Artificial Intelligence Bias Traits in the Urgency of Human Discrimination: Cases in Race and Gender Minorities
For the last decades, the world has been moving forward with tech-based development and changes. Artificial Intelligence is one of them, taking lots of changes that took place in terms of medical development, manufacturing, construction development, and any other aspect of development. The existence of this tech-based development with its various function gives society a brand new helper in life which has been producing dependency. While technology is something that should be bias-free, it also has the potential to become like the maker, a human, who is error-prone and thick with subjectivity. In this article, we will talk about the existence of AI. AI is defined as a piece of knowledge and system that is used in a domain and adopts a different approach. AI is a study that is still young, with a multidiscipline that is developing faster than enough for the past decade. It is interesting to analyze how this subject will soon be developing more and more, better than the data and algorithms that are bias-free and discriminatory-free for the sake of a more inclusive function.
The algorithm is known as one of the crucial development of the internet and media ecosystem. The algorithm is a process or rule that is followed on a computer system, and it became one of the reasons why we can see ads on a certain product that we are currently looking for. In general, the algorithm is a machine learning subset that is expanded to tell the computer how to learn and operate themselves. Then the data from AI could be identified to be used for automation and fasten the data preparation including the data model making and also data exploration. So it could be concluded that AI is a bunch of algorithms that differ according to what data is received. In this case, it is still relevant to say that bias and discrimination against AI are still according to the maker.
Few case studies done by researchers show that there is discrimination and racism potential in technology. Take the case of facial recognition as an example. This case was written on a thesis by Yavuz in 2019. Despite the debatable notion about privacy rights threats, facial recognition potentially could be biased and discriminative. Take a searching platform on the internet, on one of its servers, they have the photos feature that is defined as a cloud that automatically stores and categorizes photos and videos to ease the user for searching desired photos. A case appear in 2015 when a software developer upload two black people, and the system from facial recognition labeled that as ‘gorillas’. In solution, the searching platform has erased the primate labels from the system to fix that issue.
I will add another notion while we are on Trans Visibility Day. In this modern society, the occurrence of bigotry and biases still happen to a lot of marginalized groups of people, including people of gender minorities. A general concept that we should understand foremost is that there are identities other than the binary genders that exist around us. Not to mention, the conversation about non-binary genders and gender non-conforming people in any scope possible is still limited, hence the technology itself is still adapting in a questionable sluggish way. The algorithm seems to take place to this vigorous agenda in a way. Of course, we could say that the culprit is humans with a lack of diversity in their input. Thus it is still our responsibility to further fix this problem in the technical scope.
The case in this notion flies around the fact that transgender individuals are not even classified on a simpler human system like the governmental system, let alone on the automation that requires big data. There is a circumstance where gender issues are narrowed to a uniform experience of people that belongs to one gender that could result in false identification. We could sum up that taking intersectionality into the AI principle is meaningful, but it should start from the base, which is the identification system itself. If we are not ready to put other gender options on our IDs, we still got a long way to assemble more inclusive technologies.
From this case, we should understand that label and labeling became one of the steps that are crucial in automation. Other than that, machines may learn about the bias itself because they receive inputs from the sampling bias which did not go goon on in the data practice processing. The challenge is how can we contribute to fixing the system that strengthens the bias in society because of the incompleteness of the data collection. But we surely agree that this cause exists because human itself is a biased species. Therefore, companies that use AI are most likely to take steps and action to use data practicing that is more representative with auditing system accordingly for the biases that could affect certain groups of society.
One can conclude that technology that is made by human experience a similar flow to adapt just like a human. The urgencies that are known is for individuals and institutions to be aware of maintaining technology, especially in training them to process what is already input by humans. This is important because to reach the goal to operate something manual for ease of life, humans should build a helper that objectively could reach every corner of the human background. If everyone is included in the system, everyone could use it effectively because it functioned as a tool that is free of bias and discrimination.
UNESCO. 2020. Artificial Intelligence and Gender Equality. Key Findings of UNESCO’s Global Dialogue
Yavuz, Can. 2019. Machine Bias: Artificial Intelligence and Discrimination. International Human Rights Law
“No pride for some of us without liberation for all of us.” — Marsha P. Johnson
Happy Trans Day of Visibility 2022