Ethics of AI in Hiring: What to Look for

The use of Artificial Intelligence (AI) in hiring is growing in prevalence. But talent acquisition professionals remain divided on how to implement AI tools in a principled way — or whether ethical AI hiring is even possible. 

A recent Mercer survey of people who lead in-house recruitment found 40% were happy to use AI to source and engage talent to fill their pipelines, while 42% said they don’t use it or wouldn’t consider using AI during hiring. That makes sense given AI bias concerns are high, and because organizational compliance and brand reputations are on the line.

Adopting AI poses both a technical challenge and a moral dilemma. And the two go hand-in-hand, because understanding the capabilities and limitations of AI tools is the cornerstone of building an ethical AI hiring practice. Discover how to conduct an effective AI hiring tools evaluation to get a head-start.

Core Ethical Concerns in AI Hiring

HR budgets are stretched. It’s easy to see why the lure of saving time and cutting costs holds appeal for busy recruiters. The Academy to Innovate HR found 49% of HR professionals use AI in recruitment, but it was largely driven by vendors, such as AI features within recruitment platforms.

Discerning HR leaders are asking themselves: to what extent do the predictive models used in AI recruitment tools impact their organization’s commitment to building trust, facilitating human connection, and practicing inclusive hiring? 

Rather than avoiding AI entirely, HR leaders can help mitigate organizational risks by ensuring they’re across these key ethical issues:

1. Algorithmic Bias and Discrimination

One of the most pressing ethical concerns facing HR is the unfair treatment of candidates (or employees) through algorithmic bias, which is introduced through poor training data or the design of the predictive model used in AI tools. 

For example, studies of facial recognition technology have found it’s most accurate in identifying white male faces. It’s significantly more error-prone when it comes to recognising faces of color, especially dark-skinned females (MIT Media Lab). That’s because the systems were trained using mostly images of white men. 

AI-led decisions also create a situation where proxy discrimination can occur. Recently, an unsuccessful job applicant sued Sirius XM Radio, alleging its AI-powered Applicant Tracking System assigned scores based on data points that proxy for race, such as home zip code, which led to biases that resulted in his rejection for multiple positions. 

2. Transparency and Explainability

The so-called ‘black box’ problem — being largely in the dark about the quality and accuracy of the code and data underpinning an AI tool’s outputs — can make errors and biases hard to uncover.  

Not being able to easily interpret AI outputs also makes hiring decisions made by AI less defensible. That makes it harder to communicate your process to candidates, or provide strong evidence if your HR team is hit with claims of negligence or discrimination. 

Transparency and explainability will only become more important with the growing implementation of responsible AI recruitment legislation. Multiple US states now have, or are considering laws, that require employers to track and share how and when they use AI in hiring, and how they mitigate its risks.

3. Privacy and Data Protection

Around the same amount of people believe that companies’ use of AI will lead to their personal information being used in ways they wouldn’t be comfortable with.

To be equipped to comply with a complex web of international and local data privacy and security laws, and maintain an upstanding employer brand, employers need to get clarity on:

• AI platforms’ policies around the collection, processing and disposal of candidate data.
• How organizational data is used to train/modify solutions, provide services or make money.

4. Human Dignity and Autonomy

How humans feel during your organization’s recruitment process is central to the ethics of AI in hiring. Heavy use of AI can make candidates feel disrespected, especially if an organization hasn’t proactively explained AI’s role in the process.

Automated screening tools like AI-led interviews — which may exclude applicants without human review — are particularly ethically dangerous because they dehumanize the recruitment process. Job seekers and recruiters may both suffer from a lack of agency, as having little control over the choices that shape an outcome can be disillusioning.

What to Look for in Ethical AI Hiring Tools

While plenty of individual recruiters dabble with AI, a 2024 survey of almost 500 HR leaders found just 14% had actually incorporated AI technology into their organization’s recruitment stack (Mercer). A key reason cited was poor understanding about the tools and their efficacy. 

Leading organizations are committed to responsible AI adoption and governance. Part of that commitment is properly evaluating third-party tools, across multiple criteria, before they’re systematically introduced.

The core evaluation criteria to cover when exploring AI hiring tools include:

1. Human-in-the-Loop Design

Since AI is incapable of matching the wisdom, decency and compassion of a real human, ethical AI requires human decision-makers being involved. AI systems designed with an understanding of the importance of human-in-the-loop are more likely to provide opportunities for people to consider insights and fix incorrect outputs, to avoid mistakes and to help a system improve over time. 

2. Bias Testing and Mitigation

Organizations should request information about how a tool’s training data is sourced and updated, the cadence and type of monitoring and testing that vendors perform on their AI models/features, and evidence of regular bias audits. Where available, also take into account independent validations, reviews or studies on a tool’s effectiveness and fairness. 

3. Transparency and Explainability

Has the vendor made it clear what the AI analyzes or actually does in the background to produce its outputs? Organizations should be wary of tools enabling AI-led decisions that cannot provide a robust and easy-to-follow explanation of its technical workings and reasoning process. Transparent and explainable tools also help recruiters communicate well with candidates about how AI influences hiring decisions.

4. Data Practices

Look for precise and comprehensive data collection and usage policies that instill confidence that candidate and organizational data will be held and processed securely, and used only for its intended purposes. Check that the AI system is compatible with relevant data privacy and security legislation. Obtaining candidate consent and giving people control over their personal data is best practice, and central to a number of privacy laws, such as the EU’s General Data Protection Regulation (GDPR) and the California Consumer Privacy Act (CCPA). 

5. Validation and Accuracy

It’s a myth that all AI tools on the market have been carefully tested and are guaranteed to measure or predict what they say they measure or predict. If a tool makes bold claims about an AI’s predictive powers, but how its algorithm works is hidden or unknowable, it’s essential to obtain prediction accuracy metrics. Organizations should also seek clarity on the construct validity of different tools — what is the evidence that they measure a specific trait or quality?  

6. Vendor Accountability

Employers will face the consequences of any mistakes, deceptive conduct or compliance breaches resulting from the use of AI tools. Therefore, in addition to assessing the functionality of specific tools, it is vital to have trust in the vendor — including their ethical AI ethos. Has the vendor stated a commitment to developing and refining fair and fit-for-purpose tools through ongoing testing and monitoring? Do they provide clear service level agreements and responsive support?

Implementing AI Hiring Ethically

Once adopted, AI tools should be evaluated regularly. Do this by requesting and reviewing bias audit documentation from the vendor, but also carefully review your own hiring outcomes statistics and interview scoring records for adverse impacts. 

Useful methods for auditing AI recruitment outcomes for bias could include:

• Filtering data on recent candidates by protected characteristics and comparing the success rates and scores across groups to identify discrepancies in how they were assessed.
• Changing minor details (non performance-related) of a candidate’s application and running it through AI tools again to see whether the score/recommendation alters.
• Pulling together a diverse mix of stakeholders to manually read and compare AI-screened applications for a role to sense-check the tools’ decision-making from a human-led perspective.

Conclusion

Ethical AI hiring isn’t an oxymoron. Despite lingering questions over the validity and fairness of AI in hiring, organizations that thoughtfully choose and apply AI tools can benefit from faster and fairer recruitment processes.  

HR leaders that care about doing the right thing, yet want a competitive advantage from AI, should evaluate their current and prospective AI recruitment tools against the guidelines provided here before progressing from experimentation to full-scale implementation.