INFORMATION TO TECHNOLOGY

Disscussion 1.

Artificial Intelligence is commonly used today; from using Siri’s voice detection to Tesla’s self-driving ability. The development of intelligence machines that are created to think and work like humans is ever-evolving, but not without apparent errors. These errors include bias, racism, and sexism as a few examples. These ideas will be explored in more detail in this discussion post.

Joy Buolamwini was moved to write an article on her experience with face recognition when she found that there was a bias in the software. She is a dark-skinned woman and found that some face recognition softwares would not recognize her face unless she put on a white mask. This sparked her interest to look into the algorithms used to create face recognition softwares which she found were predominantly trained with and created best for light-skinned males (Boulamwini, 2019).

Not very long ago, Amazon created a version of Artificial Intelligence to allow people to submit their resume’s to be hired on to Amazon. Amazon’s AI algorithm was created to separate the most suitable resumes and the incompatible resumes. This algorithm ran into a large issue when they found that the people who were selected all had very similar characteristics in the fact that almost none were women candidates because the version of AI Amazon had created and trained didn’t include many women characteristics.

These two examples just explored support the claim that AI system advantages are not worth the biases if uncorrected. If the face recognition AI system was not corrected, it would prove to show that the software creators are biased, racist, or sexist. This would create negative publicity for the software creator and would be proven when they see a sudden decrease in buyers. Amazon’s error would widely be spread through technology and publicity. This would most likely cause a significant decrease in customers and partnerships. Fairness can be built into AI systems by training the algorithms to pick up various types of information about the people they are trying to include. Such as face recognition algorithms trained and balanced with dark-skinned men and women, Asian men and women, and white men and women faces.

References:

BUOLAMWINI, J. (2019). Artificial Intelligence Has a Problem With Gender and Racial Bias. Here’s How to Solve It. Time. Retrieved from https://time.com/5520558/artificial-intelligence-racial-gender-bias/

Siwicki, B. (2021). How AI bias happens – and how to eliminate it. Healthcare IT News. Retrieved from https://www.healthcareitnews.com/news/how-ai-bias-happens-and-how-eliminate-it

Discussion 2.

 It is only a matter of time until artificial intelligence systems are integrated into most facets of our lives. Most machine learning comes from compiled data it is important to think about the implications current social circumstances have on what data gets imported and how that can create a skew that would cause AI bias. In terms of AI in healthcare there are two noticeable ethical implications that need to be addressed; healthcare data availability and AI’s role in healthcare.  

   Healthcare data availability can be an issue with bias in A.I due to the unavailability of proper healthcare data. The National Center for Biotechnology reports that minorities are less likely to receive a diverse range of procedures and experience worse medical care than whites. (Anderson. 2004) This lack of medical history can lead to a negative skew in how AI can react to care of ethnic groups. For example, if the data used for machine learning shows that there is a low percentage of CAT scans being performed on a Hispanic person medical A.I will be less likely to suggest a CAT scan to a medical provider even though the relevance of that information may not be due to health at all rather, to systemic issues.

   Another consideration of ethics that must be considered in regards to AI healthcare is the role they play in healthcare. How automated would we want our healthcare to be? Is it ethical for a machine to determine the level of care we receive? Doctors are often passionate about their professions, they will suggest care that has a low chance of success if it can have a large impact on a persons’ life. Algorithms have human influence behind them, most of the time that influence is coming from a business perspective, can we trust that these AI created will recommend a time intensive, costly procedure if it has a low chance of success for the patient? Many doctors would prefer for AI to perform duties that take them away from the care of their patients, rather than perform the roles of doctors. (Nelson, 2019)

   AI could bring many advantages to healthcare if bias is addressed. Unfortunately the bias in healthcare data goes far beyond the reach of AI programmers. The advantages of AI could still be beneficial if the AI’s are built to not account for things such as race and only focus on truly pertinent information when performing machine learning functions.

 

David Woodall

References

 Nelson, Gregory (2019) Bias in Artificial Intelligence.  NC Medical Journal volume 80, No. 4 http://ncmedicaljournal.com

Anderson, NB (2004) Understanding Racial and Ethnic Difference in Health in Late Life. National Center for Biotechnology Information http://www.ncbi.nlm.nih.gov/books/NBK24693/