In the data collection process the Selfiecity team used Amazon’s Mechanical Turk workers to identify some qualities about all of the photos. The Amazon’s Mechanical Turk workers are English speaking individuals from other countries willing to do work and research for small amounts of money. The team randomly selected 120,000 (20,000-30,000 photos per city) from a total of 656’000. They then asked the Turk workers a simple question,”Does this photo show a single selfie”? Once they Turk workers selected about 1,000 selfies they also tagged the location of each selfie. Once the team had their selected selfies they asked the Mechanical Turk “master workers” (more skilled workers) to verify that a photo shows a single selfie, and to guess an age and gender of the person in each selfie. With the results of these photos the Selfiecity team then ran a face analysis, looking at the eyes, nose, emotion, and facial positions of each selfie. After all the photos were analyzed by the Turk workers, the Selfiecity team did a double-check of each selfie to make sure there was no mistakes, as there was a few. With the help of the Mechanical Turks the data curation process (cleaning phase) was much easier for the team. This process helped make the Selfiexploratory, like the age, gender, and the overall break down of the selfies. This process helped the data set in the end, coming to a conclusion on all the selfies. All of the credit is documented in the data set section in the project, stating every question asked by the team to the Turk workers and the response and overall process. This method process can slightly fit within DH best practices by having a collaborator, the Selfiecity team used and documented the questions and help of the Mechanical Turk workers.
