Exploring Big Data Ethics: The Impact of Algorithms on Lending
Written on
Chapter 1: Understanding Big Data Ethics
Have you ever engaged in a discussion about the ethical implications of big data? During my time studying Econometrics in university, we learned how to conduct statistical regressions and identify correlations within vast data sets. However, we never delved into the ethical use of these insights, which, at the time, seemed somewhat irrelevant.
Today, the conversation surrounding big data ethics has gained significant traction, particularly with the rapid advancements in AI technology. This has propelled these concerns into public discourse.
Section 1.1: Algorithmic Decision-Making and Its Consequences
A primary concern is the reliance of algorithms on data to forecast and infer future behaviors. This practice carries the risk of leading to discrimination and systemic oppression.
Recently, a simple yet illuminating example was shared with me. A bank utilizes an algorithm that analyzes a vast array of mortgage applications to identify correlations with mortgages that ultimately default.
You might assume it focuses on factors like household expenses, existing debt levels, or prior credit history. While it does consider these aspects, it also evaluates a multitude of other variables. What’s surprising is that the algorithm discovers a striking 63% higher likelihood of default among applicants who include "thank you" at the end of their applications compared to those who do not.
Section 1.2: The Implications of Discrimination
The bank has complete discretion over its lending practices. If it chooses, it can deny loans to those who express gratitude in their applications. According to the algorithm, this could substantially boost profitability across its mortgage portfolio.
Now, consider this scenario: you and your neighbor are both renting. You both have identical jobs, perfect credit histories, and the same amount saved for a deposit. Yet, your neighbor secures a mortgage while you are denied—simply because you included "thank you" in your application.
Is this ethical?
If you think, “it’s not ideal, but it’s not outright evil,” let’s take it a step further. Imagine there exists another correlation that the algorithm fails to capture because it's not a data point on the application form. Research shows that Hispanic applicants are statistically more inclined to write "thank you" in their applications.
This situation creates a scenario where an algorithm indirectly discriminates against individuals based on ethnicity—a practice that is illegal. Proving and prosecuting such discrimination is complex. However, without proper oversight and regulation, it could lead to systemic oppression of certain groups within the community.
Chapter 2: The Need for Ethical Oversight
As we progress with AI developments, it is crucial to continuously evaluate what inferences, conclusions, and decisions we permit computers to make, and what should be reserved for human judgement.
The first video titled "Priceless: Against Explainability in AI/Big Data Finance" explores the ethical dilemmas associated with AI and big data in financial sectors.
The second video, "Multiple Regression (2022 Level II CFA® Exam – Reading 2)," provides insights into regression analysis, a fundamental concept in understanding data relationships.