> I would be surprised if it's not a completely subjective experience—one person's synesthesia probably doesn't match another person's, per se.

I was wondering this too. Since they used machine learning instead of medical data, it means that each synesthesia case may be unique enough to throw off a synesthesia-color/shape catalog.

Kandinsky might have ended painting abstract art because normal art was ruined by his synesthesia, just as his abstract art is incomprehensible to someone without his kind of synesthesia.