Disclaimer: this is a thought exercise. I don’t encourage anybody to actually implement this kind of a system.
In the city where I live, you pay the parking fee by purchasing a parking slip from a nearby vending machine and putting it on the windshield. The slip shows how long you have, and you’ll get a ticket if the parking enforcer happens to walk by after that.
Outside of office hours, or outside the city center I sometimes don’t bother to pay the fee. It seems it has paid off; I haven’t gotten any tickets. For example, in an area where there’s a 1% chance of getting a ticket of 40 euros during 1 hour parking, it doesn’t make sense to pay a fee of more than 40 cents per hour. But how do I know the probabilities accurately without trying it out 100 times, separately for each parking zone and different times of the day?
By harnessing the power of the masses of course, providing us with enough data to make more fine-tuned decisions. Here’s how it would work:
- Park your car
- Launch the mobile parking fee optimizer application in your GPS-enabled handset and indicate that you have parked
- The app tells you whether you should buy the parking slip or not, do as it says
- Go do your business
- When you return to your car, you launch the application again and indicate whether you got a ticket or not
The back-end will collect information about the parking incidents. It knows how long, when and where you parked, whether you payed the fee, and whether you got a ticket. Assuming the enforcers patrol some areas and hours more heavily than others, the system will learn where and when it is beneficial to pay the parking fees, and where it’s statistically cheaper to not pay the fee and take the occasional parking ticket instead.
The system should of course “forget” older data, so as to reflect changes in the patrol schedules. And inevitably also “malicious” data (from whose perspective? ) would be injected, but assuming the majority of the data was clean and consistent, I would guess that it’s possible to filter the junk out as noise.