How To Overcome AI Biases
June 30, 2022
Unfortunately, biased AI plagues the tech industry. AI is not intentionally biased, instead, they are accidentally programmed with limited data libraries that contain the biases. These biases include prejudices towards dark-skinned people, poor individuals, and everyone who is not a white male. The Next Web discusses how “AI Has A Dangerous Bias Problem-Here’s How To Manage It.” AI is used to determine many things in society, including prison sentences, loan applications, job candidates, gambling predictions, etc.
Chief Scientist Alejandro Saucedo of The Institute of Ethical AI and Engineering Director at Seldon is aware of the risks that biased AI poses and warns that organizations need to research their AI before deploying algorithms. He suggested tips to mitigate risks.
Machine learning needs to be transparent and explainability is a tool used for it. Explainability, however, is hard to prove, especially when its tools are biased too. Identifying the processes that need close monitoring is essential and having humans inspect them regularly prevents bias.
Accountability is also important:
“The level of human intervention should be proportionate to the risks. An algorithm that recommends songs, for instance, won’t require as much oversight as one that dictates bail conditions. In many cases, an advanced system will only increase the risks. Deep learning models, for example, can add a layer of complexity that causes more problems than it solves. ‘If you cannot understand the ambiguities of a tool you’re introducing, but you do understand that the risks have high stakes, that’s telling you that it’s a risk that should not be taken,’ says Saucedo. The operators of AI systems must also justify the organizational process around the models they introduce. This requires an assessment of the entire chain of events that leads to a decision, from procuring data to the final output.”
Saucedo advises that people use best practices and human intervention to mitigate security threats to algorithms. These suggestions work not only for AI algorithms. but also other developing issues too.
Whitney Grace, June 30, 2022
Comments
-
- Subscribe to Beyond Search
Feature archive
News archive
-
