Week 7 Session: The Button That Doesn't Work

Buttons that visibly acknowledge a press while changing nothing at all — real crossing buttons, elevator door-close buttons, and one you'll build yourself.

5 April 2027

A button that does nothing to the outcome it appears to control isn’t rare, though neither example below is a claim about buttons in general — both are specific, documented cases.

New York City’s Department of Transportation has stated publicly, and outlets including the New York Times have reported, that most of the city’s roughly 3,250 pedestrian crossing buttons are mechanically disconnected from the signal timing — left on the pole after the lights were automated, because removing thousands of buttons cost more than leaving them. This is documented for New York specifically; it is not a claim about crossing buttons everywhere, and many installations elsewhere do control the signal.

Elevator door-close buttons are a separate, similarly reported case — covered by outlets including the New Yorker and IEEE Spectrum: in many U.S. buildings the button does nothing, because building codes require doors to stay open long enough for boarding regardless of the button, so a working button would undercut that requirement. This isn’t true of every elevator; some, particularly outside U.S. building-code requirements, do have a functioning door-close button.

The point isn’t that buttons are secretly fake. It’s that a button can retain every signifier Week 3 named — the raised icon, the click, the light that comes on when you press it — while making a promise, in Week 4’s terms, that nothing on the other end can keep.

What a working press can’t teach you#

A button that works confirms the wrong thing: that pressing it and the outcome happening are connected. A button that doesn’t work is the only way to find out how long you’ll trust that connection anyway, and what you do once you stop.

A crossing button that doesn’t cross#

Build a digital pedestrian crossing button. On press, give it full acknowledgement — a light change, a sound, a vibration, whatever your platform can do — exactly as a real one would. Underneath, wire it to nothing: whatever “signal timing” your version simulates should run on its own schedule, completely unaffected by the press.

The suspicion log#

Have someone use it without telling them what it does. Let them press it as many times as they want, across as long a session as you can arrange. Log the press number, or the elapsed time, at which they first do something that shows they suspect the button isn’t working — pressing repeatedly and fast, giving up, pressing something else instead, or asking you directly.

Note, in one sentence, which real signifier your version borrowed from an actual crossing button — the exact detail that made it read as functional before anyone doubted it.

Keep this button#

Don’t discard it after logging the result. Week 9 comes back to this exact crossing button.