[Der_Einzige]

It'd be nice if vector based approaches (think MBTI tests being originally created using PCA) would use state of the art dimensionality reduction techniques instead of stuff from the 70s so that way each vector actually matters and explains far more variance in the data.

If you need interpretability just use an autoencoder or simese network instead of PCA. This way the new "personality" vectors are highly meaningful and visualizations are better...

[currymj]

You know throwing a neural network at things doesn't just magically make them better... lot of advantages to linear techniques... not even clear that there's enough structure in a big 5 questionnaire to benefit from a fancy NN... also dubious that autoencoder features are particularly interpretable...

[thesehands]

This is what the mypersonality dataset collected on Facebook was. The 5 axes of OCEAN - Openness, Conscientiousness, Extroversion, Agreeableness and Neuroticism for all those who took the test.

[timwaagh]

thats called the 'big five' test and is the most standard way to test personality for psychologists. its been in use in job assesments since forever.