- Published
I have lost track of the number of people who have told me they have applied for hundreds of jobs, only to be met with rejection after rejection.
asserted
they → publish → rejection
Often, they said it felt like a human had not even looked at their applications.
asserted
human → say → applications
Some match candidates to jobs, others grade and screen CVs.
asserted
others → match → CVs
Employers use them to save time sifting through applications, but there are growing concerns that these tools could be biased.
uncertain
tools → use → applications
I want to find out how fair AI recruitment tools being used to help decide our futures really are - and what happens when we click "apply".
asserted
we → want → futures
A London-based non-profit organisation, People Like Us, is raising awareness of people's legal right to challenge automated decisions when they disagree with the outcome.
asserted
they → base → outcome
A survey conducted by Censuswide among 2,000 UK workers who had applied for at least one job in the past 24 months found that 75% were unaware of their right to challenge an automated rejection.
asserted
% → conduct → rejection
Sheeraz Gulsher, CEO of People Like Us, is leading the campaign called Reject the Rejections, which calls on employers to improve transparency and accountability in the use of AI and automated recruitment systems.
asserted
which → lead → AI
He describes the current situation as an "AI war", with candidates "resorting to AI to bolster their applications" while "employers are dealing with huge amounts of applications".
asserted
employers → describe → applications
Gulsher said: "We can't say for certain that every company is operating its automated or AI systems without bias.
asserted
company → say → bias
That's why we're calling for companies to audit these systems, publish the results and have them independently assessed.
asserted
them → call → results
"In our research, we found guidance from the Information Commissioner's Office, external stating that people subject to certain automated decisions have the right to request a human review.
asserted
people → find → review
We're trying to remind people of that right, as many simply aren't aware of it."
asserted
many → try → it
The research also found 70% of ethnically diverse jobseekers had altered or concealed information on their CVs, including their country of origin, religion or name, out of fear that recruitment software or AI tools would reject their applications.
asserted
software → find → applications
This compares with 57% of white candidates.
asserted
This → compare → candidates
Would people ask for a human review?
asserted
people → ask → review
Business consultant Vivi, who lives in Lewisham, was made redundant earlier this year.
asserted
who → live → Lewisham
She says she has applied for so many jobs that she's given up and started her own company.
asserted
she → say → company
"There's one instance where applications closed on Friday and I got a reply on a Saturday morning basically saying that I should work on my skills and come back in six months.
asserted
I → be → months
But it was so obvious that nobody had assessed my skills.
asserted
nobody → assess → skills
"
If she had known she could ask for a human to review her application, she might have done things differently.
uncertain
she → know → things
"I would definitely go back and ask for a human to look at my application and if it is a 'no', that's fine, but feedback helps me as a jobseeker."
asserted
feedback → go → jobseeker
Hema, 49, from Surrey, is an HR and talent professional who has worked internationally across multiple sectors.
asserted
who → work → sectors
She was made redundant in June and has since tracked more than 100 job applications and recruitment opportunities.
"
asserted
She → make → applications
A lot of them have been automatically rejected within 24 to 36 hours."
asserted
lot → reject → hours
She added: "I've always landed a job in 10 applications, so to be nearing 100 is completely out of this world."
asserted
nearing → add → world
While she is "hesitant" about asking for a human review in case she jeopardises her application, she believes "a human recruiter can see the fuller picture".
asserted
recruiter → ask → picture
"If you've got an in-house recruitment team, they know the culture, they can see the type of language that you're using in your CV, in your cover letter, during a phone screen."
How much do recruiters use AI?
Of the 425 recruiters CV Library spoke to, about 90 said they used AI in the CV screening stage.
asserted
they → get → stage
Alec Stanwell, director of marketing, believes there is a "perception gap".
asserted
Stanwell → believe → marketing
"Candidates almost universally believe that employers are using AI to screen their CVs, filter them and to screen them as they go through the process and that is happening but nowhere near as much as candidates think it is.
asserted
it → believe → process
That perception gap is a problem because candidates are now writing their CVs, and applications, through the lens of 'an AI bot is going to be reviewing this rather than a human'.
asserted
bot → write → human
He adds this is a problem because recruiters are seeing all CVs look the same because "they're all written by AI, or polished by AI".
asserted
they → add → AI
He says this creates a "perfect storm" because AI is then more likely to be used to filter through a large volume of applications.
asserted
AI → say → applications
"We're going to end up with computers hiring computers, but we're actually in the world of hiring people."
asserted
we → go → people
In the four years to December 2022, Stanford researchers analysed four million job applications processed by an AI hiring tool, external.
asserted
researchers → analyse → tool
Some candidates were rejected from multiple roles more often than would have been expected if they had been assessed by each employer independently.
asserted
they → reject → employer
The analysis also indicated black and Asian applicants were recommended for jobs at lower rates.
asserted
applicants → indicate → rates
They only looked at one AI hiring tool, and other factors such as education and immigration status could have also played a role.
uncertain
factors → look → role
The hiring tool assessed did not respond to our requests for comment.
asserted
tool → assess → comment
The people I have interviewed are not talking about any company in particular, but concerns over AI used in recruitment are also playing out in the courts.
asserted
concerns → interview → courts
…and 25 more, not listed.
The Aporia
