The Hidden Age Filter Checklist

01 The four places an age filter hides

According to the U.S. Equal Employment Opportunity Commission (EEOC), one company's software was set to automatically reject women aged 55 or older and men aged 60 or older. The company denied it and settled. That rule was typed. Today, nobody has to type it. It sits in settings that look harmless:

The knockout question. "How many years of experience do you have?" A career coach on 60 Minutes Australia described someone who answers twenty years going straight to the no pile, before any person reads the resume. The filter. A University of Melbourne study found tools that let recruiters filter by year of graduation, years of experience and seniority, "all of which are proxies for age". A proxy is a detail that stands in for the thing you are not allowed to ask. The date window. "Work experience in the last five to ten years." Applicants are already cutting everything before it. The ad. Phrases like fail fast, fast-moving and a fresh perspective tell older applicants not to bother.

One person's bias touches the resumes they read. A setting touches every single one, and nobody's watching it.

02 List every setting that rejects without a person

Copy every rule in your hiring software into one sheet. Most teams have never seen them all on one page.

Auto-reject inventory

Setting | Type | What it rejects | Real job minimum? | Keep / Change
1. [ ] | Knockout question | [ ] | [YES/NO] | [ ]
2. [ ] | Filter | [ ] | [ ] | [ ]
3. [ ] | Minimum or maximum range | [ ] | [ ] | [ ]
4. [ ] | Ranking cutoff | [ ] | [ ] | [ ]

The decision rule. You don't drop the floor. A minimum the job actually needs, like three years running a team, stays. What goes:

Any ceiling on years of experience. A maximum rejects people for having more.

Any date window. "Experience in the last five to ten years" hides everything before it.

Anything that works like "graduated after". Graduation year is one of the proxies the Melbourne study names.

03 Read your job ad as if you were 55

Cut the age-coded words. U.S. rules on job ads already say that words like young or recent college graduate break age discrimination law, unless an exception applies. Watch for the softer ones too: fail fast, fast-moving, fresh perspective, fast paced environment.

If a chatbot helps you write it, never ask it which age group to hire. When a Melbourne professor asked ChatGPT that, it suggested 21 to 30 and 30 to 45, and nobody over 45. Ask it what the job needs.

Ask what the job needs

Here is the role: [JOB TITLE AND THE WORK IT DOES].
List the skills and the minimum experience this job actually needs.
Then write the ad without any words that hint at an age,
like young, recent graduate, fresh, fast-moving or fail fast.

04 The twin test and the four-fifths smoke alarm

Run the twin test. It's how that first case was caught.

Make two resumes that are identical except for one thing: a graduation year from the nineties on one, from the twenty-tens on the other. Run them through a test posting, never a live job. Compare what comes out. If one scores better, you've found your filter.

Count, don't guess. Ask your vendor for its adverse impact report, broken out by age. If you look at your own numbers, keep them as totals, never as notes on people.

The rule of thumb regulators use for race and sex: if one group passes a step at less than four-fifths (80 percent) the rate of another, look closer. Worked example: applicants under 40 pass the knockout at 60 percent, applicants 40 and over at 45 percent. 45 is three-quarters of 60, below four-fifths. That's a smoke alarm, not a verdict.

I'm not your lawyer. At any real size, do this with an employment lawyer.

05 A person at the exit, and a log of what you changed

Anyone a knockout rejects who still meets the real minimum gets read by a human. That's the answer to volume, not an age filter.

Every week, a person reads a small sample of the lowest-ranked pile. Let the software sort the pile, and have a person check the bottom of it.

If you find something, change it that day. Write down what you changed and when.

Ask your vendor two questions. What does it score on? Has it been tested for age?

What applies depends where you are. In the U.S., federal age discrimination law protects people 40 and over, at employers with 20 or more employees, and state laws can go further. Australia has a federal Age Discrimination Act. New York City's yearly bias audit counts sex and race, not age. California's rules cover age and say records from these systems have to be kept for four years. In Mobley v. Workday, a federal judge in California let an age discrimination claim move forward for applicants 40 and over. Nothing has been proven, and Workday says the claims are false and that its tools don't make hiring decisions. Either way, the settings in your account are yours.

Sources

U.S. EEOC: iTutorGroup to Pay $365,000 to Settle EEOC Discriminatory Hiring Suit Sullivan & Cromwell: EEOC Settles First AI-Discrimination Lawsuit 60 Minutes Australia: How AI is causing an unemployment crisis among middle aged job seekers Sheard, Journal of Law and Society: Algorithm-facilitated discrimination Blackham, University of Melbourne: Over 45 and looking for a job? AI thinks you might be too old 29 CFR 1625.4: Help wanted notices or advertisements 29 CFR 1607.4: Information on impact (the four-fifths rule) 29 U.S. Code 631: Age limits and 630: Definitions Australia: Age Discrimination Act 2004 NYC DCWP: Automated Employment Decision Tools FAQ Mayer Brown: California Adopts New Employment AI Regulations Mobley v. Workday: order on the age discrimination claim

Mathew's LibraryLibrary card

Free access

Keep reading free with your library card

One email opens this checklist and all 3 resources in Mathew's library. No account, no password.