Working together with machine learning to investigate the specifics of my fear of spiders.
‍
I have a severe fear of spiders. However, I am also aware of how irrational this fear is. Especially in the Netherlands where the big eight-legged critters
 are rare and venomous spiders do not exist. With this project I wanted to investigate my phobia, teach the specifics of it to artificial intelligence and have it help create exposure therapy specific to myself.
‍
To start the project I first set out to hold a big spider to see it up close.
 I noticed that I was much less scared of the tarantula in such a controlled setting, but also that tarantulas don’t scare me as much. They’re fluffy and slow. After this revelation I decided to go through thousands of photos of spiders and select a few hundred of the ones I thought were scariest. 
I noticed that I am scared of a few specific characteristics, such as long, bald and irregular legs, a thick disproportionate body and a brown coIour.
I then fed these images to a generative adversarial network (GAN) which is a class of machine learning which can generate new images from a given training set of images. My idea was that the GAN would generate a spider which has all the elements I am most afraid of and create my “meta-fear”. If I can face this spider comfortably, all other spiders are less scary. I made the GAN’s spider into a 3D model, and 3D printed this for therapy. The entire process was very therapeutic.

Concept and execution by me.
Image Generation done with support of the Willem de Kooning Academy.