- perfect accuracy (no FP; no FN)
- automated enforcement via a fine, without knowing the identity of the perpetrator (just assume it's possible; I know it's hard)
- no potential for the device or underlying technology to be used or repurposed for any other form of surveillance or enforcement
Would such a system (and automated enforcement) be OK?
Yes. But as you pointed out, `automated enforcement via a fine, without knowing the identity of the perpetrator` is hard.
And as others have pointed out, Point 1 and Point 3 are also difficult to accomplish.
Question for you is, are you okay for the government to roll out a mechanism to detect and prosecute people changing lanes without giving you any guarantee of respecting the 3 points you mentioned above?
If your answer is no, you are not okay with it, then we are all on the same side.
If your answer is yes, then you should read the article above by moxie (very interesting article)
I'd be happy to compromise a bit on point 1. I could accept a high (20%) false negative rate. If the fine were small enough then I'd be OK with a small number of false positives (as long as they're randomly distributed, not hammering one person who happens to live near a faulty camera).
I wouldn't want to compromise on point 2 or 3, particularly for this example (catching impolite drivers).
But what if the example were about a highly accurate system for detecting kidnappers or people traffickers or something? It seems harder for me to argue against enforcement on the grounds of privacy or slippery slope.
That is always the argument for total surveillance state. In Singapore, crime rates are close to 0, but to be effective it means no privacy and everybody is monitored. I'm not okay with this.
Would there be a mechanism in place to remove your fines and points? Would that system be automated? Would it be susceptible to lying? How long would it take to get the record corrected? Would the effort cost more than the fine?
Just from a technical/automated point of view, removing the moral and societal aspects, the real world is currently just way to complex and ambiguous to automate things like this.
Also on a side note: "...the fine could be small enough (e.g. $2) to be insignificant for most of us..." -> I once said something like that to someone and he pointed out that statements like that was very hurtful to hear, especially coming from someone who does not understand his financial struggles.
Your second example (speeding) doesn't incur a fine because in this hypothetical example the automated enforcement is only used for signalling violations.
For your third example (self defence), maybe you are forced to change lanes without signalling. How often? Once per year? What if the fine were $2, and people like you had to pay it once per year, but people who persistently refuse to signal end up paying 1000s of $2 fines per year?
Seems like a reasonable trade-off.
My point is that by automating fine dispensaries, you are subversing the underlying idea itself (automatron can't know what or why is it enforcing) and reasonability.