I'm in Love with a Stereotype

The simple game

Just another cute little dating sim

Using an OkCupid dataset, players are presented with a series of generated profiles that include descriptions and a representative animal photo that they can swipe on.

For any profile that they identify as wanting to match with, they are presented with a summary of what their first date with this person would look like: a terrible scenario based on the stereotype that the person's description fell under, as determined by machine learning.

The takeaways are:

• dating apps are terrible and you will have a bad experience trying to find someone to date
• there are so many red flags that you should be able to spot in people's profiles
• dating apps determine your experience based off of ML algorithms that arbitrarily interpret how you appear, as well as how others appear
  

Love learning about dating apps? I wrote a whole paper on them which you can read here ☞

Powered by Machine Learning

Mapping data to arbitrary latent space

Machine learning usually relies on the ability to map data onto latent space, which requires assigning values to the data points along axes or measures that the developer defines. The data itself may have bias--something that's been well-documented in critiques of machine learning and AI--but so can the way it's interpreted.

By mapping the OkCupid data along very arbitrary axes such as soft boy to fuckboy,  the latent space created is shaped by the way that I, the data researcher, want to interpret the data.

Clustering this data, I could interpret how the algorithm determined how people were similar to each other.

Elen síla lúmenn’ omentielvo