Gender Shades & AI Biases Reflection
The Gender Shades video significantly influenced my perspective on the intricate challenges surrounding biases in artificial intelligence algorithms. While strides have been made in implementing strategies to mitigate these issues, the faults and biases are still evident in AI applications, particularly in areas like school admissions and employment.
Despite advancements, biases still continue, and in some instances, AI systems may exhibit more pronounced biases than humans. A case to demonstrate this is Amazon's use of AI for employee selection, where the algorithm, trained on a dataset skewed towards male hires, displayed gender bias against women. This showcases that there can be underlying biases that AI won't pick up on, and we as humans have to be more cautious with regards to utilizing and creating these programs in everyday life.
The crux of the matter lies in the weakness of algorithms to biases when trained on unrepresentative datasets. The key to addressing these issues lies in the awareness and proactive measures we make for ourselves. To forge a more equitable future, it is imperative to incorporate diverse and balanced datasets during the algorithmic training process. By ensuring equal representation, we can prevent the perpetuation of biases in AI systems, creating a more just and unbiased technological landscape.
While developing the project for this semester, it is paramount that we lookout for these biases and are vigil about this issue as we move forward as to prevent the continuation of these issues. With awareness of the matter, hopefully these issues will be able to be subsided.
Three Paper Synthesis
Going over Teaching Tech to Talk: K-12 Conversational Artificial Intelligence Literacy Curriculum and Development Tools by Van Brummelen , chosen from Saniya and Can Children Understand Machine Learning Concepts? The Effect of Uncovering Black Boxes by Tom Hitron article chosen from Cesar, and with my own chosen article AI Literacy in K-12: A Systematic Literature Review by Lorena Casal‑Otero , it's very interesting to see the many correlations between the articles. While the article that I chose an article that focused less on a specific research conducted teaching to the K12 students, it did go over a broad range of research that has already been conducted on this topic and condenses it down to learn from it. There is some connection to be had with the article of Teaching Tech To Talk, with regards to how we can teach AI technology to K12 students and with my article. However the Teaching Tech to Talk was more of a hands-on approach to teaching children some concepts of AI, and it showcased some primary sources of engagment since it was detailing a workshop that was conducted. But I believe the conclusions were the same, that AI literacy is important and should be taught to K12 students and teachers alike. My article emphasized the importance that teacher's role play in teaching to the students and the Teaching Tech to Talk article emphasized this too in a different way. In the Van Brummelen article, the teachers found teaching AI concepts to children a daunting task, but after the workshop they found a new perspective, and no longer felt this way. In the Tom Hitron article, it didn't go over teacher's role in the subject but it did go over the actual teaching aspect of it all. One correlation that I noticed among these three articles is how they emphasize the importance of uncovering the black box and showcase the importance of actually messing around with the algorithim by designing things and tinkering with it. This seems to be the best way to get K12 students to understand the basic concepts of AI. I found it interesting in the Tom Hitron article that when they were getting the students engaged in the workshop, they didn't explicity mention any of the AI concepts they were teaching them while they were doing their tasks. This seems different in comparision to Van Brummelen's article where it seemed more structured around learning concepts and understanding that as they were doing the workshop. This method seems fesasible if given the time to do so and the resources, however with our project we will be given about twenty minutes or so to demonstrate a concept and teach it. With this restriction, it seems best to let the student tinker with the program and explain the concepts after exploring some of the concepts so the student can have some background. Overall though, it seems that with some basic hands-on tinkering, a student is able to understand the basics of AI concepts and we should focus on that in our project to ensure students aren't lost or aren't invested in the concept to learn it.