There is much we can learn from data. And machine learning can be powerful. However when we use data from the privileged to understand the vulnerable, we’re unlikely to succeed. It would be like studying the spending habits of billionaires to draw conclusions about how people on welfare spend their money.
Seems silly doesn’t it? And yet, according to findings by UNICEF, this happens with mobility data. The richest 20% of users create 54% of the data. Please note this is 20% of users not of the world’s population – the percentage would be even lower. UNICEF identified this and worked to de-bias this data. Certainly important work.
It also makes me think again about how biased our datasets are. And how important it is to realise this, understand it, and do something about it.