Because they are famous for low specificity and low sensitivity.

Is a score on a video game quantifiable enough for you? There are so many "brain training" games that do nothing but score you on your observation, working memory, reflexes, etc. Back in college, everyone had to participate in the psych experiment test pool, where many of the experiments were coded in extremely rudimentary Visual Basic, to show video clips and record the time intervals between on-screen events and keypresses.

That's more objective than a human checking your eyeballs to see if there are too many involuntary movements, or making a subjective determination that there weren't enough social displays of submission to the tester. And it scales better to have X computers set up for employees to test and clock in at the same time than to have a supervisor processing the whole line one by one.

I suppose the system could compare your current score with your baseline / sober scores. Or, mentally slow people would simply have less tolerance for being fucked up at work.

All everyone is talking about is weed which I’m sure most of us agree shouldn’t be tested for but we don’t even know what the failed tests failed for.

And no one is required to work for you at the wages you can afford to pay...

Want affordable workers? Increase supply or decrease demand. If you can't do the latter, focus on the former.

If labor is turning into a seller's market, buyers are going to have to adjust.

Hell, you could set up an old Atari with Frogger in it, and require employees to score X points before clocking in. It might not have a very strong correlation, but it sure would be cheaper and more goal-specific than drug testing.

It would more likely take the form of "watch this simulated surveillance video of the work floor, and press the space bar to stop the machinery on the entire production line". When someone steps over the warning tape on the floor around the person-shredder, you press the button. If someone gets their sleeve stuck in the dude-grinder, you press the button. If the part-spewer starts stamping out misaligned doohickeys, you press the button. If your response time is too slow, you aren't fit to clock in for today's shift. Or it could be a simulation of operating the actual machine that employee uses, attempting to predict their productivity for the day.

It has the added benefit of sneaking safety training videos into the daily workflow.

I am not even aware of any jury trial that has decided on the impact of random drug screenings. As far as I am aware, most claims relating to workplace injuries are settled long before the jury pool is even tapped, and many of the remainder are settled before the jury gets a chance to reach a decision. So I think the question you should be asking is if it is likely that any actuary would be willing to analyze the experimental data from such a program, to adjust their company's insurance premiums for anyone using similar software.

If a correlation can be shown, the premiums would be discounted. Actuaries are better with the math than juries are, after all.

A labor union or government regulator could possibly demand such a system in lieu of random drug tests, but your average company does not have enough commitment to worker safety--for its own sake, rather than for its impact on their worker compensation insurance premiums--to experiment in this fashion. They would need a reason that increases revenue or cuts costs.

Does such a thing exist? Is there any independent evidence that it actually works?