Discrimination in Algorithms

Sandra highlights the potential for discrimination in algorithmic models, emphasizing that socioeconomic status and identity factors can be embedded in the data without explicit categorization. She suggests that by analyzing the outputs, such as loan approvals or hiring trends, we can identify biases even when the inner workings of the models remain opaque. The conversation also touches on the surprising ease of personal identification through minimal transaction data, raising concerns about privacy and the depth of insight that can be gleaned from spending habits.