Materializing Data Bias(8)| FMP
6/10/21–30/10/21
🤝Teammates: Tiana Robison, Jinsong (Sylvester) Liu,Luchen Peng.
Tutor John said a workshop plan as outcome is not enough. This kind of workshop is too much attention to individuals, we need to let more people involved. What is more, the outcome created from the workshop is too abstractive to be self-explained without maker ‘s introduction. Therefore, our group decided to develop one of critical objects to the level of playfulness, interaction and interpretation.
Prototype
I decided to combine my previous facial training equipment idea with Siqi Xander and Jasper’s idea(dynamic background and camera filters). With the help of my group members, we quickly prototype the new idea — an impression checker. More and more companies are using AI-hiring interviews to select job seekers. Some researches show that AI will judge applicants’ personalities and score their performance according to their facial expression, gestures, background, and their wearing, while a large number of potential gender biases , racial bias, and disabilities bias have been exposed. Therefore, the Impression Checker Machine aims to train jobseekers’ behavior to cater bias to get the job.
In oder to reveal wider databias, Syl design a data fish tank where we put a large number of interview pictures, most of which were white men.When the machine spits out the evaluation and suggestion instructions, the box in the data tank will eat the pictures to show the process of data feeding.
Making the function real
I and Tiana first asked for help from Creative Tech Lab at LCC and had the webcam, thermal printer and initial code. Sylvester focused on overcoming the key problem — how to press the one button to control different texts to be printed from the Thermal printer. In the technical part, we spent a lot of time, and its successful operation was benefit from Syl’s help mainly.
Testing
We brought the prototype to the tutorial inviting Wang, Greg and other classmates to experience it.
Feedback
Most of them thought it was interesting and could reveal bias behind the algorithm. They also provide some useful suggestions to let this experience more playful and creditable. Wan advised we could create more text according to different industries. Tiffany suggested that props need to highlight gender bias or racial bias. For instance, The machine could require users to use foundation to make their look lighter.
