Voice assistants are already everywhere. With Amazon leading the pack. And while still a bit gimmicky, their influence and capabilities are only growing.

But to process our language, the microphones are always listening. So we’re not in control of what parts of our lives are recorded, stored and analyzed.

So how to solve this technical dilemma?

The idea was born in late 2017 in an interaction design class at FH Potsdam.

My original plan was to use voice interaction to influence the growth of a plant-like structure. But soon I was more interested in my fellow students’ concerns about Alexa and their sense of losing agency.

So I thought: What if there would be a physical barrier around Alexa? And what if this cage would only be opened and closed by a second voice recognition? One that’s independently working offline – and therefore is out of Amazon’s and others’ reach.

A simple offline voice module opens the cage to a range of new wakewords like “hey wiretap” or “hey nsa!” – so the user is aware of how to talk past this point.

The first cage is a biscuit box and not soundproof at all. But I chose to focus on the interaction first.

Alexa’s physical action key is hotwired and reacts to the new wakewords.

I currently work with hobby.denkwerk and thinx lab on version 2.0