Data bias: when the information is favorable to a group

Daniel Kahneman brought to attention that humans are biased and noisy. The former means it tends to decide on a direction, whereas the latter, it can be influenced by random factors. One can be indentified, the other not.

Some biases we may have one datasets:

1. Gender bias;

2. Color bias;

3. Nationality bias

4. Surname bias

5. Past bias

Make sure if you a data scientist, keep you dataset well-represented to all the group sets you want to model!

Biased dataset in Portuguese creates prejudice on machine

See: "How much do we have to know to predict? do not mistake understanding with prediction! The illusion of validity"

---

Did you know: our world is designed for men

It may sound like dummy feminism, but we have real effects. One is, say, for medical treatments, or even female officers.

Of course, AI should be the biggest concern, since it is here to stay, and soon may be taking decisions. Initial voice recognitions could not recognize female voice.

---

Listen

---