Mix of Instagram, Affordances, Social Media, Enthographic Research, Face Analysis, Human Labour, Data Visualization?

Now that we are wrapping up the semester, I would like to share an exploratory data visualization project of 'selfies across 5 cities'.

I watched a video of Moritz Stefaner on Vimeo who explains the process of this study: their team collected 656,000 images on Instagram between December 4 to 12, 2013, in New York, Sao Paulo, Berlin, Bangkok, and Moscow and pinned it down to 640 selfies from each city to compare and contrast differences mainly between age and gender.

(You can check out the video here: https://vimeo.com/87797013)

He recognized that people have been taking self-portraits for hundreds of years; however, he questions how the practice has been changing, as the topic of ‘selfies’ explode across social media.

In the introduction of his talk, he referenced images of “duckfaces, Justin Bieber, Kim Kardshadian, blondes in bikinis [and] the whole shebang” to foreground the top Google Images that appear when searching "selfies" on google.
He describes the act of selfies in a creative way and gives "respect" to some creafully crafted and hilarious selfies.

 

 Then he turned to “less respectful” selfies including ones that was captioned “funeral selfie”. The most salient to me was the one in the center of two younger children taking a selfie and smiling in the backseat of a vehicle.

I had some issues with the actual methodology of the study which I would like to share with you.

To narrow down all these images on Instrgram to only ‘selfies’, Stefaner and his team put all pictures on mechanical turking websites and had people identify whether or not it was a selfie. Stefaner claimed that if his team only followed the hashtag #selfie on Instagram, their data set would be limited to only those who used the hashtag itself. Through this method they found that 3 to 5 percent of all images were selfies.

They then used heuristics to determine roughly the age and gender of the represented participant in the image, where 2 to 4 people had to agree on the category each image was labelled as.

To have an accurate representation of the demographics and the gender included in these images, I would argue that could be hard to determine by just looking at their face in the represented participants.

They then ran a face analysis using different algorithms to indicate if people are smiling, looking up or down, which only can result in "some degree of accuracy"

He even touched on how “creepy” the images looked when complied in a large data set when zooming in on just the images of the faces for the algorithms and for his team to spot patterns in the image collection.

He pointed out the small error in which a ‘stuffed animal-like figure’ was included in the data set and claimed that this was a way to help him and his team “feel better” about the mass surveillance project that they were conducting. – somehow this eased their speculations of conducting this massive collection of personal data

When asked about the methodology and the purpose of this project, he said that it mixes art, practice, human inspection, theory, and science.

Although I did enjoy the uniquely dynamic approach to collecting and sorting information, I had many issues about the accuracy of both the algorithms and human error. Both Striphas and Gehl suggest that algorithms can amplify and reiterate human errors and ways of seeing.

I have an issue with the human labor and turking site that was used to analyze this data, within what I am assuming were very exploitative environments. Fuchs draws on the idea of turking and how it underpays users by offering very little compensation for their time and efforts.

I also thought that by conducting this type of experiment, the team fails to address the purpose of such an intrusive project which takes these selfies out of the context of their post in a “mass surveillance” type of project. Is this a fair and ethical way to appropriate their publicly posted selfies on Instagram? This draws on at least half the readings we have discussed in class about the ethics of data collection and mass surveillance.

What about the affordances of Instagram itself? Does that not have an impact in the way these selfies were taken and posted, as opposed to a different platform? What does that do to the data?

What about the “respectful and disrespectful” dichotomy he outlines in his introduction – how can the practice of taking a selfie demonstrate how social and cultural practices are embedded in technology itself.

I believe this study is RICH with many topics for discussion. I will be exploring some of these ideas in a few weeks, during my presentation, but I’d like to get you guys thinking about how this project in particular culminates many of the ideas we have discussed throughout the semester! (Then you’ll hopefully have a lot to talk about in our last class!) – You can check out the full study here : http://selfiecity.net/