What people learn from punishment: A cognitive model / hacker news