Hirevue: Artificial Intelligence + Hiring

Ethical artificial intelligence (AI) usage has the ability to provide powerful data that supports smarter hiring decisions— helping us understand jobs and people better than ever before and offering strong and objective decision support. There are various solutions on the market that employ AI, including job matching, conversational AI, assessments, and more. Solutions range from easy deployment with low impact to more sophisticated deployment with higher ROI. 

For example, a basic, scripted chatbot is simple to incorporate on a careers website and keeps the process moving quickly for both parties, offering the ease of convenience for candidates. Solutions like assessments driven by AI frequently require a more complex deployment with an ATS integration and provide powerful data by assessing skills, work styles, and personalities to determine fit for a role.

Hirevue products used for candidate engagement and interview insights employ  generative AI, or Large Language Models (LLMs) with additional domain knowledge and strict guardrails. We have also trained our own proprietary models for job matching and candidate assessment. We believe that algorithms should be highly controlled and tested by experts (IO Psychologists and Data Scientists) throughout the entire lifecycle of the model. As such, they can then help to reduce hiring bias and make the process more fair.

We deploy AI Assessments that are both static and deterministic, meaning algorithms are not being re-trained or “learning” on the fly and they provide a repeatable output each time they are used; this should be the standard in hiring. This includes our language proficiency assessments through a third-party provider, which like our proprietary systems, uses language models and undergoes rigorous bias testing. With all of these AI systems, the algorithm is trained and tested in the “lab” and then locked before deployment. The system can only “learn” new things if someone chooses to update it.

It is absolutely true that if care is not taken, AI algorithms can mimic human bias. Both Hirevue data and algorithms are audited to check for bias and mitigate accordingly. There’s an entire field of research around algorithmic fairness, but it’s a complex topic—no one definition of “fairness” actually exists. However, in the hiring space we benefit from decades of work on pre-hire assessments in I/O Psychology. Computers allow us to formalize fairness/bias requirements that we define in an automatic way. Rather than simply incentivizing an algorithm to predict an outcome (in this case, a job-related competency), we simultaneously penalize it for having significant demographic group differences in the outcome.

Read Hirevue’s industry-first AI Explainability Statement to learn about the methodology for mitigating bias in assessment algorithms.

AI has sparked both curiosity and concern among hiring teams and candidates alike. Break down the myths circulating around AI in hiring, and navigate the evolving landscape of hiring with confidence and efficacy.

Myth: There are tricks to “game” my AI interview.

Fact: Due to the capabilities of modern natural language processing (NLP) systems and the robustness of our training data, there are no magic keywords to trick the system—it’s much more nuanced than that. There are no secrets to trick interviews, assessments, or a chatbot. Additionally, the best thing candidates can do to perform well on a recorded interview is to reference standard interview tips and tricks. Practice your responses to standard questions about your past experience, challenges, and how you’ve overcome them, and what makes you valuable as an employee.

Myth: AI decides who gets the job.

Fact: Although AI may play an important role in evaluating candidates, humans make the final hiring decision. Our algorithms are built to bolster human decision-making in a structured and consistent way that combats human biases.

Myth: AI replaces a face-to-face interview.

Fact: AI-assisted recorded interviews occur rather early in the hiring funnel–at what is typically the resume review or phone screen part of the process. An AI interview gives all candidates a fair chance to move onto face-to-face interviews, which come later in the process.

Myth: AI is a black box and we don’t know what it’s doing.

Fact: There is an entire field of study dedicated to AI explainability. Because increasingly complex systems can be increasingly opaque, this is extremely important work. In a famous example, researchers built a model to predict whether a picture was of a wolf or a husky. The model was extremely accurate at differentiating the two, but when one uses an explainability method that highlights the most important parts of the image it was using to make the prediction, it determined it was the snow in the background of the image. In the training data, wolves were much more likely to be photographed in the snow. Similar methods can be applied to language and can help us understand how models are working and verify that they are paying attention to the data that matters.  

Myth: AI is always learning and changing in uncontrollable ways, so you can’t trust the results.

Fact: While some AI systems do learn continuously, responsible deployment varies by use case and risk level. For high-stakes hiring assessments, Hirevue models are trained and tested in the “lab” then locked for deployment—it wouldn’t make sense to use a different algorithm day-to-day when evaluating candidates for a role. In lower-stakes applications where we do use generative AI, we implement extensive guardrails and monitoring to ensure consistent, appropriate outputs. The key is matching the AI approach to the application’s requirements and potential impact.

Myth: AI is going to make recruiters obsolete.

Fact: The World Economic Forum predicts 92 million jobs globally will be displaced  by 2030, but 170 million new roles will be created during that time, leaving 78M net new jobs. As has happened many times historically, new tools increase productivity and have allowed time for more strategic tasks.  This puts an additional focus on hiring for skills and agility. The role of the recruiter is certainly changing as technology permeates the hiring space (e.g. LinkedIn). New tools have changed the role so that recruiters can use their time to focus on fewer more promising candidates rather than mundane tasks that do not require human ingenuity (like back-to-back phone screens or scheduling interviews) and often suffer due to human biases.

Myth: AI will not treat those with disabilities fairly.

Fact: At Hirevue, we provide many accommodations for candidates with disabilities including but not limited to allowing retakes, removing timers, and, where applicable, explaining before an interview that AI will be used in its assessment. Hirevue also works closely with Integrate Autism Employment Advisors to understand how candidates on the autism spectrum interact and perform with AI hiring assessments. Research has shown that interview processes with appropriate wording, structure, transparency, and adequate preparation time are highly beneficial to these candidates. Our own peer-reviewed research with Integrate Autism shows that game-based assessments are a particularly effective way to discover this untapped talent resource.

Myth: AI is unregulated. 

Fact: AI has created entirely new areas that spur nuanced conversations about how to regulate certain technologies (social media recommendation systems, who owns content produced by generative AI, etc.). However, even though AI is being introduced into many different areas of our lives, existing laws still apply. Particularly in the US, laws against discrimination in hiring have been in place for decades, and AI does not give anyone a free pass to break them. There are well-established best practices and statistical methods for bias testing that are standard in the IO Psychologist toolbox.

Myth: AI on my video uses face scanning technology which judges me on behaviors and attributes I cannot control.

Fact: Hirevue video assessments don’t use facial analysis, video, or audio data to evaluate candidates. Instead, we analyze only the transcript of your responses, focusing on how you describe your experiences and past actions related to key job competencies (e.g. team orientation or adaptability).

Hirevue is committed to the science, research, and regulation of ethical AI. Access Hirevue’s industry’s first AI Explainability Statement, Ethical Principles, audits, whitepapers, and other resources.

• Hirevue AI Explainability Statement
• Hirevue AI Ethical Principles
• ORCAA Algorithmic Audit
• Landers IO Psychology Audit
• Hirevue Whitepapers
• Hirevue and the NYC law on AEDT & AI regulation
• Webinar | AI in Hiring: the Fact, the Fiction, and the Fundamentals
• AI in Hirevue products
• Infosec Trust Center