Does under-representation always cause algorithmic bias and does collecting more data on the minority group always solve the problem? Our recent results show that it can be the opposite:
arxiv.org/abs/2306.05068
@SamiZhioua
@WiMLDS_Paris
arxiv.org/abs/2306.05068
@SamiZhioua
@WiMLDS_Paris