Comment by icegreentea | Hacker News Reader

Comment by icegreentea | Hacker News Reader

Parent Full thread

Once you have a sufficient percentage of a job market gated by these types of algorithms, the individual worker literally cannot win. Adversarial approaches to 'beating' the algorithm can be detected and compensated for within the algorithm. Ironically, it'll just be right back to forming unions, except you'll be negotiating against algorithms.

Seriously, almost any type of 'optimization' on the half of companies/employers right now is bad news for employees and job seekers. We call it optimization because its nice and clean. But the effect of every optimization is to squeeze out every last bit of productivity out of a given work force as possible.

Honestly, I hope this type of thing is never developed and deployed to the degree to which some of the proponents in the article wish it to. They can paint starry eyed pictures of a future where everyone gets to work the job that fits them the best, but all I can see is a future where everyone who has a job is scared shitless of losing it, since it'll blackmark them forever.

Thats a highly cynical view. There's plenty of optimization problems that aren't about squeezing out productivity at the cost of anyone, but rather about improving productivity and employee satisfaction at the same time.

For instance - lets say that I realise I have a problem with turnover that I want to fix, because my costs of replacing staff are too high. To address that I want to spend 1M on retention activities in a year. Should I spend that on additional vacation days, or should I spend that on extra events for the staff, or on extra training opportunities? If I spend that money, will the impact on my turnover costs be positive enough to warrant the spend?

That's the kind of discussions that always pop up. Being able to quantify the impact would make it easier to do the right thing that both benefits employees and the employer.

Oh, I agree that there are ways to improve productivity and employee satisfaction at the same time, and some companies might choose to do that.

But let's be real - there are likely more people that jobs that they are willing to do in the US for the short to medium term. As the 'jobless' recovery showed, companies were perfectly able to squeeze out additional 'productivity' (ok, I'll agree that the way we measure market performance isn't really that great, but it's the metrics that we largely agree to play the game by) while cutting labour force, and maybe even per-worker pay, and likely driving down employee satisfaction, other than how glad they were to still have a job.

As long as 'hey, look be glad that you have a job' is a legitimate threat, for the majority of workers, they're really out of luck, cause again, let's be real. The right thing that benefits both employees and the employer is not the same thing as the right thing that benefits both the employees plus the guy you just fired and the employer.

This type of optimization has been happening since the beginning of the assembly line or even before it. Plus, employee/employer relationships are always adversarial. I want as much as I can get out of the company and the company wants to get as much out of me as possible.

As for it going into production I don't see why it is a bad thing. You could turn the whole thing around and help job seekers find companies where they know they would be happy and productive. After all it is a similar kind of optimization but from a different perspective.

When I saw the the headline that's what I thought the article was about: "Oh joy, someone has found a way to use hadoop to better inform job seekers."

Think of the algorithm as playing a game against a human, where the human players' goal is to maximize compensation/opportunities/generally feeling good or at least ok about work. The algorithm's goal is to maximize the business's surplus (productivity minus cost). This game is competitive (but not zero sum).

You are claiming that the human cannot beat the algorithm, because the algorithm can always adapt. But humans are still better than computers at some games (Arimaa, Havannah, Hex, Go) and worse at others (Chess, Checkers). So your claim needs a lot more justification.

The comparison is worrisome: depending on how good these algorithms get and how zero-sum the competition is, we could be screwed. But we don't know.

I wonder about that. There'll be a motivation for companies to defect.

If I can get awesome people to work at my company just by offering them a reasonable amount more than at Evilcorp, I'm just gonna do that. If it forces Evilcorp to raise their wages, then that's good for me - EC is spending more to get what they'd get anyway - and if they come work for me then that's good too.

Wage fixing, is a problem, don't get me wrong. However, it's a problem regardless of how efficient you are in hiring.

Ironically, it'll just be right back to forming unions, except you'll be negotiating against algorithms.

An algorithm can't 'negotiate' with a strike, or other more extreme forms of trade union actions (like burning down the compeditors, shunning scabs, etc.)

Reply on news.ycombinator.com