Original Article https://www.npr.org/programs/ted-radio-hour/580617765/can-we-trust-the-numbers
Nia Imani Pillow May 5th, 2022 Professor Saidi Data for Good
Can We Trust the Numbers?
Can We Trust the Number? presented by Ted Radio Hour, is an hour-long mash-up of previous Ted Talks with data scientists, editors, mathematicians, and visual analysts. O’Neil, Buolamwin, Smith, Chalabi, and Milgram all share various issues and solutions to compiling and interpreting data accurately. Speakers advise on how to improve statistical analysis and interpretation of data throughout the hour while sharing the successful application of these machine learning techniques. Cathy O’Neil and Dr. Joy Buolamwin share similar concerns about addressing prejudices to decrease algorithmic bias against POC, women, and those who suffer from mental disparities. O’Neil shares her experience as a financier who oversaw the sorting of individual people through class, gender, and race to determine major financial decisions – for example – mortgage loans.
Key Points
• Algorithms can bureaucratic decisions other people are making for you.
• Algorithms are not only predicting outcomes, but they are causing outcomes.
• Focus on accuracy vs fairness. The data may be accurate but is the data fair? - Cathy O’Neil.
• Improving facial recognition technology to recognize darker skin – Machine Learning Techniques – Dr. Joy Buolawmin.
• Read the methodology so you know what questions were asked and how the data may have been collected. – Mona Chalabi
Personal Experience and Community Affect
My own experiences with facial recognition have been very faulty. Recently I was trying to set up my Bitcoin wallet through Cash-App when I was prompted to scan the front and back of my state-issued identification card to complete the transaction. After several unsuccessful “scanning” attempts to verify my identity to due “lighting issues”, I emailed Cash-App to explain the problem and let them know I felt as if I was experiencing some sort of bias because of my skin color. The Cash App Support team reached out to me immediately offering to manually input my information into the system.
I have always been aware of how statistics can and have been used to significantly keep people within the African American community at a disadvantage. African Americans are still suffering from the discriminatory practices of Redlining - a practice that occurs when lending institutions refuse to make loans to people with lower incomes or of a certain race. Who was collecting the data on these neighborhoods and making decisions? It would take another 30 years for this practice to be outlawed in the 1960s – but remains a major factor in today’s wealth gap between Black and white families across the country.1How Redlining Works • “Home Owners’ Loan Corporation assigned grades to, and color coded, residential neighborhoods to indicate their “mortgage security.”
• Neighborhoods that received the highest grade of “A”—colored green on the maps —were considered “best.”
• Those that received the lowest grade of “D,” colored red, were considered “hazardous.”
• Urban areas with a large share of Black families were most likely to be redlined, while neighborhoods made up mostly of white families were most likely to be deemed “best” and colored green. In redlined neighborhoods, it was virtually impossible to get a loan.2
The Fair Housing Act of 1988 – 34 years ago – was put in place to combat these unfair housing practices, but Cathy O’Neil points out that these decisions are now it’s becoming an algorithmic issue. According to European Parliamentary Research Service, “[algorithmic decision systems] rely on the analysis of large amounts of personal data to infer correlations or, more generally, to derive information deemed useful to make decisions… In many situations, the impact of the decision on people can be significant, such as on access to credit, employment, medical treatment, judicial sentences, among other things”3
Works Cited
1-2. Anderson, D. (2020, October 15). Redlining’s legacy of inequality: $212,000 less home equity, low homeownership rates for black families. Redfin Real Estate News. Retrieved May 5, 2022, from https://www.redfin.com/news/redlining-real-estate-racial-wealth-gap/