The Hiring Revolution: A Conversation with Mitch Hoffman on the Growing Use of Algorithmic Hiring Tools and Potential Impact on Outcomes

Assessing the potential effects of algorithmic systems on recruiting and hiring

How do you choose the right person for the job? Hiring has always involved uncertainty. Employers must evaluate large applicant pools, relying on imperfect signals – cover letters, resumes, interviews – and their own judgment to assess candidates’ qualifications and fit. Increasingly, companies are turning to algorithmic and AI tools to help answer that question.

According to a report from the World Economic Forum, more than 90% of companies now use some form of algorithmic screening for initial candidate review – helping manage application volumes that can reach hundreds of thousands or even millions in a single recruiting cycle. But as these systems reshape how hiring decisions are made, important questions arise: Can algorithms improve hiring outcomes? How should they interact with human judgment? And what are the implications for productivity, efficiency, and fairness?

To better understand how algorithmic and AI tools may impact hiring outcomes, Managing Principal Jee-Yeon Lehmann and Vice President Yeseul Hyun sat down with Mitch Hoffman, the Richard F. Aster Jr. Professor of Economics at the University of California, Santa Barbara, and director of the Personnel Economics Working Group at the National Bureau of Economic Research (NBER). Professor Hoffman conducts research on hiring, workplace productivity, and the broader implications of employment decisions for workers and firms.

They discussed how algorithms can help address challenges employers face in hiring, how these systems interact with human decision makers, and rising concerns that they may introduce bias into hiring decisions. Professor Hoffman also highlighted factors companies should consider when integrating such tools into their hiring processes and explained how economic research can inform questions about their use and potential impact in the context of litigation and regulatory debates.

What are algorithmic hiring tools, and why are firms increasingly interested in them?

Algorithmic hiring tools are automated systems that help predict which candidates are most likely to succeed in a role based on historical data, including resumes, assessments, and past employee performance. These tools are part of a broader shift toward data-driven decision making in human resources management, and they can complement or, in some cases, replace human review at various stages of the hiring process, including screening and ranking applicants. These systematic processes can also help identify promising candidates who might otherwise be overlooked. Given the high volume of applications many companies receive, these algorithmic systems can help firms evaluate candidates more consistently and systematically.

What challenges do employers face when trying to predict candidate quality?

Employers often contend with large applicant pools and limited visibility into important worker characteristics – such as motivation, adaptability, and teamwork. As a result, hiring managers rely on noisy signals – including education, experience, and behavioral interviews – to draw imperfect predictions about an applicant’s fit for a given position. Human evaluations and decision making can also be inconsistent – different managers may evaluate candidates according to different criteria, and even the same manager’s judgments may vary over time. Data-driven tools can help firms identify patterns in past hiring outcomes that individual decision makers might miss, although the effectiveness of these algorithmic tools depends on the context in which they are used, as well as how the tools themselves are built and monitored.

Your research studies how algorithms interact with human hiring decisions. What motivated that work?

Much of the debate in hiring focuses on the potential for algorithms to replace human decision makers and the impact of such a shift on hiring outcomes. In practice, however, most organizations use a hybrid approach, in which algorithms provide recommendations while managers retain discretion.

My coauthors and I wanted to understand whether that human discretion actually improves decisions. To answer this question, we examined hiring outcomes when firms started using algorithmic hiring tools and how those outcomes changed, if at all, when managerial judgment was applied alongside them.

How did you study this, and what did you find?

We focused on the introduction of pre-employment job tests that are used as inputs into algorithms to generate structured recommendations about candidate quality across multiple firms. After introducing these standardized signals, firms saw improvements in worker quality.

This suggests that data-driven tools can help firms identify more productive workers and reduce errors in hiring decisions.

You mentioned improvement in worker quality. How exactly did you evaluate that in your research?

We examined job tenure – how long workers stayed at the firm – as one proxy of their quality. Workers who left quickly were considered lower-quality hires, while those who remained longer were considered higher quality. We also examined productivity metrics where available.

Your research also examined the impact of managers who override algorithmic recommendations. What did you find?

Managers sometimes override algorithmic recommendations when they believe they have additional information about a candidate. However, our research suggests that these overrides may not always add value. In our data, on average, managers who frequently hired against test recommendations tended to hire workers who had shorter job tenure, although this pattern was not uniform across all managers. This indicates that human overrides may sometimes reflect variable judgment quality across managers rather than superior insight. However, whether and the extent to which this is true depends critically on how the overall hiring process and algorithmic tools are designed and monitored.

Some have raised concerns about the potential for algorithmic tools to introduce bias into the hiring process. How does economic research approach this issue?

From an economic perspective, the key question is how different decision-making systems perform in specific contexts. Both algorithmic and human-driven processes can exhibit inconsistency or bias, depending on their design and implementation. No approach is inherently superior; outcomes depend on the specific processes and how they’re built and monitored.

Empirical research, including my own, examines whether structured, data-driven systems can improve outcomes in particular settings. Studies typically evaluate worker productivity, turnover, and demographic differences. The evidence suggests that structured signals can reduce decision-making variability in certain contexts, although careful validation and ongoing monitoring remain essential.

What lessons does economic research offer for companies implementing algorithmic hiring tools?

A key lesson is that implementation and governance matter as much as the algorithm itself. Firms need to decide how much discretion to give managers when algorithms provide recommendations.

The evidence suggests that human judgment, in certain contexts, can sometimes weaken the benefits of structured tools. However, this does not mean algorithmic processes should replace human decision making entirely; rather, firms should actively guide human judgment alongside these tools to fully realize their benefits. Organizations should also evaluate how these tools perform over time and adjust accordingly. More broadly, firms should treat hiring as an ongoing, data-informed process that enables continuous assessment and refinement based on context.

 

These issues are increasingly appearing in litigation and regulatory discussions. How can economic research inform those debates?

Many legal and regulatory questions focus on whether hiring processes produce inefficient or biased outcomes. Economic research can shed light on how these systems actually impact hiring results in practice.

By comparing outcomes under different approaches – such as greater reliance on algorithms versus giving managers more discretion – empirical analysis can help move the conversation beyond assumptions and toward evidence about what works in which settings. ■