by Isabella Grabski
figures by Nicholas Lue

It’s no secret that bias is present everywhere in our society, from our educational institutions to the criminal justice system. The manifestation of this bias can be as seemingly trivial as the timing of a judge’s lunch break or, more often, as fraught as race or economic class. We tend to attribute such discrimination to our own internalized prejudices and our inability to make decisions in truly objective ways. Because of this, machine learning algorithms seem like a compelling solution: we can write software to look at the data, crunch the numbers, and tell us what decision we should make. 

In reality, these algorithms can and do fall prey to the same biases as humans. One particularly chilling example is COMPAS, an algorithm used in several U.S. states to determine how likely a given defendant is to commit another crime in the future. This risk assessment is used to help determine high-impact consequences like probation and parole, but an analysis from ProPublica demonstrated that the algorithm’s decisions can replicate racial discrimination.  Researchers found that COMPAS is almost twice as likely to incorrectly predict black defendants as high risk than white defendants. Using this algorithm, then, can reinforce the same biases we are afraid of in our human decision-making. 

COMPAS is not an isolated example. There are far too many to comprehensively list, leading to disparities in who is eligible for same-day Amazon deliveries, who will be shown science career opportunities, and who sees Facebook advertisements for certain types of housing (Figure 1). In short, machine learning algorithms and the biases they pick up can affect a huge component of our day-to-day lives. This issue has not gone unnoticed in the machine learning community and is referred to as the fairness problem. 

Fairness is difficult to pin down, and its exact definition is the subject of much contention among researchers. One simplistic way to think about it is that a fair algorithm will make similar decisions for similar individuals, or similar decisions regardless of what demographic an individual belongs to. This definition is vague, of course. Part of the challenge is that we can’t even define what a just and unbiased society should look like, let alone the decision-making processes that will bring us there. Nevertheless, even if we can’t state exactly what fair should look like, we often have a good idea of what unfair is. But where does the unfairness in machine learning algorithms come from, and how can we address it?

Figure 1. Examples of how bias in machine learning can affect our daily lives.

What causes unfairness?

Machine learning algorithms may seem like they should be objective, since decision-making is based entirely on the data. In a typical workflow, an algorithm is shown a large amount of representative data to learn from, and its decision-making process is refined by what it sees. However, any data we give the algorithm is describing, directly or not, the choices that have already been made in society. If black defendants are already falsely determined to be higher risk than white defendants, then an algorithm will learn that from the data as if it were factual. This bias in available training data creates a feedback loop, where the algorithm will make unfair decisions based on what it’s learned, perpetuate further discrimination in society, and thus further taint data used down the road.

Unfairness can also arise from too much homogeneity in the data. A classic example is what happened with Nikon’s facial recognition algorithm, which automatically detected the presence of blinking in photos. However, this algorithm mistakenly flagged Asian people as blinking at a substantially higher rate than other demographics. Although the exact reason was not explicitly revealed by Nikon, this situation is a textbook example of what might happen when an algorithm is primarily shown data from only one segment of the population. If the algorithm did not see many examples of Asian people, then it would not have been able to correctly learn what an Asian person blinking looks like. 

Figure 2: Nikon’s blink detection algorithm may only have been trained on certain types of eyes, leading it to misclassify new types of eyes it hadn’t seen before.

How can we correct unfairness?

Some forms of unfairness may be easier to correct than others. In the case of Nikon’s facial recognition algorithm, balancing the data initially shown to the algorithm may have prevented the issue from arising at all. But in many other cases, when the data will always reflect pre-existing discrimination in society, it is much harder to prevent an algorithm from learning those same biases. 

One approach is sometimes referred to as fairness through blindness. Here, the attributes at risk of discrimination are left out of the data entirely. In our criminal justice example, we might remove race entirely from the data we show an algorithm like COMPAS. The hope is that if COMPAS never sees what race a defendant is, it can only make decisions based on other characteristics. 

The problem with this approach is that something like race does not exist in a vacuum. Many other attributes of a defendant are likely to be associated with their race, such as zip code or profession. These other attributes can then be used inadvertently as proxies for race and lead to essentially the same unfair results. 

Another approach takes a completely different tack, and is sometimes called fairness through awareness. Instead of removing attributes that could lead to discrimination from the data, this approach focuses on these protected attributes and forces the algorithm to make comparable decisions among these different segments of the population. This idea can be implemented in several different ways, but the simplest approach would be to ensure that a positive outcome is given at equal rates across demographics. For example, an algorithm like COMPAS could be constrained to predict low risk among black defendants at the same rate as among white defendants. 

This idea may seem promising, but it can lead to problems as well by oversimplifying what fairness looks like. There could still be imbalanced decisions within the attributes we correct for. For instance, even if there is an equal rate of low risk predictions between black and white defendants overall, lingering gender or class bias could still result in penalizing black men or low-income black defendants more frequently. 

This significant shortcoming highlights a key challenge in trying to solve the fairness problem: our inability to identify what, exactly, fairness should look like. Fairness through awareness attempts to enforce a very simplistic ideal, but in doing so, it only creates more problems. 

Some researchers are trying to approach fairness from a different angle. Instead of striving for literal equality at every step of the process, one way to think about fairness is determining how to make decisions that will improve the lives of disadvantaged demographics over time. If researchers can figure out what fairness should look like, then maybe the right machine learning algorithms can guide us there.

Isabella Grabski is a second-year Ph.D. student in the Biostatistics program at Harvard University

Nicholas Lue is a third-year Ph.D. student in the Chemical Biology program at Harvard University. You can find him on Twitter as @nicklue8

For More Information:

6 thoughts on “Fairness in Machine Learning

  1. The article I read was a thought-provoking piece on the effects of social media on mental health. The author provided an in-depth analysis of the various ways that social media can negatively impact individuals’ mental well-being, including increased anxiety, depression, and decreased self-esteem. The author also discussed some of the potential solutions, such as limiting social media use or taking breaks from it altogether.

    I appreciated the author’s nuanced approach to the topic and their use of credible sources to back up their claims. The article was well-structured and easy to follow, making it accessible to a broad audience. I found the insights provided in the article to be both informative and relevant, especially in light of the current pandemic, where many people have been spending more time online than ever before. Overall, I thought the article was well-written, thought-provoking, and timely.

  2. While I wholeheartedly agree with the assessment that lack of diverse data is an inherent weakness to the abilities of machine learning, your assertion that COMPAS would flag black Americans at higher rates specifically because it’s unfair is preposterous. It simply uses statistical analysis to make its decisions. Facts are not biased. It is a subjective folly to say so.

    The African American community is attempting to heal from years of oppression and substance abuse. Instead of sweeping this under the rug, we need to support their communities by providing opportunities, safety and education in the fight against gangs and criminality. The rampant murder and violent crimes need to be openly discussed without race being brought into it. We need to focus on culture. Ideology and culture are the only serious dividers of humanity. You want machine learning to be fair? Take care of the issues that result in the data in the first place. Stop trying to hide it in the interest of perceived equality. It is neither fair, nor honest.

    Statistics aren’t perfect but they aren’t racist. To be clear, in my humble opinion, machine learning should not be implemented in court cases at all, or for deciding on sentencing and review. The problem with COMPAS is machine learning is not capable of gleaning the data required to make any fair decision on this type of situation. You’re replacing one problem which I believe to be human bias, with another one, AI’s most glaring weakness.

    Machine learning should not be implemented in any meaningful way in society except for the most simple of tasks, until we as people, realize our look does not describe who we are. Our personal ideas and actions dictate how we should be seen by others and AI.

    Thank you for your consideration.

    1. “Statistics aren’t perfect but they aren’t racist.” Agreed and this is the point that most people tend to hang their hats on because they fail to consider the data generation process which produces the data on which the models are performed.

      Consider the nature of policing. There is no doubt that communities of color are overpoliced. Laws like stop and frisk in NYC were biased against communities of color (not to mention illegal) and didn’t impact white people as frequently. Crime is universal, I’m sure if the FBI decided on a new policy of “stop and search” for financial crimes committed on computer devices in/around Manhattan, we’d see an overwhelming majority of those cases would involve white people. If the FBI were to then train a model using this data it would be biased against white people.

      This is the problem…it starts before statistics.

Leave a Reply

Your email address will not be published. Required fields are marked *