I thought about this a while, and then had an epiphany last night. Without a perfect test, the answer is almost definitely no. It's a fairly simple statistical calculation called Bayes' Theorem. The end result is that you'll end up preventing more people from driving when they aren't drunk than preventing drunk drivers.
I'm going to plug in numbers, but since I'm (likely correctly!) assuming drunk driving is a rare event, the numbers don't really matter that much. I'm also going to assume the test is extremely accurate.
Let's say that in 1/10,000 car trips, the driver is too drunk to legally drive. This is probably an underestimation by a factor of 100, if not more, if you think about how many car trips there are in a day. Let's assume that the when the test is positive, the driver is drunk 99.9% of the time. And then assume that when the test is negative, the driver is sober 99.9% of the time (in other words, if the test is negative, the driver is drunk 0.1% of the time).
We can use this to plug in probabilities for each event.
Probability that a driver is drunk: .0001
Probability that a driver is sober: .9999
Probability that a drunk driver gets a positive test: .999
Probability that a drunk driver gets a negative test: .001
Probability that a sober driver gets a positive test: .001
Probability that a sober driver gets a negative test: .999
Bayes' Theorem applies here. It says:
The probability that someone is drunk driver given a positive test is equal to the probability of a drunk driver gets a positive test times the probability of a drunk driver; that divided by the following: the probability of a drunk getting a positive test times probability of a drunk driver plus the probability of sober driver getting a positive test times the probability of a sober driver.
In mathematic terms (DD=drunk driver; SD = sober driver; + = positive test):
P(DD | +) = (P(+ | DD)*P(DD))/((P+ | DD)*P(DD)+P(+ | SD)*P(SD))
Plug in the numbers:
P(DD | +) = ((.999)*(.0001))/((.999)*(.0001)+(.001)*(.9999))
P(DD | +) = .0908
In other words, the probability of a drunk driver given a positive test is only 9%. Meaning that out of a 100 people that test positive under this test, 91 of them would actually be sober.
Because the test is imperfect and drunk driving is rare, it's going to impact more sober drivers than drunk drivers. Even if the test is 99.99% accurate and as a false positive rate of 0.01%, the probability of a drunk driver given a positive test is only 50%. Note that I'm assuming that 1/10,000 car trips is one by a drunk driver. If you assume 1/100,000 car trips are by a drunk driver, the probability of a drunk driver given a positive test is 0.9%.
(You can also use this calculate to find out the odds that a drunk driver will have a negative test, but I have other stuff to do now...)
So, without a nearly perfect test, it's a bad idea for the entire population.
If drunk drivers were more frequent, then it would make more sense. Hence, it makes sense for someone who is more likely to drive drunk, and why the current policy probably makes sense.
I'd go with this system over either of those. I have seen automatic braking systems malfunction before and it turns catastrophic in an instant. The incredibly unfortunate part is that automatic braking systems also have a disturbingly high number of ways they can be fooled. The two I have seen personally were leaves covering up the sensor (slammed the brakes on someone in town and caused a collision) and bugs from I-70 covering the sensor (the car locked up and the brakes remained engaged until the sensor covering could be cleaned).
As far as GPS-enforced speed limits, this also introduces danger on the roads. It prevents drivers from making effective evasive maneuvers when driving at the speed limit. Malfunctions for this system would also be incredibly dangerous, considering the number one cause of traffic accidents is a differential in speed between the two cars that collided. If one car is limited to 10mph under the speed limit because their GPS glitched out then they just became a sitting duck on the road, though not as bad as the automatic braking malfunction.
I'm fine with mandatory safety measures that don't risk lives compared to the alternative of not having them, such as seat belts and air bags. If those fail you may die, but if they fail you are no worse off than you would have been if the safety measures were never installed.
I draw the line at mandatory safety measures that will actively risk your safety or life when they fail. Automatic braking systems that will slam the brakes in highway traffic. GPS-enforced speed limits that can hamper evasive maneuvers and cause the same symptoms as automatic braking system failures (if an error displays a limit lower than the true limit). And yes, mandatory BAC interlock devices for law-abiding citizens that can leave them stranded and stuck with a very costly repair bill in the best case scenario and death in the worst case scenario.
If we want to talk about personal anecdotes about why it's incredibly important to be able to start you vehicle at any time, I've got the perfect example of how this can risk lives in real scenarios that actually happen.
When I was 17 I took the bus with my friends down to the annual Denver Avalanche game and we hung out at the 16th Street Mall afterwards until we caught the last bus back to where our cars were parked. Having parked in opposite corners we parted ways getting off the bus and went to our cars, my friends having no issues driving home. I, on the other hand, had some trouble with starting my vehicle. You see that year the temperature was 15 degrees below zero and my car was an old (1979) Mercedes 240D diesel. Diesel engines don't particularly like the cold, so I cycled the glow plugs several times before trying to start. No dice, so I repeated that. This went on until my car battery died at around 2 AM, and the worst part of it was that stupidly I was only wearing jeans, a t-shirt, and a sweatshirt with no jacket or coat. The buses had finished their schedules and the park and ride was empty (I was the last car) in the middle of nowhere without areas I could take shelter nearby. I was lucky to have a mylar blanket and a comforter in the trunk of my car that I kept there only because my Grandpa insisted I'd need them if I was ever stranded in the cold.
I wasn't able to get assistance at my location until 5:30 that morning because it was located in the mountain, a lovely cell phone dead zone. 3.5 hours spent in -15 degree weather with only jeans and a sweatshirt. Even sitting in my car without exposure to wind I would have risked frostbite in 30 minutes or less, and that temperature presents a high risk of hypothermia even with proper winter clothing. While wearing winter clothing at that temperature you'll lose one degree of core body temperature about every 30 minutes, sooner if you have no hat. Below 95 degrees (2 hours) is the beginning of hypothermia, below 93 degrees (3 hours) is when amnesia sets in. Profound hypothermia is 90 degrees (4.5 hours) and you'll find yourself no longer even shivering to keep warm. At 86 degrees (6.5 hours) your heart starts to pump arrhythmically. At 85 degrees (7 hours) you'll rip off your clothes for your final minutes of life. Those times are for proper winter clothing.
When an ignition interlock device fails, it WILL kill people in the mountains every single year. People who went camping, skiing, hiking, or hunting and get back to their car in the evening only to have it refuse to start. Cell service is sparse at best in these areas, meaning only those prepared with extra blankets/gear and the ability to start fires will survive through the night without heat from their vehicle. I say when, not if, because the failure rate will be above 0%. 15 million new cars are sold each year, and if the failure rate is 0.01% annually then you'd see 1,500 failures in the first year, growing by another 1,500 every year and providing 1,500 more opportunities to kill in either what was described or other scenarios.
This is exactly why using emotional arguments is dumb, because realistically the number of deaths would be small but a personal anecdote carries additional weight. The point is that any deaths that directly result from a safety device are unacceptable even if that safety device may save lives in other circumstances. Trading lives of innocent and law-abiding citizens because of a small number of criminals is morally reprehensible on every level.