Job candidates strongly crave transparency in the use of ADM in recruitment, wanting to know when and how automatic decision-making is used, what data is being processed and how decisions are being made.
These findings are from new research by the Information Commissioner’s Office into understanding public perceptions towards ADM in recruitment, released yesterday alongside the ICO’s strategy on the use of AI and biometrics.
As recruitment processes increasingly use Al to screen CVs and application letters, conduct initial interviews, and assess candidate suitability, concerns have been raised about fairness, transparency, and accountability.
ADM in recruitment
CIPD to lead research into responsible AI adoption
The ICO said it is particularly interested in understanding how these technologies impact individuals, especially marginalised groups who may be disproportionately affected by biased algorithms.
It commissioned Revealing Reality to gather qualitative evidence of how people experience ADM in recruitment and the effects they feel it has had on them. Research explored reactions to different types of ADM tools, whether they had experienced them directly or not.
They also captured evidence of people’s expectations around the use of ADM – particularly in relation to Article 22 of the UK General Data Protection Regulations, which safeguards individuals’ rights in relation to ADM.
The study involved remote focus groups with 33 job seekers, divided into four groups to represent a range of employment experiences and attitudes.
While many of the job seekers acknowledged the increased use of ADM, they generally had a limited understanding of how it worked, often perceiving it in a binary way of either human decision-making or fully automated.
Some saw potential benefits in early-stage filtering but there were significant concerns about using ADM for assessments and final decision-making. All participants were strongly negative about fully-automated decision-making in recruitment.
Most expressed concerns about the confusing nature of assessments, where tests were perceived as illogical, and expressed frustration with the design of recruitment tasks, such as games.
One member of the graduate group said: “I remember one of the games I got was you had to pump up a balloon. You had like two minutes and you pumped up balloons consistently by just clicking your mouse and they would burst at different points …
“The point was to try and pump them up as much as you could, then let them go and keep on going without bursting … for a graduate role, I thought, is this really … a measurement of skill?”
Some participants believed they had experienced ADM tools, but felt the overall process lacked transparency. “The speed on the turnaround time of a response and it’s an automated rejection makes me think it’s ADM,” said one participant.
“Getting identical rejection emails across many different jobs is a sign [that it might be ADM],” said another.
However, while participants thought ADM was being widely used, none could recall seeing any information about this while making applications. Many believed there was a lack of transparency from employers.
“I’m confident that it’s being used, but I’ve not seen any transparency,” said one of the graduates.
The researchers highlighted key areas of concern and emphasised the need for careful consideration of ethical implications and best practices in the design and implementation of ADM systems:
- Transparency is paramount: Participants stressed the need for transparency regarding the use of ADM in recruitment processes.
- Human oversight is essential: Despite recognising potential efficiencies, participants believed that human involvement remained crucial for ensuring fairness, addressing biases, and providing individual feedback.
- Concerns about bias must be addressed: The potential for ADM to perpetuate or amplify existing societal biases was a major concern. There was a strong expectation for ADM systems to be fair, unbiased and non-discriminatory.
- Candidate experience matters: Participants’ experiences in the recruitment process significantly influenced their perceptions of ADM systems. Issues, such as lack of communication, vague feedback, and impersonal processes negatively impacted attitudes towards ADM.
- ADM’s appropriateness is contextual: Participants had mixed views on the appropriateness of ADM at different stages of recruitment. While “light-touch” ADM for initial filtering was seen as acceptable, there were significant concerns about more comprehensive or fully automated systems.
Over the next year, the ICO said it will consult on an update to its ADM and profiling guidance, and develop a statutory code of practice on AI and ADM.
Sign up to our weekly round-up of HR news and guidance
Receive the Personnel Today Direct e-newsletter every Wednesday
Recruitment and resourcing opportunities on Personnel Today
Browse more recruitment and resourcing jobs
