Automated hiring software is mistakenly rejecting millions of viable job candidates
| Automated resume-scanning software is contributing to a “broken” hiring system in the US, says a new report from Harvard Business School. Such software is used by employers to filter job applicants, but is mistakenly rejecting millions of viable candidates, say the study’s authors. |
https://www.theverge.com/2021/9/6/22659225/automated-hiring-software-rejecting-viable-candidates-harvard-business-school
Last edited on
The AI is slowly taking over decision making..
I already see headlines of the future: "It was the AI that did this" etc..
I already see headlines of the future: "It was the AI that did this" etc..
"The AI"?
Also, "taking over" isn't quite right, either. It's more like there's groups of people trying to avoid work and delegating tasks to computers when they shouldn't.
Also, "taking over" isn't quite right, either. It's more like there's groups of people trying to avoid work and delegating tasks to computers when they shouldn't.
A good book about this sort of thing...
Hello World How to be Human in the Age of the Machine
by Hannah Fry
Hello World How to be Human in the Age of the Machine
by Hannah Fry
Last edited on
| It's more like there's groups of people trying to avoid work |
Right, but that "group of people" is even more scary than the AI.
Ive been on again, off again with AI for many, many years.
My non PHD conclusion is that
- for many difficult problems, it can get between 85% to 90% correct results
- for every problem where it gets above the 90% mark (by a good margin, eg up towards 95-100% results), all the AI does is save writing complex code to do the same thing (eg OCR, you could do it without the AI, its just a lot of crunching).
AI does not understand text documents. It does not understand anything. It is just doing some clever pattern matching/memorization tricks and voodoo that look smart on the surface. The machine cannot think, and nothing we have done to date gets close to 'thinking'.
So nothing here is unexpected. The tool is being misused, and not working as desired.
My non PHD conclusion is that
- for many difficult problems, it can get between 85% to 90% correct results
- for every problem where it gets above the 90% mark (by a good margin, eg up towards 95-100% results), all the AI does is save writing complex code to do the same thing (eg OCR, you could do it without the AI, its just a lot of crunching).
AI does not understand text documents. It does not understand anything. It is just doing some clever pattern matching/memorization tricks and voodoo that look smart on the surface. The machine cannot think, and nothing we have done to date gets close to 'thinking'.
So nothing here is unexpected. The tool is being misused, and not working as desired.
AI is only as good (or bad) as the people who create and maintain it as well as use it.
AI makes it easier for those tasked to review applications to avoid doing any actual work.
AI makes it easier for those tasked to review applications to avoid doing any actual work.
There is a difference between having a task automated and AI. My definition of AI includes some sort of learning capability. So for the resume scanning software, IMO for this to be classed as AI then there would need to be some way - without changing the program - for it to learn what not to reject and conversely what to accept based upon some kind of feedback to the program. So that the program 'learns' and progressively rejects fewer viable and accepts more viable candidates.
Job aggregator websites would seem to be at least partly to blame for this.
If an applicant can do any random search for "shoe buffer", get a list of matches each with a tick box, and a super handy "select all" check box and "apply for this job" button, then 100's of companies are going to get spammed from 5 seconds of effort on the part of the applicant.
Companies, in an effort to deal with the deluge, resort to any automated system that promises to filter out the irrelevant.
Applicants, realising that companies are using filters, resort to even greater keyword stuffing, applying to many more companies in the hope that just one message makes it to a human.
If an applicant can do any random search for "shoe buffer", get a list of matches each with a tick box, and a super handy "select all" check box and "apply for this job" button, then 100's of companies are going to get spammed from 5 seconds of effort on the part of the applicant.
Companies, in an effort to deal with the deluge, resort to any automated system that promises to filter out the irrelevant.
Applicants, realising that companies are using filters, resort to even greater keyword stuffing, applying to many more companies in the hope that just one message makes it to a human.
Supposedly companies in the US are having a major problem of finding even warm bodies to hire, people are eager to not work, getting paid by the government.
Do you think it'd be better if instead of candidates making the first contact to employers, for employers to make the first contact to posted candidates? The way it's done now is kind of silly if you think about it.
We have the tech to do that now. Everything is still done as if it were 1930 at most places. A lot of peoples' heads would explode at that suggestion, but I think someday we may get to that.
I've had companies contact me without my asking, so it is a thing, albeit rare. But wouldn't you still have the same problem? The company only has so many resources to spend on contacting potential employees, so it will prune the list of people to ask, via automation.
robots are hiring now and failing big time ? just wait until they start firing people ...
Last edited on
Same problem?
I suppose its similar. If you had a large master site of people looking for work (eg, linked in ish) and fixed tags "use these tags for your skillset" then filtering on tags is going to be *fair*. The current problem is that no one knows what tags which AI is using, at least in part. Do you lose out if you don't have crap like "Emerging Technologies" (one of the tags stuffed into my last resume by the resume fixer guy who knew what was needed to get past AI)? Possibly! If you had a master site with known tags all that goes away.
I suppose its similar. If you had a large master site of people looking for work (eg, linked in ish) and fixed tags "use these tags for your skillset" then filtering on tags is going to be *fair*. The current problem is that no one knows what tags which AI is using, at least in part. Do you lose out if you don't have crap like "Emerging Technologies" (one of the tags stuffed into my last resume by the resume fixer guy who knew what was needed to get past AI)? Possibly! If you had a master site with known tags all that goes away.
Topic archived. No new replies allowed.