The Internet Shadow

Date: 2021

Type: Individual project, Technical Study

Programme: Architecture MArch, Architectural Association School of Architecture

Software: Google Takeaway, Jupyter Notebook, NumPy, NLTK

Internet Shadow is the sum of all one's actions performed on the internet over a period of time.

During the project, four of my friends gave me their complete Google Search and YouTube history. I then mined that data to visualise their shadows, alongside my own dataset.

I did this by plotting all words in 3D space according to their meaning, using a Machine Learning algorithm called Word2Vec. Each person's shadow turned out to be incredibly unique.

 

All five internet shadows mapped in one space

 

I then hypothesised that the plotted pattern would predict personality. I thought, for example, that a more spread-out internet shadow would signal higher trait Openness (from the Big Five Personality Traits).

I then asked my friends to take the personality test, the results of which are seen below:

Eeda

Ash

Siyue

Erika

Veronika

 

The width of the internet shadow turned out to be inversely correlated with the trait Openness. 

​While that is the opposite of what I predicted, a correlation still exists. 

​Larger studies are needed to investigate this correlation, which is challenging due to Google API limitations to making an application for performing this analysis.

Other datasets could be used, such as Spotify data, as their API allows for use of personal music data in applications.