
How to Reduce Bias in AI-based Banking Services?
AI systems can only be as good as the data we feed them. Input data biases, such as gender, racial, or ideological biases, as well as insufficient or unrepresentative datasets, will hinder AI's ability to be objective.
By
CIO Applications Europe | Thursday, June 03, 2021

AI systems can only be as good as the data we feed them. Input data biases, such as gender, racial, or ideological biases, as well as insufficient or unrepresentative datasets, will hinder AI's ability to be objective.
FREMONT, CA: The application of AI in financial services introduces new ethical concerns, such as the danger of unintended biases, prompting the industry to reconsider the ethics of new models.
While the benefits of AI are obvious, the potential unexpected repercussions are less so. As AI introduces new decision-making tools, there is the possibility of discrimination, putting societies' moral commitments to the test if judgments harm diverse groups.
Stay ahead of the industry with exclusive feature stories on the top companies, expert insights and the latest news delivered straight to your inbox. Subscribe today.
Understanding AI behavior is crucial for detecting and preventing models that discriminate against or exclude marginalized people or groups. While the mysterious nature of AI can appear to be "magic" at times, it has the potential to dramatically aggravate societal challenges in financial services.
There are a number of ways bias can manifest itself in AI:
First, consider the input data. AI systems can only be as good as the data we feed them. Input data biases, such as gender, racial, or ideological biases, as well as insufficient or unrepresentative datasets, will hinder AI's ability to be objective. Uncertainty about input utilization is also a problem. Some AI training methods may disguise how data is utilized in judgments, potentially leading to discrimination, for instance, if race or gender data were utilized in credit decisions or insurance premiums.
Second, it is still in the works. Many AI systems will continue to be trained with specific data, resulting in a never-ending loop of bias. Subconscious bias or a lack of diversity within development teams may influence how AI is trained, introducing bias into the model. Models can also be unpredictable; as market conditions change, it may be difficult to forecast how models would respond, with portfolio and macro ramifications.
Finally, in a post-training setting, continued learning tends to favor discrimination. As AI systems evolve and learn, they may develop new behaviors that have unforeseen implications, such as if an online lending platform began rejecting loan applications from ethnic minorities or women at a higher rate than other groups. Finally, this has the potential to erode confidence between financial institutions, humans, and machines.
More in News
Weekly Brief
I agree We use cookies on this website to enhance your user experience. By clicking any link on this page you are giving your consent for us to set cookies. More info
Be first to read the latest tech news, Industry Leader's Insights, and CIO interviews of medium and large enterprises exclusively from CIO Applications Europe
THANK YOU FOR SUBSCRIBING


