4 ms·

What if Google used all the data they collected and used machine learning to show the voting reminder only to users who align ideologically with Google? That’s

by k_sze 7y ago

What if Google used all the data they collected and used machine learning to show the voting reminder only to users who align ideologically with Google? That’s a big what-if, of course. I’m not even sure if it would be violation of any privacy law because that’s not selling your data to 3rd-parties.

evrydayhustling 7y ago
Companies sponsor campaigns and publicly support politicians all the time. Personally I think campaign contributions should be way more limited, but I'm not sure what is special about Google here.
dantheman 7y ago
Imagine if a newspaper endorsed a candidate?
- k_sze 7y ago
  Except it would be more subtle. Google would not be telling you who to vote for because they already have an idea who you'll be voting for, based on machine learning results. If they think you would vote for someone disadvantageous to Google, they would not show the voting reminder.
  - pvg 7y ago
    It sounds a bit like this degree of subtlety: https://i.imgur.com/OkVMmQi.jpg https://i.imgur.com/OkVMmQi.jpg
superseeplus 7y ago
Even assuming Google is theoretically capable of showing the voting reminder only to users who would vote for a Democratic candidate doesn’t mean that they did that. Dr. Epstein doesn’t present any evidence (even anecdotal) that the reminders were not shown to Republicans.
tptacek 7y ago
It's more than a big what-if: it's also not something Epstein claims has happened.
kingludite 7y ago
Exactly! This is the interesting question to ask! And you should expect this to take a much less straight forwards form than outright malicious intend. Machine learning stuff, to us mere mortals, is a black box that seems to output that what we desire. However, we don't need a single black box to completely hide the effect from everyone involved. It can be simply spread out over many processes without anyone noticing the sum of countless, by themselves very sensible, pro-google decisions made by google and its employees. Say on a fuzzy edge some human or automation has to decide if a website gets banned from the index or not. Since they use adsense or google analytics it becomes possible to contact someone running the site or even, long before the offense, present them with a machine generated report suggesting decline of quality. The google hating webmaster doesn't want all that tracking crap on his site and therefore he might be punished by not having the warning in advance. It doesn't seem harmful, it seems perfectly reasonable, if however you create 1000 such effects the meaningless 0.1% adds up to 100%. Say, we take a user profile that use to frequent sites now banned from the search index. If that profile moves to a new website. Does that merit close monitoring? If the AI is a black box you don't even get to ask the question. It all seems to be fair and logical but the sum of those 2 would boil down to close monitoring websites not using analytics. We would praise the AI for noticing it so soon and so efficiently. And so things further escalate without anyone connecting the dots. To make it more sinister: We have a system of pro-consumerism entities that collectively indoctrinates us to enjoy consumerism. We don't know how efficient or effective that system is. But all we need to know is that we can improve it. How hard is it really to plant a seed, water it and eat the tomatoes? How hard is it to mix flour, yeast and water then shove it in the oven? It sure seems like an infinitely complex task to me, infinite as in: ill never do it. That apathy is great news for consumerism. Its behavior that deserves to be praised and rewarded by the entities benefiting from it. There is no evil man pulling levers behind a curtain. Its just the way things are.
UncleMeat 7y ago
What if reddit did this? Or Amazon? Or literally any popular web site. Is this reason to break up all of them?

by k_sze 7y ago

evrydayhustling 7y ago

Companies sponsor campaigns and publicly support politicians all the time. Personally I think campaign contributions should be way more limited, but I'm not sure what is special about Google here.

dantheman 7y ago

Imagine if a newspaper endorsed a candidate?

k_sze 7y ago
Except it would be more subtle. Google would not be telling you who to vote for because they already have an idea who you'll be voting for, based on machine learning results. If they think you would vote for someone disadvantageous to Google, they would not show the voting reminder.
- pvg 7y ago
  It sounds a bit like this degree of subtlety: https://i.imgur.com/OkVMmQi.jpg https://i.imgur.com/OkVMmQi.jpg

superseeplus 7y ago

Even assuming Google is theoretically capable of showing the voting reminder only to users who would vote for a Democratic candidate doesn’t mean that they did that. Dr. Epstein doesn’t present any evidence (even anecdotal) that the reminders were not shown to Republicans.

tptacek 7y ago

It's more than a big what-if: it's also not something Epstein claims has happened.

kingludite 7y ago

Exactly! This is the interesting question to ask! And you should expect this to take a much less straight forwards form than outright malicious intend. Machine learning stuff, to us mere mortals, is a black box that seems to output that what we desire. However, we don't need a single black box to completely hide the effect from everyone involved. It can be simply spread out over many processes without anyone noticing the sum of countless, by themselves very sensible, pro-google decisions made by google and its employees. Say on a fuzzy edge some human or automation has to decide if a website gets banned from the index or not. Since they use adsense or google analytics it becomes possible to contact someone running the site or even, long before the offense, present them with a machine generated report suggesting decline of quality. The google hating webmaster doesn't want all that tracking crap on his site and therefore he might be punished by not having the warning in advance. It doesn't seem harmful, it seems perfectly reasonable, if however you create 1000 such effects the meaningless 0.1% adds up to 100%. Say, we take a user profile that use to frequent sites now banned from the search index. If that profile moves to a new website. Does that merit close monitoring? If the AI is a black box you don't even get to ask the question. It all seems to be fair and logical but the sum of those 2 would boil down to close monitoring websites not using analytics. We would praise the AI for noticing it so soon and so efficiently. And so things further escalate without anyone connecting the dots. To make it more sinister: We have a system of pro-consumerism entities that collectively indoctrinates us to enjoy consumerism. We don't know how efficient or effective that system is. But all we need to know is that we can improve it. How hard is it really to plant a seed, water it and eat the tomatoes? How hard is it to mix flour, yeast and water then shove it in the oven? It sure seems like an infinitely complex task to me, infinite as in: ill never do it. That apathy is great news for consumerism. Its behavior that deserves to be praised and rewarded by the entities benefiting from it. There is no evil man pulling levers behind a curtain. Its just the way things are.

UncleMeat 7y ago

What if reddit did this? Or Amazon? Or literally any popular web site. Is this reason to break up all of them?