AI Bias is a small research project, which uses a fine art photography context to explore questions of AI bias in search engines. According to our research, image recognition AI produces more biased results, thus, we selected three of the world's most popular search engines, Baidu (China), Google (U.S.), and Yandex (Russia), and used their image recognition searches to test for AI bias. The project applies a typological approach, presenting AI bias through an objective lens for the viewer. Twenty-two participants, including members of our group (Tima, Lei, Tutu, Zihan), were asked to take their portraits in three different dress styles. Their portraits were put into three search engines to get the results. Of the results found, we selected the first four images from those and edited them to 1" size. We logically presented their results in four rows (Baidu, Google, Yandex, mixed). We expect the audience to recognise for themselves the AI bias in those search engines as the images are compared in the zine.

As this was a small study, we were not able to reach major conclusions. We managed to find major trends such as long hair giving female outcomes, and accessories such as large chains and caps tend to give outcomes that are leaning towards dark-skinned people, or famous actors. If we were to repeat this study at a larger scale, might put search settings for medium to large file sizes in order to get higher resolution files.