In recent years, technological bias has been showing up in the news more frequently. Problems like gendered voice recognition software are often presented as new issues but these problems have much deeper roots.
I recreated the earliest magnetic recording experiment and created an interactive exhibit that allowed visitors to perform the experiments and experience early voice recording bias.
I talk more about these experiments and voice recognition bias in my article titles Are You There Google? It’s Me, a Woman.
There is very little available historical information about the first magnetic recordings so I prototyped the device based on this illustration by Marvin Camras. The goal was not to exactly replicate the conditions of the first magnetic recordings but rather to imitate them in order to understand the contingencies and biases that surrounded the experiment.
Using Marvin Camras’ illustration and historical user descriptions, I began to map out how the experiment was carried out. In 1898, Valdemar Poulsen used piano wire, an electromagnetic trolley, and rewired telephone parts to magnetize the wire with his voice. Holed up in a small cabin, he recorded, replayed, erased, and rerecorded the word “Jacob.”
Poulsen ran a trolley and an electromagnet along a wire. He attached a telephone transmitter to the trolley to record his voice and listened to the recording by attaching a telephone receiver. To wipe the recording, Poulsen ran a permanent magnet along the wire.
Once I understood how the experiment was carried out, I worked with my partner, Katherine Goertz, to assemble a rapid prototype of the mechanism. We laser cut the trolley, rewired old telephone parts, and tested the recording mechanism on mounted piano wire.
After building a functional recording mechanism, I began to design the exhibit. On the far wall, there was a mounted telephone, which could be dismantled, inspected, and reassembled. The trolleys were mounted on both sides of the room and on the left side of the room, eight sequential photographs showed how the experiment was carried out. These photographs were the only instructions at the exhibit because I wanted to encourage users to test the mechanism in their own way, uncovering any biases against their natural inclinations.
The exhibit revealed that a person’s height, tone of voice, speech volume, pitch of voice, walking speed, talking speed, stature, and accent all affected the way sound was recorded on the wire. Most notably, lower pitched voices produced a recording while higher pitched voices did not. Because the first magnetic recording experiments were carried out by one man, all subsequent iterations of this technology are based on one voice. So, voices that are different than Poulsen’s will record on the wire in drastically different ways and this is still true for current recording technologies.