---

Artificial intelligence is rapidly becoming part of the hiring process, promising employers a faster way to sort through hundreds or even thousands of job applications.

But for older job seekers, that efficiency may come with a hidden cost.

AI-powered recruiting systems can screen résumés, rank applicants, administer assessments and recommend which candidates should move to the next stage. Depending on how those systems are designed and trained, they can also identify characteristics associated with age and potentially push older applicants down the list — without anyone explicitly telling the computer to reject people because they are older.

The concern is becoming more significant as AI spreads through the workplace. An AARP survey published in March found that 88% of U.S. employers said their organizations were already using AI in some capacity, with another 10% planning to adopt it.

---

How AI can figure out an applicant’s age

An employer doesn’t necessarily have to ask, “How old are you?” for an automated hiring system to make a fairly good guess.

A résumé may contain a college graduation date, decades of employment history or technologies and job titles associated with a particular period. Years of experience can also become a proxy for age.

In testimony before the Equal Employment Opportunity Commission, AARP’s Heather Tinsley-Fix warned that algorithms can pick up on information such as birth dates, years of experience and graduation dates. Those patterns can cause a system to lower an older applicant’s ranking or eliminate the person altogether. 

Even the language used on a résumé can matter.

AARP recommends that employers using automated résumé screening include terminology common over the previous 15 years rather than relying only on the latest workplace jargon. Otherwise, an experienced worker who describes a skill using an older term could receive a lower score even though the underlying skill is the same.

---

AI can learn an employer’s previous biases

Perhaps the bigger risk comes from the data used to train an AI system.

Suppose a company historically hired mostly workers in their 20s and 30s for a particular position. If an AI system is trained to find applicants resembling previously successful candidates, it could learn that younger candidates are preferred.

No programmer has to write a rule saying “reject applicants over 50.”

Instead, the algorithm may discover combinations of characteristics that correlate with younger workers and give those candidates higher scores.

AARP has warned the EEOC about exactly that type of feedback loop. If recruiters have historically favored younger applicants, an algorithm trained on their decisions can detect those patterns and reproduce them on a much larger scale. 

AI-based assessments can create another problem. Systems analyzing online tests or even interview performance could disadvantage older candidates if the definition of a successful applicant was developed primarily from data involving younger workers.

---

Older applicants may never know

For job seekers, one of the most troubling aspects is that the discrimination can be nearly invisible.

A person submits a résumé and receives an automated rejection. The applicant may never know whether a recruiter saw the résumé, what score the AI system assigned or what factors caused the rejection.

The problem can occur even earlier. Automated advertising and recruiting systems can determine which people see a job advertisement. The EEOC says it is illegal to recruit workers in a manner that discriminates based on age, and job advertisements generally cannot express preferences that discourage people 40 and older from applying.

---

Using AI doesn’t erase an employer’s legal obligations

Federal law provides some protection.

The Age Discrimination in Employment Act generally prohibits covered employers from discriminating against workers and job applicants who are 40 or older. That protection applies to hiring as well as firing, promotions, compensation and other employment decisions. 

And turning a decision over to software doesn’t necessarily provide a loophole. The EEOC says employment tests and selection procedures can violate federal law when they intentionally discriminate against older workers. Even apparently neutral procedures can run afoul of the law when they disproportionately exclude applicants 40 and older, and the employer cannot establish the required justification.