The Algorithmic Bias Dilemma: When Algorithms Discriminate Against Minorities

In today’s data-driven world, algorithms are increasingly influencing various aspects of our lives, from hiring decisions to credit approvals. However, a concerning issue has emerged: algorithmic bias, where these algorithms can perpetuate and amplify existing societal biases against minority groups.

The Impact of Bias:

It’s important to note that algorithmic bias can impact not only racial and ethnic minorities, but also other marginalised groups, including the LGBTQ+ community. The potential for bias against LGBTQ+ individuals arises from similar factors, such as limited representation in training data, biased language in prompts, and the application of algorithms in areas like hiring and lending decisions.

It’s important to note that proving specific instances of algorithmic bias against the LGBTQ+ community can be challenging due to the complex nature of these algorithms and the difficulty in accessing their inner workings. However, research suggests potential for bias, and it’s crucial to be aware of this risk as algorithms become increasingly integrated into various aspects of our lives.

The Root of the Problem:

There are multiple factors contributing to this problem:

• Data Imbalance: Often, training datasets lack diverse representation of minority groups. This can lead to the algorithm “learning” patterns based on the dominant group, resulting in inaccurate or unfair outcomes for minorities.
• Data with Limited Representation: Training data for algorithms often lacks adequate representation of diverse sexual orientations and gender identities. This can lead the model to make biased decisions based on incomplete or inaccurate information.
• Input and Prompt Biases: The way data is presented and the prompts used to guide the algorithm can also be biased. For instance, using gendered language in job descriptions might inadvertently favour male candidates.The language used to describe individuals or situations within the algorithm’s input or prompts can unintentionally perpetuate harmful stereotypes about LGBTQ+ individuals. For example, using predominantly heteronormative language in housing applications could disadvantage LGBTQ+ individuals seeking housing.
• Algorithmic Bias in Hiring and Lending Decisions: Similar to the examples mentioned for racial and ethnic minorities, algorithms used in hiring and lending decisions could potentially discriminate against LGBTQ+ individuals by misinterpreting data points or relying on biased training data.

The consequences of algorithmic bias can be far-reaching:

• Hiring: Studies show that facial recognition software used in recruitment can have a higher error rate in identifying people of colour, potentially leading to unfair hiring practices.
• Credit Scoring: Algorithmic bias in credit scoring can disadvantage individuals from minority groups with lower credit scores, making it harder for them to access loans and mortgages.
• Life Insurance: Biased algorithms might overestimate the risk of health issues for certain groups, leading to higher life insurance premiums or even denials for coverage.

Combating the Bias:

Thankfully, there has been taken some steps to mitigate this issue:

• Data Diversity: Ensuring training datasets are diverse and representative of the population the algorithm will be used for is crucial.
• Debiasing Techniques: Techniques like data augmentation and fairness-aware machine learning algorithms can help identify and address biases within the data.
• Human Oversight: Implementing human review processes alongside algorithms can help catch and correct biased decisions.
• Starting a regulation: In many countries, regulations are starting to surge to avoid inequalities and injustice.

Statistics:

• A 2018 report by the Algorithmic Justice League found that facial recognition software used by law enforcement misidentified black individuals at a rate 10 times higher than white individuals. It is possible to watch the documentary on Netflix
• A 2019 study by the Brookings Institution found that a widely used credit scoring algorithm was biased against black and Hispanic applicants: for borrowers with similar creditworthiness, the algorithm predicted defaults at a higher rate for these groups.
• A 2020 investigation by ProPublica revealed that an algorithm used by a major health insurer in the US unfairly penalised black patients by assigning them higher risk scores, potentially leading to them receiving less care. In 2016, they investigated a bias in a system to identify future criminality.
• A 2022 study by the Center for Democracy & Technology found that algorithms used by some online platforms in the United States disproportionately removed content created by LGBTQ+ users compared to content from other groups.
• A 2024 a Racial Discrimination in Housing: How Landlords Use Algorithms and Home Visits to Screen Tenants paper highlighted concerns about the potential for algorithmic bias in housing discrimination, with individuals potentially facing higher difficulties securing housing due to biased algorithms used by rental platforms or landlords.

It’s important to remember that these are just a few examples, and research on algorithmic bias against various groups is ongoing. As the field of AI continues to evolve, it’s crucial to stay informed about these issues and advocate for responsible development and deployment of algorithms to prevent further discrimination.

Algorithmic bias is a complex issue with serious consequences. It is crucial to remember that discrimination in real life should not be replicated anywhere, and certainly not within the sophisticated systems that are increasingly shaping our world. By acknowledging the existence of algorithmic bias, implementing robust mitigation strategies, and promoting ongoing research and development in this field, we can ensure that algorithms are used ethically and fairly, benefiting everyone, regardless of their background.

If you believe you have been discriminated against due to algorithmic bias, GayLawyers, Giambrone & Partners’ LGBT+ division, can advise and guide you. We understand the complexities of this issue and are committed to fighting for your rights.

Cynthia Cortés Castillo, Digital Marketing Executive