We had to chose an innovative field of research, still not well explored, about digital footprints, the data that is left behind by users on digital services. The we had to develop our concept on a dataset, built by collecting the digital traces left by a member of the group.

 

 

 

One of us noticed that, using an alternative keyboard app on an Android device, it was possible to retrieve parts of past conversations, using only the previsions suggested by the keyboard. Then, when she linked the app to the cloud (including Facebook, G+, Gmail, Twitter, sms service, Evernote), the predictions improved and became more accurate.

 

 

 

First of all, we needed a dataset. It is possible to access your personal data held by SwiftKey by sending your request together with a check for £10 Sterling payable to TouchType Limited (click here). Unfortunately, because of time constraints, we couldn’t buy this data, so the collecting was manual.

 

We calculated that, by choosing each time one of the three suggestions for six consecutive times, we would have had sentences of six words, and 1092 words overall. Unfortunately, we could not go any further, since the seventh “level” consists of 3279 words (n=173n). (To calculate the amount of the elements for each level, we used WolframAlpha.)

Since the predictions are often articles and prepositions, in order to obtain a sentence with a logical sense, sometimes we had to tap the suggestion more than six times.
We had to delete some predictions from time to time, since the more we tapped on a prediction, the more the keyboard suggested that word, repeating previous patterns and leading us to not-so-interesting results.
Moreover, the suggestions changed within the seven hours spent to complete the collection, since the user needed to use her smartphone. So, it was quite unstable. Those were the problems relative to the manual collection.

The dataset was designed, since the beginning, to be loaded on Raw, a data visualization tool developed by DensityDesign Lab, a research lab in the Design Department of the Politecnico di Milano. We chose the Circular Dendrogram chart, among the sixteen proposed. It was the best way to show the ramification of our results, and the connection between words. With it, we realized an infographic. The result were classified by their level of confidentiality: level 0, marked in blue, for non-confidential sentences and recurring expressions in the user lexicon; the first level, marked in orange, for generic references about user’s life, daily routines, nicknames; the second level, for specific references to the user’s life, names, places, private conversations.

In order to communicate our project on the internet, we shot and uploaded on YouTube two short videos, portraying in an ironic way what could happen if your keyboard and its informations fell in someone else’s hands.

These video, the infographic and some informations about the project were put on a website.

 

A long message generated by using the words suggested by the keyboard

IMAGES.