German Credit Data Bias and AI Decision Making Critique

German credit datasets are those used to determine a customer's credit risk within the country of Germany, and thus are used to decide if people get access to certain lines of credit. Each row represents an individual's financial status and history, with the inputs contributing being make or break for some looking for a loan. German credit data analysis reveals potential biases, specifically within the inputs used to determine customer credit risk. Outdated inputs, such as those based on gender, marital status, and others, disproportionately affect certain communities, and can marginalize certain groups within the current system. 

Simply put, the data set lacks certain social and political contexts that should be considered when determining someone's credit risk, such as the current political climate and social and educational opportunities. These inputs can be reinforcing of current power dynamics, as social norms and standards are determined by an overarching system, and if this system is unjust, then it will unjustly affect people's standards of living and the opportunities provided to them. For example, some citizens do not have the same access to education, work opportunities, etc, because of circumstances out of their control. Being a foreign worker, for example, comes with many challenges within immigrating and finding work in another country, which is a prime example of how this group (foreign workers) could be marginalized based on the inputs used in the data, as the data provides no context for what they have gone through to get into the position they are in. Examples like this remain consistent across the dataset, consistently marginalizing, as it lacks context for customers' situations.

Given these factors, we concluded that AI should not be used in determining credit risk and who gets access to credit, as it lacks the necessary context in order to make equitable and just decisions. The dataset, given the biases in determining the factors that go into credit risk, should not be coupled with AI for decision-making. A huge part of this is the fact that determining how to group people is an inherently biased practice, as there are infinite ways to group people, and defining groupings in a way that is driven by data points becomes biased when determining what data points to use. Whoever codes the data codes their personal bias into it. If an AI continues to make inequitable decisions in this system, it will continue to fuel inequality. It takes an actual person with an understanding of the happenings going on within Germany to make these decisions; again, an AI should not be allowed to do so.