The AI technology is man made to replicate human behavior and make human lives easier on a day to day basis. Overall, Artificial intelligence (AI) is the development of smart machines that are capable of carrying out operations that require human intellect— such as perception, problem-solving, reasoning, learning, and natural language processing. AI is achieved by combining large amounts of data with fast, iterative processing and intelligent algorithms, allowing the software to learn automatically as it experiences different patterns of features in the data (Artificial Intelligence: What It is and Why It Matters, 2022). AI is a very broad subject but can be categorized into these main fields: natural learning process, machine learning, and deep learning. Natural learning is teaching computers how to understand human language. Machine learning involves algorithms and data for computers to learn patterns and relationships. By doing so, AI is able to predict human behavior and actions. Lastly, deep learning entails making neural networks that learn and make decisions independently such as . Although AI has the ability to simplify work and improve in the future, it also has societal and ethical issues, such as bias in algorithms, that need to be addressed. 

There are several ways that bias can arise in technology. One common way is through skewed or prejudiced data used to train algorithms. If the data used to train an algorithm is not representative of the entire population, then the algorithm may not accurately represent everyone. For example, if facial recognition technology is trained on predominantly white faces, it may not be as accurate in identifying people with darker skin tones. 

Another way bias can arise is through developer subjectivity. If the developers who create a technology have unconscious biases, they can be reflected in the technology they create. This is why it is crucial for developers to be aware of their biases and take steps to mitigate them. 

Additionally, bias can arise through the unintended effects of automated decision-making methods. For example, an algorithm designed to screen job applicants may inadvertently discriminate against certain groups of people, such as women or people of color, because of how the algorithm is designed. 

The consequences of technology bias can be significant. It can perpetuate social inequalities, exclude certain groups of people from using certain technology, and result in discrimination. To ensure that their products are fair and open to all users, developers must proactively identify and address bias in their technology. This includes using representative data to train algorithms, being aware of their biases, and designing transparent and accountable algorithms. By taking these steps, developers can help to create a more just and equitable society through AI.

Training data is a common problem when AI algorithms are being built and in their beginning stages of learning. Training data is data used to build an algorithm before it is tested worldwide. Oftentimes there are flaws in this training data and bias has an easy pathway to sneak in. If the training data lacks different perspectives or a wide range of data, this is when problems can occur.  If this bias is not detected early on, it can be built into an algorithm and very hard to detect and eliminate later on. To summarize, AI bias is created by human input and early technological flaws. Humans choose which data is used in the algorithm and humans also input their own data into it. Due to AI’s advanced learning abilities, it can often mistake bias and stereotypes for fact. Dr. Sanjiv Narayan is a professor at the Stanford University School of Medicine and has recently been studying AI and how bias plays a factor in it. He said, “All data is biased. That is not paranoia. This is fact” (Narayan, 2021). It is clear to see that artificial intelligence and its algorithms due in fact contain bias. 

Racial Bias - Racial bias in AI is very common in programs such as facial recognition. These systems have been found to be less accurate for people with darker skin tones due to lack of database images of these people. This leads to a higher likelihood of misidentification and false arrests for people of color. A study done by the American Civil Liberties Union found this to be true, stating that the AI systems failed to tell black people apart from one another (Burton-Harris, Mayor, 2020). This can have major consequences such as incorrect arrests for misidentified people (Brodkin, 2023). Another example of this was actually found in one of the biggest companies in the world. Amazon had an employee screening AI created that aimed to increase diversity in the workplace. However, it was created by a group of majority men and therefore the training data was based on these men. This created problems in the AI as it started to prefer male candidates. This was due to female candidates having possible time gaps in their resume due to possible leaves such as birth. The AI was unable to detect this and did not take it into account. (Narayan, 2021).

Gender Bias - There are many AI systems that aim to create equal opportunity systems for hiring or job search. However, these systems often do the exact opposite. These hiring or decision making algorithms tend to be biased against women. They either tend to prioritize males or penalize women for taking breaks in their careers (such as during birth or after). The Amazon example also falls into this category as it was created by men and therefore the data failed to create diversity of genders. 

Socioeconomic Bias - Many AI systems are used for credit scores, mortgaging, and lending. These systems are often biased towards people from low-income backgrounds. They may not have access to the same financial resources and opportunities as others and can be viewed by the algorithm as bad candidates. In addition, many of these systems perpetuate racial bias as well. A study at UC Berkeley found that mortgage algorithms charged latino and black borrowers higher interest rates (UC Berkeley Public Affairs, 2018). 

As you can see, there are many different forms of AI that can be found in different areas of technology. Bias can easily sneak into any type of algorithm. It seems that despite the goal of AI removing discrimination, Ai rather perpetuates it in similar ways that humans do, due to the nature of AI creation.

Going off of that example, the biases of AI can negatively affect people and can be harmful through the discrimination that can be present in AI for certain groups of people. Based on human interaction, groups such as people of color, women, and other marginalized communities in areas like hiring, lending, and healthcare could be negatively affected. This can be harmful in the future when we start to utilize the efficiency of artificial intelligence to our advantage, and people not in these marginalized groups reap the benefits. In contrast, those who are in these groups get fewer opportunities to thrive. 

Another example of how the biases of artificial intelligence can negatively affect people and be harmful is the inaccuracy in AI's responses. Artificial intelligence systems with bias can produce inaccurate or incorrect results, leading to poor decision-making and potentially harmful outcomes. For example, as we ask AI for answers to different things, the biases it gains can cause it to give incorrect answers, which can ultimately lead to incorrect decisions that can harm others based on the situation. For example, we can look back at the Amazon example where the company was trying to use AI to find and select candidates to work there. The AI programming, made by a bunch of men, was found to be discriminatory towards women as it chose mostly all men for eligibility for the job. 

These different ways that the biases within artificial intelligence can be harmful also bring up the legal and ethical issues that AI can get into with these biases present. AI bias can lead to legal and ethical issues, as some forms of bias may be illegal or violate ethical principles, leading to potential lawsuits and reputational damage for organizations. This can then lead to reduced trust in artificial intelligence as AI bias erodes trust in AI systems and can make people less willing to use or rely on them, limiting the potential future benefits of AI technology.

Three ways that we can look to prevent the biases that we see in artificial intelligence are by diversifying human inputs, establishing clear guidelines and ethical principles, and by considering using adversarial training techniques or other methods. Diversifying your dataset helps to ensure that it represents a variety of individuals and groups, including underrepresented populations. Establishing clear guidelines and ethical principles for data collection, data processing, and model development help to ensure that AI systems are designed to respect the rights and dignity of all individuals. Furthermore, using adversarial training techniques or other methods helps intentionally introduce counterfactual examples that challenge the model's assumptions and reduce the potential for biases.

We can look to eliminate biases within artificial intelligence by conducting an audit on the AI system at hand, implementing changes to different datasets, and regularly monitoring and updating the AI. Conducting an audit of your AI system to identify and evaluate any biases or inaccuracies in the results it produces. Implementing changes to your dataset, model, or algorithm to address the biases identified in the audit. Regularly monitoring and updating your AI system ensures that it remains unbiased and accurate over time. 

One of the reasons why these biases within AI matter for our future as humans is social progress. AI bias can slow down or hinder social progress, perpetuating existing inequalities and preventing people from achieving their full potential. Another reason is technological advancement, as bias in AI can limit the potential of AI technology to solve complex problems and make breakthroughs in fields such as healthcare, climate change, and space exploration. In the future job market, as AI becomes more prevalent in our society, bias in AI can result in job discrimination, perpetuating inequalities in the job market. We can also see why these biases matter in safety and security. AI bias can have real-world consequences for safety and security, such as biased decision-making in autonomous vehicles or biased profiling in law enforcement. When thinking about the ethics and morals of artificial intelligence, AI bias raises essential ethical and moral questions about the role of AI in our society and how we ensure that AI is used fairly and justly. And lastly, the effects that the biases have and the education and the awareness needed to use artificial intelligence as a whole correctly. We will all need to know about AI systems' potential biases and limitations and how to mitigate these biases to ensure a fair and just future for all humans.