Many people are proposing alternative solutions to this problem in this thread so here is my attempt:
I see the driving force of the data collection as the desire to make more money selling ads which is done by constantly trying to come up with new kinds of ads to sell and with trying to sell ads as better targeted (or offering better metrics on whether they worked or not) and therefore worth more money. I propose 1. a way to make increased targeting less valuable and 2. a way to make this data collection offer something slightly closer to the privacy people had before the internet:
1. An entropy based tax on selling advertisements. The tax should be proportional to minus the logarithm of the proportion of the ad-seller’s audience whom the ad is expected to reach. (Plus some requirement of auditing these numbers or updating them if they turn out to be wrong). This would mean that eg newspapers would not have a high tax because their ads go out to all their readers (but if they sold ad-slots which were region-specific then the tax would be higher), but some firm uploading a (small) list of emails they collected to google or fb would have to pay a higher tax as google/fb have a lot of users not on that list.
2. Behavioural data may not be used or sold when it is more than 30 days old. It may be used (in a non-ad-targeting) capacity after this time when it is part of the official records of the company that has it (so eg your bank can keep your transaction history and let you query it). Same goes for inferences or other derivatives from that behavioural data. This would hopefully add some forgetfulness back into the system.
I see the driving force of the data collection as the desire to make more money selling ads which is done by constantly trying to come up with new kinds of ads to sell and with trying to sell ads as better targeted (or offering better metrics on whether they worked or not) and therefore worth more money. I propose 1. a way to make increased targeting less valuable and 2. a way to make this data collection offer something slightly closer to the privacy people had before the internet:
1. An entropy based tax on selling advertisements. The tax should be proportional to minus the logarithm of the proportion of the ad-seller’s audience whom the ad is expected to reach. (Plus some requirement of auditing these numbers or updating them if they turn out to be wrong). This would mean that eg newspapers would not have a high tax because their ads go out to all their readers (but if they sold ad-slots which were region-specific then the tax would be higher), but some firm uploading a (small) list of emails they collected to google or fb would have to pay a higher tax as google/fb have a lot of users not on that list.
2. Behavioural data may not be used or sold when it is more than 30 days old. It may be used (in a non-ad-targeting) capacity after this time when it is part of the official records of the company that has it (so eg your bank can keep your transaction history and let you query it). Same goes for inferences or other derivatives from that behavioural data. This would hopefully add some forgetfulness back into the system.