week4-Discussion

Yinuo Pan 

WEEK3

COLLAPSE

Top of Form

Mathematical models and algorithms are constantly developing in various industries. A lot of work is being replaced by machines, and automated programs and algorithms that simulate human thinking and processing data are increasingly being used. However, many models are biased.

As stated by Knight (2017), Opaque and potentially biased mathematical models are remaking our lives and neither the companies responsible for developing them nor the government is interested in addressing the problem. According to the principles of automation and machine learning, the bias of the model comes from the data itself rather than the algorithm. The algorithm only reflects human bias. 

For example, in the case of salesforce, if machine learning is used to pay new employees, the algorithm must first build a salary model based on historical data. However, in historical data, under other conditions being equal, female employees will receive less wages than male employees. As a result, algorithms and mathematical models with gender bias are generated. Perhaps some people think that the prejudice can be corrected by deleting the characteristic value of gender. Unfortunately, this idea is too naive.

The problem is not only the complexity of the algorithm-for example, even if the gender is not written, the identity of female employees can also be discovered from other aspects-such as gynecological items in insurance or maternity leave plans recorded in the system. The most terrifying problem is that the companies or governments that use these models are not interested in eliminating bias. Related experts O’Neil previously worked as a professor at Barnard College in New York and a quantitative analyst at the company D. E. Shaw. However, "I’ll be honest with you," she says. "I have no clients right now." (Knight, 2017) 

Obviously, these algorithms have brought benefits to the institutions that use them, at least they have not brought them losses, while the modified algorithms may bring them losses. Data scientists can change algorithms, but they cannot change Interests supremacist and the society. 

References: 

1. Knight,Will(2017, July). Biased Algorithms Are Everywhere, and No One Seems to Care. MIT technology Review From https://www.technologyreview.com/2017/07/12/150510/biased-algorithms-are-everywhere-and-no-one-seems-to-care/

Yinuo(Richard) Pan

Bottom of Form

Mengqiu Yao 

week 3 dicussion

COLLAPSE

Top of Form

Bias is the biggest factor affecting the accuracy and credential of data analysis, and it’s commonly existed in every step of analysis. Common bias includes information bias, selection bias and confounding bias. Algorithmic is a deep learning technique supported with tremendous data. Error is highly possible to occur if data isn’t sufficient, especially when certain group of data is expelled by mistake. Algorithmic bias may magnify potential problems because people tend to over trust AI. Bias could cause serious negative consequences especially for poorer communities and minorities if algorithmic bias makes ever-more- important decision goes unrecognized and unchecked (Knight. M 2017). In 2016, an investigation by ProPublica found that law enforcement officers using artificial intelligence technology-driven tools to investigate reoffends between the whites and the blacks. In some cases, the judge depends on AI to decide who stays in prison and who walks freely even without further investigation. It could save time on judicial process if algorithm is correct, but it could cause disastrous consequence if the algorithm is biased. Algorithm bias may implicate the innocents and increase the public’s prejudice against them. 

Personally, I think government should pay more attention and take necessary regulation on AI which is applied on important strategical decision making. Also, companies that developing such AI should make their product more transparent, but not hide the internal operation of their algorithms as secrets. Individual auditing team should be promoted to examine these algorithms and find potential errors. People have the right to know the logic behind any AI algorithm that affecting their daily life.  Reference: 

Knight, W. (2017, July 12). Biased Algorithms Are Everywhere, and No One Seems to Care. MIT Technology Review. Retrieved July 23, 2020, from https://www.technologyreview.com/2017/07/12/150510/biased-algorithms-are-everywhere-and-no-one-seems-to-care/

Larson, J., Angwin, J., Mattu, S., & Kirchner, L. (2016, May 23). Machine Bias. ProPublica Retrieved July 23, 2020, from https://www.propublica.org/article/machine-bias-risk-assessments-in-criminal-sentencing

Bottom of Form