Algorithmic Hiring and the New Labor Market Matching Function

Few things are more stressful than searching for a full-time job as an individual who is currently unemployed.  In previous generations, finding a job meant looking for open positions in the “Help Wanted” section of the classified ads in newspapers.  Over the last twenty-five years or so, however, the job market has gone increasingly online.  Today, job-seekers browse digital job listings on a myriad of job sites, ranging from LinkedIn to Indeed to social media.

Efficiency of the Factor Market for Labor

Pre-Internet Era: Inefficiency due to Asymmetric Information

Up through the 1990s, the job search, conducted through newspaper classified ads, job fairs at secondary and post-secondary schools, and posted Help Wanted signs on business doors, was primarily local, especially for entry-level jobs.  Without Internet communication tools like Zoom and Google Meets, most job interviews had to be conducted in person.  Although posted jobs in the pre-Internet era had fewer applicants, this was not necessarily inefficient due to the fact that these applicants were local and could begin work immediately.

However, there were plenty of complaints regarding a local- and in-person-only job market: data about applicants could be lacking.  Before the Internet, it could be difficult for employers to get accurate and up-to-date information about applicants’ backgrounds.  In economics, this is a situation known as asymmetric information, which creates inefficiency.  Prior to the convenience of high-speed Internet, more hiring decisions had to be made based on “gut feelings” and assumptions rather than verifiable data.  This could result in many candidates landing jobs based more on charisma and appearance, or a stylish but potentially fraudulent resume, than objective qualifications and performance metrics.

Internet Era: Inefficiency due to Information Overload

The rise of high-speed Internet in the early 2000s made it easier for employers to verify data about applicants.  Files and queries could be sent via email, previous employers could be looked up online, and academic institutions could be easily vetted.  Job openings could be posted on websites, allowing applicants to apply from anywhere.  At first, this generated excitement for employers - the talent pool was potentially unlimited!  Prospective employees no longer had to be sourced locally; talent could be found anywhere.

Unfortunately, the ease of posting jobs for applicants nationwide quickly led to an overwhelming influx of applicants.  Soon, employers were wading through online applications bearing reams of data, making it hard to process it all.  Instead of picking among three or four local applicants, a business owner might struggle to pick among thirty or forty applicants from across the region.  And, similar to the pre-Internet era, fraud and “puffery” abounded.  Although it was now easier for employers to verify applicant data, applicants could also use digital tools to more easily back up their claims through forged data.

AI Era: Automation Handles Information Overload, but Creates New Concerns

In recent years, as applicants began spam applying to hundreds of jobs using new digital tools, employers responded by resorting to restrictive filters that automatically deleted most applications and resumes.  This factor market “arms race” led to frustration for both laborers and firms: laborers felt that they were applying for new jobs uselessly into a void and firms felt that they were being deluged by applications and resumes of dubious quality and interest.  For applicants, the marginal cost of applying to additional jobs was near zero, while the firm also faced near zero marginal cost for setting overly restrictive filters - there would always be more applicants!

Now, AI has entered the picture and is a tool for both job-seekers and employers.  As opposed to traditional filters, AI can theoretically review applications and resumes holistically and deliver better applicants to employers.  However, concerns arise when it comes to the final steps of reviewing applications: are hiring managers making the final decisions, or are they letting AI handle the whole process?  This potential over-reliance on AI as hiring managers could lead to adverse hiring, especially as applicants learn how to “game” AI or the AI algorithm resorts to bias or prejudice against certain groups of qualified applicants.

Asymmetric Information in the AI Era: Algorithm Gaming is the New Charisma

For better or worse, “gaming” the hiring market still exists in the AI era.  While charisma and [assumed] personal connections helped under-qualified applicants land localized jobs in the 1990s and earlier, understanding AI algorithms and the key words they seek can do the same for potentially under-qualified applicants today.  Prior to high-speed Internet, physically attractive and charismatic applicants could wow employers with a firm handshake and smooth talk.  In the AI era, tech-savvy applicants can use AI to optimize their resumes and cover letters, allowing digital “smooth talk” to hide insufficiencies in their actual experience and performance data.

Therefore, despite dramatic advances in technology, the sorting function of the labor market remains imperfect.  Because many job applicants can access similar digital tools to those used by employers, they can continue to use puffery and fraud to mark insufficient skills and experience to land desirable positions.  In all eras, this job applicant masking advantage has gone to those with more wealth and resources, continuing the trend of low-income and minority applicants struggling to land desirable jobs for which they are qualified.  People with higher income are more likely to be familiar with productive AI tools, giving them continued advantages in the job market separate from their job-related skills.

Potential Societal Dilemma: Long[er]-Term Unemployment

The growing use of AI by hiring managers creates the possibility of an AI divide in the labor market.  Those who understand how to use AI to create resumes, cover letters, and fill out online applications will be perpetually advantaged, quickly rising through the employment ranks.  Meanwhile, those who do not effectively use AI to optimize their job applications may be perpetually blocked from full-time jobs for which they are actually qualified.  This could result in long-term unemployment for those who fail to use AI to pack their resumes and cover letters with the key words sought by hiring managers’ AI programs.

Speed may also create an AI divide, with hiring advantages going to applicants who can use AI tools to fill out job applications meeting select criteria as soon as they become available.  If hundreds or thousands of AI-savvy applicants are using this tool, their applications and resumes may flood hiring managers’ inboxes within seconds, leaving the non-AI applicants far down the queue.  This could mean older unemployed workers, years past learning AI in school, never having their applications entered quickly enough to make it into the window of applications viewed by a hiring manager.