Two critical actors have been almost entirely missing from research on AI hiring bias: the outsourced recruitment firms and the technology developers who build these systems. A new conceptual framework published in the Journal of Information, Communication and Ethics in Society reveals that the fairness and validity of AI-enabled recruitment depends far more on the complex relationships between these players than on the algorithms themselves.

For years, the conversation around AI hiring has focused narrowly on whether algorithms discriminate. But researchers Penny Williams and Paula McDonald argue that this approach misses the bigger picture. AI doesn't exist in a vacuum; it's designed, deployed, and used by different actors working within different institutional and cultural contexts. When you ignore these relationships, you're essentially ignoring where bias actually gets baked in.

Why Are Outsourced Recruiters and Tech Developers Being Left Out of the Conversation?

Most research on AI in hiring has concentrated on either the technology itself or how job candidates react to it. But the people actually building and selling these systems have largely escaped scrutiny. Outsourced recruitment firms, which handle hiring for thousands of companies, and the technology developers who create the software, occupy a middle ground that researchers have overlooked. Yet these actors make critical decisions about how AI gets configured, what data trains the algorithms, and how the systems are actually used in practice.

The problem is structural. When a company outsources recruitment to a third-party firm, and that firm uses AI tools built by yet another vendor, the responsibility for fairness becomes murky. Who ensures the algorithm isn't biased? The employer? The recruitment firm? The software developer? The answer, according to the research, is that all three need to work together, but they rarely do.

How Do Technical and Social Factors Interact in AI Hiring Systems?

The research applies a framework called social informatics, which examines how technology and human actors shape each other. In AI recruitment, this means understanding that the same algorithm can produce very different outcomes depending on who deploys it, how they deploy it, and in what organizational context.

Consider a CV-screening tool. The algorithm itself might be neutral, but if it's trained on historical hiring data from a company that has systematically hired fewer women in technical roles, the tool will learn to replicate that pattern. The developer might not have caught this bias. The recruitment firm deploying it might not have tested it on diverse candidate pools. And the employer using it might not even know the problem exists. The fairness of the system depends on all three actors catching and correcting for bias at different stages.

The research identifies several AI-enabled recruitment technologies now in widespread use, including vacancy prediction software, CV-screening tools, AI-powered psychometric testing, automated video interviews, and candidate engagement chatbots. Each of these involves design choices, deployment decisions, and usage patterns that can either mitigate or amplify discrimination.

Steps to Strengthen Fairness in AI-Enabled Recruitment

• Transparency Across the Supply Chain: Employers need to demand that recruitment firms and technology developers disclose how algorithms are built, what data they use, and what testing has been done for bias. This includes understanding the historical data used to train the system and whether it reflects the diversity of today's job market.
• Shared Accountability Frameworks: Rather than leaving fairness to chance, organizations should establish clear agreements about who is responsible for monitoring bias at each stage. This includes regular audits of hiring outcomes by demographic group and mechanisms for correcting problems when they arise.
• Diverse Input During Design: Technology developers should involve recruitment professionals, HR experts, and representatives from underrepresented groups when building hiring tools. This helps catch potential fairness issues before the software reaches the market.
• Ongoing Candidate Feedback: Recruitment firms should systematically gather feedback from job candidates about their experience with AI tools and use that data to identify potential fairness concerns that might not show up in statistical audits alone.

What Does This Mean for Job Seekers and Employers?

For job seekers, the takeaway is sobering: the fairness of AI hiring systems depends on decisions made by people you'll never meet, at companies you may not even know exist. A resume might be rejected by an algorithm trained on biased historical data, screened by a tool deployed without proper testing, and evaluated by a system whose fairness no one is actively monitoring.

For employers, the research suggests that simply buying an AI hiring tool and assuming it's objective is dangerous. Organizations need to understand their entire recruitment ecosystem, including who is building the technology, how it's being deployed, and what safeguards are in place to catch bias. The research emphasizes that AI is fundamentally reshaping how talent is attracted and assessed, with potentially uneven social consequences, yet there has been rapid, largely unregulated expansion of these technologies.

The broader context matters too. As AI and data science reshape employment across industries, workforce transformation depends on more than just reskilling workers. It requires responsible AI implementation and collaborative policymaking among governments, industries, and academic institutions. In hiring specifically, this means ensuring that the systems deciding who gets access to jobs are built, deployed, and monitored with fairness as a core principle, not an afterthought.

The research presents a new conceptual framework for understanding AI in recruitment that accounts for the design, deployment, and utilization stages, and the different actors involved at each stage. This framework is intended to guide future research and help organizations understand where fairness can break down and how to fix it. Without attention to these hidden players, the promise of objective, efficient hiring through AI risks becoming a mechanism for scaling existing biases at unprecedented speed.