Well, you are right that uber users is not a representative sample of the city population. Biased data is not inherently bad, it's just that you need to be aware of biases.