A lot of recruiters go there an think they could show their HR buddies how they can increase productivity by using AI.
Placing the CV text into chatGPT (OMG #GDPR !) and watch the magic happen.
So, they find it cool, and want a fully automated process to figure out, who is the best candidate.
It's so cool to have that, when one needs to review 200+ CVs.
Mirror mirror on the wall
Who is the most fitting of them all?
The magical bot will display you the most fitting one. It just needs to fit to the job description.
The machine has ranked it for us, so we can just pick from the top.
But that's already too much. I just want the most-fitting candidate!
What I want to see is the actual candidate. The machine should pick for me and reject the rest.
(I ignore the fact it was a different one this time!!!!11eleventy)
We train out bot to create standard rejections, so we can also automate that part. Also, we don't clean the data and remove personal data from training and also not from actual execution.
Meaning, the bot knows what address, or date of birth or gender has a higher chance of unprofessionalism.
And so for this, the EU has placed two significant laws, that recruiters and companies targeting recruiters as customers.
I hope by now, that many citizens know what GDPR is. Actually, I can't trust you on this either, I have seen some digital services that simply aren't even though claiming it.
And the EU AI Act says, that #recruitment is categorized as "high-risk".
"AI applications that are expected to pose significant threats to health, safety, or the fundamental rights of persons. Notably, AI systems used in health, education, recruitment, critical infrastructure management, law enforcement or justice. They are subject to quality, transparency, human oversight and safety obligations, and in some cases require a
...
A 2nd quote to note:
"Before placing a high-risk AI system on the EU market or otherwise putting it into service, providers must subject it to a conformity assessment. This will allow them to demonstrate that their system complies with the mandatory requirements for trustworthy AI (e.g. data quality, documentation and traceability, transparency, human oversight, accuracy, cybersecurity and robustness). This assessment has to be repeated if the system or its purpose are substantially modified"
This is the source, of the quote:
https://ec.europa.eu/commission/presscorner/detail/en/qanda_21_1683
And it basically says, you can not do AI without humans.
A thumb of rule is to follow the principle that: "Machines shouldn't decide for humans." Especially in high-risk situations.
Let's get that example from the European Court of Justice:
https://www.deutschlandfunk.de/schufa-score-eugh-urteil-100.html
where the Scoring company was sued for automating decision processes.
Fun fact: Scoring is on the level "Unacceptable risk" – AIs on this level are banned.
A decision is also then a decision
- if you get a sole suggestion
and
- if you have a ranking, where a candidate is ranked first and so on.
So, what does it mean for recruiters and AI builders for recruiters?
So AI needs to create insightful information out of the data base that does not harm the mentioned human rights.
It needs to be explainable, understandable, with all the conformity assessment properties mentioned.
With human-oversight.
devbirgit
