My amazing friend Hang asked me late last summer if I would be interested in contributing to DataSelfie. Of course I was. DataSelfie is a project that has its roots in her thesis from…2016? After that she, together with Regina Flores Mir worked on it for about a year to make it a reality - with great success.

From the DataSelfie website: Data Selfie is a browser extension that tracks you while you are on Facebook to show you your own data traces and reveal what machine learning algorithms could predict about your personality based on that data.

While you are using facebook the extension tracks how long you are looking at specific posts, whose posts those are, what you like, comment, the content of your comments etc. It then sends this data via its server to “real” Machine Learning APIs - by “real” I mean that the same algorithms are used in the real world to analyse information about you… its inferrences are then used, for example, to sell things to you or affect your life in other ways. One of the algorithms DataSelfie makes use of is Cambridge Analytica which got an extra amount of media attention since it came out that Donald Trump’s campaign employed it to gather information about american voters.

>> If you are using Facebook, you should definitely try it - it’s out as Firefox Add-on and Chrome Extension.



Hang asked me to implement an image analysis feature for the next iteration of the project. My plan was to build a own API. DataSelfie’s clients would then browse their Facebook, links to the images that scroll by are collected (or maybe only the images that are being looked at for the longest, or liked/commented) and send to the DataSelfie server. The server that send the links to the API, where they are downloaded, classified, deleted and the result of the classification is send back all te way to the client. Over the time, the client then gets to see what the algorithm thinks the images he/she looks at contains.

More on how exactly I achieved all this soon!