AI bias in hiring occurs when an AI system favors or discriminates against specific candidates. While these systems operate unintentionally, they can still impose biases in the data they were trained on or how they were developed.
Even though AI systems don’t have human emotions they can still misinterpret information, which often leads to unfair decisions on hiring.
What is Bias in AI Hiring?
Bias in Artificial Intelligence hiring is a model that unfairly favors certain applicants. This results in the rejection of qualified individuals for reasons that are completely unrelated to their job performance or skill level, including gender, race, and other personal characteristics.
How Can AI Tools Be Biased When Used in Hiring?
AI tools when used in the hiring process can become biased if they are unchecked, specifically undermining the company’s efforts for an equal, inclusive, and diverse workplace.
Qualified Candidates Rejected For Irrelevant Reasons
AI hiring bias can cause qualified candidates to be rejected for reasons that are not related to their actual abilities.
For example, a UK based makeup artist lost her position in 2020 because an AI screening tool gave her a low score based on her body language, regardless of her strong skill assessment performance.
These types of rejections can have far greater consequences with AI becoming a more common asset utilized in recruitment.
Lack of Diversity
When a company lacks diversity and uses historical data of its employees to develop an AI model, it then risks inputting hiring biases in the system.
This often leads to candidates being underrepresented or rejected simply because they are different profiles to those of previous employees.
Homogenization
Some of the finest employees come from non-traditional backgrounds or career paths as they bring in fresh perspectives and often drive innovation.
However, as AI models are being trained on existing employee data, they tend to favor candidates who resemble their current personnel. In turn, this leads to companies overlooking valuable and quality applicants.
AI Hiring Bias Examples
Here are two real world consequences in two separate high profile cases that highlight this issue:
1. Video Interview Impact Example
A German journalist tested an AI video interview platform back in 2021 with the aim to check for hiring bias. He discovered that an applicant’s hairstyle, accessories, and outfits highly influenced the personality score.
This experiment also showed that background elements, like pictures on bookshelves, and changes in saturation or video brightness could also affect the outcome of the assessment.
2. Gender Bias Algorithm Example
Going back 10+ years in time, Amazon developed a machine learning tool to assist them in evaluating job applicants.
Now, because this tool was developed primarily from resumes of men, it became highly favorable towards male candidates. It also penalized resumes that had words like “women” in it and downgraded graduates from only-women colleges.
Despite Amazon’s efforts to address this issue, they ultimately disbanded the team behind the tool in 2017.
Ethical Implications of AI Bias in Hiring Process
Hiring bias within the recruitment process can result in some decisions being unfair and exposing your company to legal and ethical risks.
Here’s what you should know:
1. Lacking Transparency
The most worrying thing about AI bias in hiring is that even AI experts are not fully sure how these systems are able to reach certain decisions. Relying on AI for hiring decisions can greatly affect people’s lives, making it an irresponsible asset to utilize.
This lack of transparency level makes it difficult for many companies to use AI tools without bias in their hiring process.
2. Unfairness in Hiring Decisions
The biggest concern is the ethical aspect of AI bias that may cause unfair hiring decisions. When people apply for jobs they take a significant effort and it would be completely unfair for candidates to be rejected or even ranked lower in the applicant process due to factors like a headscarf or wearing glasses.
3. Regulatory Penalties & Discrimination Lawsuits
If an applicant believes they have been treated unfairly because of AI bias in hiring, they may take up legal action against your company.
Moreover, governments and regulatory agencies are increasingly introducing laws that would regulate AI usage in recruitment. Proven discrimination and failure to comply will result in costly fines, penalties, lawsuits, and damages to your company’s reputation.
For example, the European Union’s General Data Protection Regulation (GDPR) includes specific guidelines on automating decision making and this can be applied to AI hiring software and tools.
Europe & United Kingdom Regulations
European Union – AI Act
The EU AI Act was approved in March 2024 and signaled the world’s first legislation that regulates AI. It classifies AI systems that are used in the hiring process as high risk and sets out strict guidelines to ensure transparency, fairness of use, and human oversight in recruitment.
Key requirements:
- Human involvement in all final hiring decisions using AI
- Informing candidates whenever AI is part of the hiring process
- Detailed documentation of AI system development and testing
- Implementation of risk management measures to prevent bias
- Use of high quality, unbiased data to train AI models
Penalties for non-compliance:
There are hefty fines for non-compliance with reported figures reaching a whopping €35 million or 7% of a company’s turnover, based on the seriousness of the violation.
Navigating AI Bias in Hiring Algorithms
Even though bias in AI systems may never be eliminated completely, companies have taken important steps in the right direction to ensure a fairer and more equal hiring process.
Here’s how a company can navigate this:
Human Oversight
We firmly believe that all hiring decisions should be led by humans, with AI assisting in feedback and guidance to the recruitment team.
For example, AI software can help hiring managers to assemble diverse interview panels and then support interviewers in maintaining complete fairness.
AI tools shouldn’t make hiring decisions just as much as an intern cannot make CEO level decisions. A human needs to be involved to properly identify and address any biases that may arise where technology overlooks.
Regular AI Systems Audit
Regular internal audits conducted by a diverse team can help address AI hiring bias while also improving a company’s AI system over time.
Additionally, external third party experts can also weigh in and ensure your company’s AI training data is accurate to reflect the population you would be hiring from.
Autonomous testing methods can also detect any biases in the present datasets. For example, MIT researchers have developed an AI system, DB-VEA, that automatically reduces bias simply by re-sampling available data.
Diversity in Training Data
Artificial intelligence is only as effective as the data that it is trained on. When a company uses larger datasets that include individuals from different races, ages, genders, religions, cultures, sexual orientations, and skills, then it can help reduce bias.
Combining both big and small data in training sets also improves AI accuracy. Small data provides user specific insights, while big data often highlights correlations that may lead to bias.
How to Decrease AI Bias in Hiring
AI usage in recruitment can be quite powerful that can quickly analyze huge amounts of data and accelerate the recruitment process.
However, it is not a perfect solution as we have seen quite a lot of errors when used blindly. HR professionals need to remain aware of its limitations, especially when it comes to bias.
When a company evaluates AI systems and uses them to support humans in decisions, then it can greatly reduce the impact of AI hiring bias, while also benefiting from its efficiency.
Frequently Asked Questions (FAQ)
What is an Example of AI Bias in Hiring?
AI bias in hiring happens when an AI tool favors candidates with specific characteristics over others. An example would be a company’s AI screening tool showing bias against older applicants when the same candidate was rejected and reapplied using an application with a younger date of birth.
What is the Problem with AI in Hiring?
The problem with AI in hiring is its potential to introduce bias. AI tools might unintentionally favor specific candidates based on their race, age, or gender, as opposed to their skills and qualifications.
How to Reduce Bias in AI Hiring?
To reduce bias in AI hiring, companies need to take a few important steps, such as the following:
- Ensure diversity in training data
- Regular updates and reviews of the algorithm
- Maintaining human oversight
- Build diverse hiring teams
- Use bias free applicant tracking systems
How is AI Affecting Recruitment?
AI is quickly becoming a game changer in recruitment as it automates tasks like resume screening, interview scheduling, and candidate sourcing. This allows HR professionals to concentrate on strategic activities, like building relationships with candidates and making well informed hiring decisions.