to reduce order effects. I measured unaided completion, ease after each task (SEQ, 1 to 7), and one comprehension check: could they correctly describe how the change affects their own trip?

• “Not changed” read as “nothing changed at all”. → Each status now carries one sentence, and “Partly addressed” names which part. → In a second round of 5, all 5 read the status correctly.
• The before/after handle went unnoticed at 200% zoom. → A labelled handle that starts in the middle, plus a “Compare in words” button. → 5 of 5 found it unprompted.
• “Why are there two 3s?” → Each alternative card leads with the one thing that sets it apart. → 4 of 5 told 3a from 3b.

Still open

• Numbers questions. When the question was parking spaces, the PDF table was faster. The comparison needs a plain numbers view.
• Who wrote the answer. “A person at the city, or a computer?” Each response now names its team, but trust in that answer is untested.
• Comments in other languages. One participant wrote in their first language and asked if it would still count. The product needs to say yes, and staff need a way to read it.

For a live project, the measure I'd track is comments that reach an answer: the share whose theme has a published response their author can find. Submission volume is the wrong target.