Google controls which political emails land in your inbox
February 29, 2020 8:33 AM Subscribe
"Pete Buttigieg is leading at 63 percent. Andrew Yang came in second at 46 percent. And Elizabeth Warren looks like she’s in trouble with 0 percent. These aren’t poll numbers for the U.S. 2020 Democratic presidential contest. Instead, they reflect which candidates were able to consistently land in Gmail’s primary inbox in a simple test."
I've donated to her multiple times and all of Warren's messages end up in the Promotions tab for me. I don't get any other messages so I can't compare to where other candidates might land.
posted by misskaz at 8:54 AM on February 29, 2020 [4 favorites]
posted by misskaz at 8:54 AM on February 29, 2020 [4 favorites]
Would letter and parcel delivery firms be allowed to get away with checking to see who the post was from and putting some in the letter box and hiding some behind the waste bin.
I thought interfering with the mail was a criminal offence.
posted by Burn_IT at 9:03 AM on February 29, 2020 [2 favorites]
I thought interfering with the mail was a criminal offence.
posted by Burn_IT at 9:03 AM on February 29, 2020 [2 favorites]
Counterpoint: Next time I donate to any political candidate, I'm setting up a burner account first. I don't need dire fundraiser warnings multiple times a day from every affiliated organization they share their e-mail lists with.
(Possibly this is because Yahoo is less competent than Google.)
posted by Huffy Puffy at 9:08 AM on February 29, 2020 [14 favorites]
(Possibly this is because Yahoo is less competent than Google.)
posted by Huffy Puffy at 9:08 AM on February 29, 2020 [14 favorites]
Ah, so that's where a certain political organizing email I specifically requested ended up.
Given that my state isn't shying away from going after big tech companies for election issues, would anyone know the right agency to send a formal complaint to?
Though the pessimistic side of me suspects this is the sort of thing where law hasn't caught up to society, and I'll be writing my state legislator, too.
posted by Zalzidrax at 9:08 AM on February 29, 2020 [1 favorite]
Given that my state isn't shying away from going after big tech companies for election issues, would anyone know the right agency to send a formal complaint to?
Though the pessimistic side of me suspects this is the sort of thing where law hasn't caught up to society, and I'll be writing my state legislator, too.
posted by Zalzidrax at 9:08 AM on February 29, 2020 [1 favorite]
I thought interfering with the mail was a criminal offence.
The end result of this line of thinking is utter chaos and madness. Email spam is almost literally zero cost, and not segregating spam means that email becomes actually literally useless.
There's a deep difference between even just a few cents per snail mail and a few cents per billion spam.
I think we can have a very good discussion of how these filters should work and now they shouldn't without handing the keys to the V|46R4 shitheads.
posted by tclark at 9:10 AM on February 29, 2020 [19 favorites]
The end result of this line of thinking is utter chaos and madness. Email spam is almost literally zero cost, and not segregating spam means that email becomes actually literally useless.
There's a deep difference between even just a few cents per snail mail and a few cents per billion spam.
I think we can have a very good discussion of how these filters should work and now they shouldn't without handing the keys to the V|46R4 shitheads.
posted by tclark at 9:10 AM on February 29, 2020 [19 favorites]
I thought interfering with the mail was a criminal offence.
Is this a serious statement?
For one thing: email is not mail. It is not regulated by the government. That it is delivered at all is purely by mutual agreement of various providers involved.
For another thing: the "interfering" is purposeful and one of gmail's primary features. Prior to gmail existing, the only way to have a remotely usable mailbox (due to literally thousands of spam mails a day) was to run your own mailserver plus whitelists/blacklists/crowdsourced blackholes/etc. I did it in the early 2000s. It was an enormous pain in the ass, but the alternative was a mailbox with a spam:real ratio of ~1000:1.
Does that mean everything gmail is doing is great? No. Does that mean it's some shadowy conspiracy or criminal act? Not remotely.
posted by tocts at 9:10 AM on February 29, 2020 [20 favorites]
Is this a serious statement?
For one thing: email is not mail. It is not regulated by the government. That it is delivered at all is purely by mutual agreement of various providers involved.
For another thing: the "interfering" is purposeful and one of gmail's primary features. Prior to gmail existing, the only way to have a remotely usable mailbox (due to literally thousands of spam mails a day) was to run your own mailserver plus whitelists/blacklists/crowdsourced blackholes/etc. I did it in the early 2000s. It was an enormous pain in the ass, but the alternative was a mailbox with a spam:real ratio of ~1000:1.
Does that mean everything gmail is doing is great? No. Does that mean it's some shadowy conspiracy or criminal act? Not remotely.
posted by tocts at 9:10 AM on February 29, 2020 [20 favorites]
I only get mail from Warren and it all goes into Promotions and I am very glad for that because they come in pretty constantly.
posted by octothorpe at 9:13 AM on February 29, 2020
posted by octothorpe at 9:13 AM on February 29, 2020
Also, it bears pointing out that one of the ways filters such that gmail uses learn is user feedback. One easy way for a lot of a given candidate's stuff to end up auto-flagged as spam is, prior emails weren't auto-flagged as spam and the people who received them manually flagged them as spam.
Which is to say a real easy way to end up auto-filtered is to send a lot of emails to people who don't want them, either because they never asked for them or because they feel the emails are not useful or overwhelming in number.
posted by tocts at 9:14 AM on February 29, 2020 [6 favorites]
Which is to say a real easy way to end up auto-filtered is to send a lot of emails to people who don't want them, either because they never asked for them or because they feel the emails are not useful or overwhelming in number.
posted by tocts at 9:14 AM on February 29, 2020 [6 favorites]
I have only two places that incoming mail can go: the primary inbox tab or the spam folder. If something lands in spam that doesn't belong there, I tell gmail. Conversely, sending inbox items to the spam folder informs gmail as well. Just like a new puppy, you have to train gmail and, eventually, it will get almost everything right. A new gmail instance is almost certain to pee on the floor...
posted by jim in austin at 9:29 AM on February 29, 2020 [6 favorites]
posted by jim in austin at 9:29 AM on February 29, 2020 [6 favorites]
Does that mean everything gmail is doing is great? No. Does that mean it's some shadowy conspiracy or criminal act? Not remotely.
Something doesn't have to be intentional or criminal to be illegal.
Complex systems have unintended consequences that even really smart people can't foresee. But once those unintended consequences are made clear, then we need to mitigate them. If we find some chemical compound poison people, we ban it from consumer goods. There's no reason not to take the same approach with information technology and democracy.
I'm guessing all those super smart people google is hiring might just find a good way to filter spam but not accidentally favor some politicians and groups over others if there are some very large fines looming over them if they don't. And if there aren't, they may very well not bother.
posted by Zalzidrax at 9:30 AM on February 29, 2020 [7 favorites]
Something doesn't have to be intentional or criminal to be illegal.
Complex systems have unintended consequences that even really smart people can't foresee. But once those unintended consequences are made clear, then we need to mitigate them. If we find some chemical compound poison people, we ban it from consumer goods. There's no reason not to take the same approach with information technology and democracy.
I'm guessing all those super smart people google is hiring might just find a good way to filter spam but not accidentally favor some politicians and groups over others if there are some very large fines looming over them if they don't. And if there aren't, they may very well not bother.
posted by Zalzidrax at 9:30 AM on February 29, 2020 [7 favorites]
Interfering with mail is criminal because mail delivery is a federal service. It has nothing to do with free speech.
posted by kevinbelt at 9:30 AM on February 29, 2020 [15 favorites]
posted by kevinbelt at 9:30 AM on February 29, 2020 [15 favorites]
I've noticed that recently more emails from the Trump campaign have been ending up in my main Inbox on Gmail (don't worry, I'm just subscribed to get access to the insane surveys his campaign sends out), when they'd been in Promotions for years. Meanwhile, no other political solicitation or info emails are ending up in the Inbox. I suppose I had never thought about the logistics of how emails get sorted, but this shit is really pushing me towards the notion of ditching Gmail completely. What a pain...
posted by wakannai at 9:34 AM on February 29, 2020 [1 favorite]
posted by wakannai at 9:34 AM on February 29, 2020 [1 favorite]
Attaining high deliverability for large email lists is similar to SEO in that it has its own knowledge base and good practices.
For example: if Warren's team is sending emails that consistently don't get opened, have low read times, low clickthrough percentages, text that's spammy or similar to other emails that don't get read etc., Gmail will assume emails from her mailservers aren't desired and will assign a greater percentage of them to spam across her entire list.
posted by Spacelegoman at 9:44 AM on February 29, 2020 [7 favorites]
For example: if Warren's team is sending emails that consistently don't get opened, have low read times, low clickthrough percentages, text that's spammy or similar to other emails that don't get read etc., Gmail will assume emails from her mailservers aren't desired and will assign a greater percentage of them to spam across her entire list.
posted by Spacelegoman at 9:44 AM on February 29, 2020 [7 favorites]
The way this stuff is purely algorithmic. Nobody's behind the scenes deciding which emails get sent where.
The algorithm is trained on a data set generated by the people that use gmail and manually categorize what types of emails get put in which folder.
For any given email, gmail asks the question "does this look like other emails that I've seen?" And then it sticks them into the same bucket.
posted by lalunamel at 9:46 AM on February 29, 2020 [4 favorites]
The algorithm is trained on a data set generated by the people that use gmail and manually categorize what types of emails get put in which folder.
For any given email, gmail asks the question "does this look like other emails that I've seen?" And then it sticks them into the same bucket.
posted by lalunamel at 9:46 AM on February 29, 2020 [4 favorites]
I suppose I had never thought about the logistics of how emails get sorted,
Are you regularly clicking links in those emails to access the surveys, such that Gmail might think those are important to you?
posted by Nonsteroidal Anti-Inflammatory Drug at 9:50 AM on February 29, 2020 [2 favorites]
Are you regularly clicking links in those emails to access the surveys, such that Gmail might think those are important to you?
posted by Nonsteroidal Anti-Inflammatory Drug at 9:50 AM on February 29, 2020 [2 favorites]
Complex systems have unintended consequences that even really smart people can't foresee. But once those unintended consequences are made clear, then we need to mitigate them.
Is it unintended? You are using an email service that's primary feature is spam filtering.
I treat my inbox as sacred, I unsubscribe or flag every single email that is not personally addressed to me by a human. So all political spam goes in the spam box, just as it should. I don't care if it's for the candidate I like, it's still SPAM.
I myself am a part of the problem, because I do a MailChimp campaign for my business. The whole point of doing email campaigns is that you can track every thing about it (who opened, who clicked, who visited my website, what they looked at and how long where they there, did it convert to sales or where did they go after? etc.). There's like a whole cottage industry, similar to SEO, on how to not get hit by Google's filter and how to get people to actually open and click. It's not like Warren wouldn't know they're getting to spam consistently - because the whole point of email spam is to generate this kind of data for the sender.
So if I, sole propreitor who hates spam yet still plays the game with the most minimal amount of effort, can get through Google's filter then surely Warren can hire someone who can do the same.
Or, if email filtering bothers you, maybe don't use an email service that filters your mail. But for me, all political spam is still spam so this system is working exactly as intended. And Google uses my choices to inform your filter. Sorry not sorry?
posted by bradbane at 9:51 AM on February 29, 2020 [3 favorites]
Is it unintended? You are using an email service that's primary feature is spam filtering.
I treat my inbox as sacred, I unsubscribe or flag every single email that is not personally addressed to me by a human. So all political spam goes in the spam box, just as it should. I don't care if it's for the candidate I like, it's still SPAM.
I myself am a part of the problem, because I do a MailChimp campaign for my business. The whole point of doing email campaigns is that you can track every thing about it (who opened, who clicked, who visited my website, what they looked at and how long where they there, did it convert to sales or where did they go after? etc.). There's like a whole cottage industry, similar to SEO, on how to not get hit by Google's filter and how to get people to actually open and click. It's not like Warren wouldn't know they're getting to spam consistently - because the whole point of email spam is to generate this kind of data for the sender.
So if I, sole propreitor who hates spam yet still plays the game with the most minimal amount of effort, can get through Google's filter then surely Warren can hire someone who can do the same.
Or, if email filtering bothers you, maybe don't use an email service that filters your mail. But for me, all political spam is still spam so this system is working exactly as intended. And Google uses my choices to inform your filter. Sorry not sorry?
posted by bradbane at 9:51 AM on February 29, 2020 [3 favorites]
I suppose I had never thought about the logistics of how emails get sorted
Bayesian probability
I'm sure Google's system is doing something more complicated than that, but the idea is a basic math problem.
posted by bradbane at 10:04 AM on February 29, 2020
Bayesian probability
I'm sure Google's system is doing something more complicated than that, but the idea is a basic math problem.
posted by bradbane at 10:04 AM on February 29, 2020
This is shoddy reporting and should never have been published. Email providers manage which emails go through on an individual basis. Setting up a single Gmail inbox won't tell you anything.
posted by Hollywood Upstairs Medical College at 10:05 AM on February 29, 2020 [5 favorites]
posted by Hollywood Upstairs Medical College at 10:05 AM on February 29, 2020 [5 favorites]
Yes, it is unintended. I am sure no one writing a spam filter has any intention of disproportionately favoring one candidate over another. However, that is apparently the outcome.
I mean I didn't think of it until just now, so I'm not going to blame anyone else for malice or stupidity, but it's clearly a major bug, and it needs fixing.
As google deployed this email sorting algorithm (without any sort of opt in that I remember), is responsible for designing it, are the only ones who have access to it, and can change it at any time, they are the ones responsible. End users providing data by customizing their settings or doing SEO voodoo or not are not really relevant to who has the ultimate responsibility.
This is just another instance of decision making algorithms amplifying bias. Just like the decision making algorithms have been shown to amplify racial bias or gender bias if not done carefully.
This is shoddy reporting and should never have been published. Email providers manage which emails go through on an individual basis. Setting up a single Gmail inbox won't tell you anything.
Sure it will! It tells you they haven't taken steps to make sure their system avoids privileging one candidate over another. Which I have stated, is perfectly understandable that no one thought to do. But once you're in a position to deploy an algorithm that can affect society at large, this sort of stuff crops up regularly. And society needs to regulate it, or it is going to break things like any other computational instability will make simulation go off the rails. Except if your algorithms are tied in to real world systems, it will break those systems, too.
posted by Zalzidrax at 10:17 AM on February 29, 2020 [8 favorites]
I mean I didn't think of it until just now, so I'm not going to blame anyone else for malice or stupidity, but it's clearly a major bug, and it needs fixing.
As google deployed this email sorting algorithm (without any sort of opt in that I remember), is responsible for designing it, are the only ones who have access to it, and can change it at any time, they are the ones responsible. End users providing data by customizing their settings or doing SEO voodoo or not are not really relevant to who has the ultimate responsibility.
This is just another instance of decision making algorithms amplifying bias. Just like the decision making algorithms have been shown to amplify racial bias or gender bias if not done carefully.
This is shoddy reporting and should never have been published. Email providers manage which emails go through on an individual basis. Setting up a single Gmail inbox won't tell you anything.
Sure it will! It tells you they haven't taken steps to make sure their system avoids privileging one candidate over another. Which I have stated, is perfectly understandable that no one thought to do. But once you're in a position to deploy an algorithm that can affect society at large, this sort of stuff crops up regularly. And society needs to regulate it, or it is going to break things like any other computational instability will make simulation go off the rails. Except if your algorithms are tied in to real world systems, it will break those systems, too.
posted by Zalzidrax at 10:17 AM on February 29, 2020 [8 favorites]
Hollywood Upstairs Medical College: Email providers manage which emails go through on an individual basis.
Note that they took pains to avoid individualization, and that's where these very disparate results were found:
posted by clawsoon at 10:25 AM on February 29, 2020 [1 favorite]
Note that they took pains to avoid individualization, and that's where these very disparate results were found:
To test how Gmail treats political email, we opened a new Gmail account using a new phone number and Tor, an anonymizing browser, to avoid sending signals about political leanings based on previous web activity.It seems that Google's algorithm is not purely individual, but is also using some "wisdom of the crowds" priors, as you might expect.
posted by clawsoon at 10:25 AM on February 29, 2020 [1 favorite]
It tells you they haven't taken steps to make sure their system avoids privileging one candidate over another.
And why exactly should political spam be privileged over other spam? It's still spam, and you chose to use a service who's main feature is spam filtering. The whole point of spam filtering is to privilege some messages over others. Like the author of the piece, you seem to not understand what is going on here at all.
It's not "amplifying bias" if the Warren campaign sucks at sending spam. It just means they suck at sending spam, to users who specifically have chosen to filter out spam.
Are you saying that political messages should be exempt from spam filtering for those who choose to use spam filtering?
This article is incoherent, and as a result so are a lot of the responses here.
posted by bradbane at 10:31 AM on February 29, 2020 [11 favorites]
And why exactly should political spam be privileged over other spam? It's still spam, and you chose to use a service who's main feature is spam filtering. The whole point of spam filtering is to privilege some messages over others. Like the author of the piece, you seem to not understand what is going on here at all.
It's not "amplifying bias" if the Warren campaign sucks at sending spam. It just means they suck at sending spam, to users who specifically have chosen to filter out spam.
Are you saying that political messages should be exempt from spam filtering for those who choose to use spam filtering?
This article is incoherent, and as a result so are a lot of the responses here.
posted by bradbane at 10:31 AM on February 29, 2020 [11 favorites]
Doesn’t Google customize its Bayesian filters for each user?
If so, this n=1 sample is basically worthless (unless you want to consider the initial conditions for a new account, which are admittedly interesting until you consider that the filters were apparently trained on a brand-new account that subscribed to every single candidate’s mailing list and nothing else).
Unless this test was substantially more rigorous than the article suggests it was, it’s junk science.
posted by schmod at 10:41 AM on February 29, 2020 [2 favorites]
If so, this n=1 sample is basically worthless (unless you want to consider the initial conditions for a new account, which are admittedly interesting until you consider that the filters were apparently trained on a brand-new account that subscribed to every single candidate’s mailing list and nothing else).
Unless this test was substantially more rigorous than the article suggests it was, it’s junk science.
posted by schmod at 10:41 AM on February 29, 2020 [2 favorites]
I note the article didn't look at the content. Is it possible that Warren isn't using proper domain signatures, or if the tool they use for HTML is generating invalid tags. Both those things will get you marked down algorithmically.
posted by CheeseDigestsAll at 10:41 AM on February 29, 2020 [1 favorite]
posted by CheeseDigestsAll at 10:41 AM on February 29, 2020 [1 favorite]
Pretty much anything from Mailchimp is going to the promotions tab, just by the fact that it came from Mailchimp, or Constant Contact, etc. I wonder if Mayor Pete is using a more obscure provider that doesn't get auto filtered like the more popular ones?
posted by COD at 10:41 AM on February 29, 2020 [3 favorites]
posted by COD at 10:41 AM on February 29, 2020 [3 favorites]
Seriously don't fuck with my spam filter 'cause your favorite politician's e-mails are too spammy.
And society needs to regulate it
This is regulation. Google, a small part of society, is regulating spam e-mails. As a person who lives in society, I find this virtuous. The politician has no right to my e-mail inbox.
posted by save alive nothing that breatheth at 10:46 AM on February 29, 2020 [4 favorites]
And society needs to regulate it
This is regulation. Google, a small part of society, is regulating spam e-mails. As a person who lives in society, I find this virtuous. The politician has no right to my e-mail inbox.
posted by save alive nothing that breatheth at 10:46 AM on February 29, 2020 [4 favorites]
Doesn’t Google customize its Bayesian filters for each user?
Exactly. If Gmail marks a Warren spam as spam, but you would actually like to see those messages, clicking the "Not Spam" button trains the filter for your individual account and messages that look like Warren messages will not get flagged as spam in the future.
Setting up a new email account and just seeing what sticks and what doesn't is meaningless. If you're getting 0% open rate on your email campaign the problem is you, not the filter.
posted by bradbane at 10:47 AM on February 29, 2020 [3 favorites]
Exactly. If Gmail marks a Warren spam as spam, but you would actually like to see those messages, clicking the "Not Spam" button trains the filter for your individual account and messages that look like Warren messages will not get flagged as spam in the future.
Setting up a new email account and just seeing what sticks and what doesn't is meaningless. If you're getting 0% open rate on your email campaign the problem is you, not the filter.
posted by bradbane at 10:47 AM on February 29, 2020 [3 favorites]
To test how Gmail treats political email, we opened a new Gmail account using a new phone number and Tor, an anonymizing browser, to avoid sending signals about political leanings based on previous web activity.
And of course nobody could possibly have any reason to assume that traffic involving a collection of Tor exit nodes might carry traffic primarily from folks who skew libertarian, or that an otherwise untrained Gmail spam filter might take that into account as a default preference for Gmails destined for those nodes. So our n=1 sampling process is all totally unbiased and scientific and shit.
posted by flabdablet at 10:48 AM on February 29, 2020 [8 favorites]
And of course nobody could possibly have any reason to assume that traffic involving a collection of Tor exit nodes might carry traffic primarily from folks who skew libertarian, or that an otherwise untrained Gmail spam filter might take that into account as a default preference for Gmails destined for those nodes. So our n=1 sampling process is all totally unbiased and scientific and shit.
posted by flabdablet at 10:48 AM on February 29, 2020 [8 favorites]
If Gmail marks a Warren spam as spam, but you would actually like to see those messages, clicking the "Not Spam" button trains the filter for your individual account and messages that look like Warren messages will not get flagged as spam in the future.
But the issue here isn’t really the spam filter. It’s the “Promotions” tab (and I think there other tabs too) that some people have turned on in Gmail to sort different categories of email. I don’t really find this to be a useful feature, so I turned it off after trying it for a few days when they introduced it. (I just unsubscribe from lists I don’t want to see regularly — like campaign emails, so I’m really not the target audience for any of this controversy.) But I think Gmail has this turned on by default these days, so for most people, these emails they might want to see are being diverted from their main inbox that they look at every day.
posted by stopgap at 10:53 AM on February 29, 2020 [3 favorites]
But the issue here isn’t really the spam filter. It’s the “Promotions” tab (and I think there other tabs too) that some people have turned on in Gmail to sort different categories of email. I don’t really find this to be a useful feature, so I turned it off after trying it for a few days when they introduced it. (I just unsubscribe from lists I don’t want to see regularly — like campaign emails, so I’m really not the target audience for any of this controversy.) But I think Gmail has this turned on by default these days, so for most people, these emails they might want to see are being diverted from their main inbox that they look at every day.
posted by stopgap at 10:53 AM on February 29, 2020 [3 favorites]
As google deployed this email sorting algorithm (without any sort of opt in that I remember)
The opt in is, you're using gmail, a free email platform that has specific features that apply only to it. If you don't want your mail sorted in the ways gmail advertises itself as sorting them, don't use gmail.
Add to this: when it comes to "Promotions", that is 100% optional. I have never even seen that thing, because it wasn't turned on for existing accounts as far as I know when they rolled it out, and any new account can turn it off. Having never even seen it, I was able to find it in about 5 seconds. The text in the UI (requiring 2 clicks - on the "settings" icon, then "configure inbox") literally says:
This entire article is clickbait designed to spark outrage.
posted by tocts at 11:12 AM on February 29, 2020 [7 favorites]
The opt in is, you're using gmail, a free email platform that has specific features that apply only to it. If you don't want your mail sorted in the ways gmail advertises itself as sorting them, don't use gmail.
Add to this: when it comes to "Promotions", that is 100% optional. I have never even seen that thing, because it wasn't turned on for existing accounts as far as I know when they rolled it out, and any new account can turn it off. Having never even seen it, I was able to find it in about 5 seconds. The text in the UI (requiring 2 clicks - on the "settings" icon, then "configure inbox") literally says:
Choose which message categories to show as inbox tabs. Other messages will appear in the Primary tab."Promotions" is one of those options. Want those in your primary inbox? You can do that.
This entire article is clickbait designed to spark outrage.
posted by tocts at 11:12 AM on February 29, 2020 [7 favorites]
I don't use the "promotions" thing either, but the fact remains that if you do not want Google to filter or divert your email, you should not have signed up for an email service who's primary selling point is that it filters your email. Or maybe you should learn how to hit that 'Not Spam' button if you're getting false positives.
No politician has a right to have their email get past my spam filter, I don't care who the politician is, I don't want any email from them period. People here are saying political spam is some special class of spam - sorry no thank you, that is some straight bullshit.
Wonder how this thread would go if it were the Trump campaign arguing that the biased dishonest liberal tech companies are using ~algorithms~ to censor conservative email messages and campaigns, and that they therefore deserve to bypass everyone's spam filters to reach voters. If political spam is somehow special, and any kind of filtering is "bias", where you going to draw the line? You want daily email updates in your inbox from Dear Leader himself and nothing you can do about it?
If you don't like spam filters, may I humbly suggest you not use them.
posted by bradbane at 11:16 AM on February 29, 2020 [4 favorites]
No politician has a right to have their email get past my spam filter, I don't care who the politician is, I don't want any email from them period. People here are saying political spam is some special class of spam - sorry no thank you, that is some straight bullshit.
Wonder how this thread would go if it were the Trump campaign arguing that the biased dishonest liberal tech companies are using ~algorithms~ to censor conservative email messages and campaigns, and that they therefore deserve to bypass everyone's spam filters to reach voters. If political spam is somehow special, and any kind of filtering is "bias", where you going to draw the line? You want daily email updates in your inbox from Dear Leader himself and nothing you can do about it?
If you don't like spam filters, may I humbly suggest you not use them.
posted by bradbane at 11:16 AM on February 29, 2020 [4 favorites]
→ No politician has a right to have their email get past my spam filter
bradbane, you seem angry about this. You've used the term “spam”. These e-mails are not spam. Paragraph 4, sentence 1 of TFA begins:
posted by scruss at 11:42 AM on February 29, 2020 [8 favorites]
bradbane, you seem angry about this. You've used the term “spam”. These e-mails are not spam. Paragraph 4, sentence 1 of TFA begins:
We found that few of the emails we’d signed up to receive …These are opt-in e-mails that people have explicitly requested. They want to see them, but they are not.
posted by scruss at 11:42 AM on February 29, 2020 [8 favorites]
I think I see some confusion here about what spam is and what spam filters do.
Like tocts, I ran a server back in the day, until it became too big a PITA.
For an email to be called “spam”, 2 things must be true:
The days when using a spam filter were optional are LONG gone. The amount of spam email far exceeded non-spam email years ago. If spam filters weren’t used, email would be completely unusable.
That said, using similar techniques to sort non-spam email, especially ones that users have little control over, is pretty sketchy in my view.
I don’t really think that the mail sorting is intentional or malicious, but that doesn’t mean it’s not due to crappy programming or unintended consequences.
One more thing - how about we don’t blame users for using Google mail, or Yahoo, or whatever? We aren’t all super-genius techies, and we’re going to use the most convenient, reliable and widely used system. Google is effectively a monopoly at this point, and for a lot of folks, there isn’t an obvious alternative.
posted by shorstenbach at 11:58 AM on February 29, 2020 [6 favorites]
Like tocts, I ran a server back in the day, until it became too big a PITA.
For an email to be called “spam”, 2 things must be true:
- It’s unsolicited - you didn’t ask for it. You might “ask for it” by either explicitly subscribing to a mailing list or by giving the sender your address and not clicking the “opt-out” checkbox (assuming there is one).
- It’s bulk. A person sending individual emails manually without using a mailing list doesn’t qualify.
The days when using a spam filter were optional are LONG gone. The amount of spam email far exceeded non-spam email years ago. If spam filters weren’t used, email would be completely unusable.
That said, using similar techniques to sort non-spam email, especially ones that users have little control over, is pretty sketchy in my view.
I don’t really think that the mail sorting is intentional or malicious, but that doesn’t mean it’s not due to crappy programming or unintended consequences.
One more thing - how about we don’t blame users for using Google mail, or Yahoo, or whatever? We aren’t all super-genius techies, and we’re going to use the most convenient, reliable and widely used system. Google is effectively a monopoly at this point, and for a lot of folks, there isn’t an obvious alternative.
posted by shorstenbach at 11:58 AM on February 29, 2020 [6 favorites]
These are opt-in e-mails that people have explicitly requested. They want to see them, but they are not.
Even this is really tricky, because there’s no way for your email client to know what emails you’ve signed up for and which are unsolicited. Not to mention the number of users who click the “spam” button on emails they did sign up for, either because they forgot or because a donation form or something tricked them into signing up.
And most political email lists look marketing email to a spam filter because they are marketing email. Albeit marketing the user might actually want to see.
I do think that this calls for regulation, though I don’t think that the government should be managing your email filters.
Even this is really tricky, because there’s no way for your email client to know what emails you’ve signed up for and which are unsolicited. Not to mention the number of users who click the “spam” button on emails they did sign up for, either because they forgot or because a donation form or something tricked them into signing up.
And most political email lists look marketing email to a spam filter because they are marketing email. Albeit marketing the user might actually want to see.
I do think that this calls for regulation, though I don’t think that the government should be managing your email filters.
- The data from this article is so bad because there’s no way to get real data on this, unless you’re actually Google. That’s bad, because while I don’t think Google is interfering with email delivery to suit their political ends, they very well could and it would be hard to detect! I think that, similar to social media companies, Google and other email providers need to start signing privacy agreements with academic researchers to provide anonymized and aggregate data on these kinds of stats so we can get independent studies of this behavior.
- Even better — I’d love to see political advertising (including political marketing emails) regulated to begin with, with penalties on campaigns for bad behavior, rather than asking platforms to effectively do the regulation themselves.
I don't see why the standard structural bias argument doesn't apply here?
Calling members' explanations here incoherent contributes to dismissal of political inquiry. It's awfully convenient as an implicitly politically motivated tactic.
posted by polymodus at 12:19 PM on February 29, 2020 [2 favorites]
Calling members' explanations here incoherent contributes to dismissal of political inquiry. It's awfully convenient as an implicitly politically motivated tactic.
posted by polymodus at 12:19 PM on February 29, 2020 [2 favorites]
These are opt-in e-mails that people have explicitly requested.
There's literally no way for your email server to know this, other than you going through the effort of telling it. It certainly can't trust the sender to inform it that you totally said the email was OK, because at that point all spammers will simply say you said it was OK.
It is worth pointing out that there's two separate things going on here:
First, something like half of what's being reported as having been filtered away is simply put on a second, 100% optional tab in the user's inbox. Any user who doesn't want this can turn it off in 2 clicks. Even if they don't turn it off, it's not gone, it's on a tab that's one click away.
Second, regarding the remainder, if the user signed up for it and didn't get it, they should be looking at their spam folder and giving feedback to gmail that they don't view this as spam. That's the only way it can tell, because otherwise these messages tend to look just like spam.
Spam filtering isn't perfect. It also isn't biased against a particular candidate, excepting insofar as much as that candidate is bad at sending mails that don't look like spam. If one candidate is doing better at this than others, it's because the people sending their mails for them are better at what they do, because sending solicited mail that avoids looking like unsolicited mail is not a simple matter.
how about we don’t blame users for using Google mail
I'm not blaming users for using gmail, but I am blaming users for complaining about literally the defining feature of it. The entire reason gmail rose to prominence was spam filtering. Talking about that feature as if it is a nefarious attempt by google to suppress certain candidates is ludicrous.
posted by tocts at 12:20 PM on February 29, 2020 [8 favorites]
There's literally no way for your email server to know this, other than you going through the effort of telling it. It certainly can't trust the sender to inform it that you totally said the email was OK, because at that point all spammers will simply say you said it was OK.
It is worth pointing out that there's two separate things going on here:
First, something like half of what's being reported as having been filtered away is simply put on a second, 100% optional tab in the user's inbox. Any user who doesn't want this can turn it off in 2 clicks. Even if they don't turn it off, it's not gone, it's on a tab that's one click away.
Second, regarding the remainder, if the user signed up for it and didn't get it, they should be looking at their spam folder and giving feedback to gmail that they don't view this as spam. That's the only way it can tell, because otherwise these messages tend to look just like spam.
Spam filtering isn't perfect. It also isn't biased against a particular candidate, excepting insofar as much as that candidate is bad at sending mails that don't look like spam. If one candidate is doing better at this than others, it's because the people sending their mails for them are better at what they do, because sending solicited mail that avoids looking like unsolicited mail is not a simple matter.
how about we don’t blame users for using Google mail
I'm not blaming users for using gmail, but I am blaming users for complaining about literally the defining feature of it. The entire reason gmail rose to prominence was spam filtering. Talking about that feature as if it is a nefarious attempt by google to suppress certain candidates is ludicrous.
posted by tocts at 12:20 PM on February 29, 2020 [8 favorites]
The entire reason gmail rose to prominence was spam filtering
I don't think this is true. I pretty certain it was the free storage. Never having to delete an email! Wow!
posted by howfar at 12:23 PM on February 29, 2020 [3 favorites]
I don't think this is true. I pretty certain it was the free storage. Never having to delete an email! Wow!
posted by howfar at 12:23 PM on February 29, 2020 [3 favorites]
If one candidate is doing better at this than others, it's because the people sending their mails for them are better at what they do, because sending solicited mail that avoids looking like unsolicited mail is not a simple matter.
But this is where a leftist or social justice progressive would deconstruct the meaning of "better" to show that e.g. having funds to pay unscrupulous advert, marketing, and/or SEO type professionals to get through spam filters is precisely an example of politically gaming a system to extract advantages and so forth, and then asserting that Gmail is free to be laissez faire about this is not political...
posted by polymodus at 12:28 PM on February 29, 2020 [3 favorites]
But this is where a leftist or social justice progressive would deconstruct the meaning of "better" to show that e.g. having funds to pay unscrupulous advert, marketing, and/or SEO type professionals to get through spam filters is precisely an example of politically gaming a system to extract advantages and so forth, and then asserting that Gmail is free to be laissez faire about this is not political...
posted by polymodus at 12:28 PM on February 29, 2020 [3 favorites]
Continuing to breathe is a political act, a choice with political consequences. Ergo, saying something is “political” is a vacuous.
posted by aramaic at 12:34 PM on February 29, 2020 [4 favorites]
posted by aramaic at 12:34 PM on February 29, 2020 [4 favorites]
So now the argument is that Pete Buttigieg is uniquely more well-funded than Bernie Sanders, Elizabeth Warren, or Joe Biden, and thus can afford to hire competent staff?
posted by tocts at 12:40 PM on February 29, 2020 [3 favorites]
posted by tocts at 12:40 PM on February 29, 2020 [3 favorites]
→ There's literally no way for your email server to know this …
and that's a failing of the technology, not the users at either end of the email chain. Seeing the way that everyone had to accept DKIM when the mail services insisted on it, some form of consent-based token registration with legit bulk mailers isn't impossible. But that smacks of self-regulation and so is something Google would never do.
The article doesn't suggest this, but it's entirely possible to pay a few hundred people to use their gmail accounts to sign up for e-mails from candidates they don't like then flag them as spam to poison the filter. It would probably be cheaper to befriend a sysop of a scummy RBL just to list the politician's mailserver for a day or so, though.
posted by scruss at 1:43 PM on February 29, 2020 [1 favorite]
and that's a failing of the technology, not the users at either end of the email chain. Seeing the way that everyone had to accept DKIM when the mail services insisted on it, some form of consent-based token registration with legit bulk mailers isn't impossible. But that smacks of self-regulation and so is something Google would never do.
The article doesn't suggest this, but it's entirely possible to pay a few hundred people to use their gmail accounts to sign up for e-mails from candidates they don't like then flag them as spam to poison the filter. It would probably be cheaper to befriend a sysop of a scummy RBL just to list the politician's mailserver for a day or so, though.
posted by scruss at 1:43 PM on February 29, 2020 [1 favorite]
So now the argument is that Pete Buttigieg is uniquely more well-funded than Bernie Sanders, Elizabeth Warren, or Joe Biden, and thus can afford to hire competent staff?
tocts, It's fine if you aren't interested in critically evaluating the value-ladenness of words such as "competency" and "better"; but, I'm just pointing out that's a very neoliberal frame of mind.
posted by polymodus at 1:50 PM on February 29, 2020 [2 favorites]
tocts, It's fine if you aren't interested in critically evaluating the value-ladenness of words such as "competency" and "better"; but, I'm just pointing out that's a very neoliberal frame of mind.
posted by polymodus at 1:50 PM on February 29, 2020 [2 favorites]
It really pisses me off that these writers decided to take an interesting question of how to make spam filtering transparent and useful, and frame it with a total bullshit clickbait narrative, to try to make sure nobody could think clearly about it. I wish I could report their article as spam to punish them.
posted by value of information at 2:27 PM on February 29, 2020 [1 favorite]
posted by value of information at 2:27 PM on February 29, 2020 [1 favorite]
It seems like this discussion divides pretty cleanly between people who think political emails should have some level of special treatment and people who have donated money to a campaign at least once ever and now get spam from every conceivable race from president to dog catcher in any part of the country no matter how far.
posted by snofoam at 2:47 PM on February 29, 2020 [2 favorites]
posted by snofoam at 2:47 PM on February 29, 2020 [2 favorites]
It's fine if you aren't interested in critically evaluating the value-ladenness of words
Condescending much? If you want to dive into a deep and pointless theoretical side discussion, have at it, but maybe actually use words to say what you think is so important instead of vaguely tut-tutting about others not being leftist enough or "interested" in The Discourse.
Meanwhile, the actual topic of this post is a scaremongering article that is insinuating that in the system we currently have, a system in which literally every major campaign has paid staff to handle what is effectively marketing, that there's some nefarious algorithmic bias towards one candidate or another, as opposed to the reality which is that you can do a good and bad job at marketing.
Is the system we have the best one possible? Not by a long shot. But, nobody at Google is putting their thumb on the scale for Buttigieg, and the author of this article shows basically zero understanding in how email filtering actually works (because knowing anything would be counter to the point of clickbait anyways).
posted by tocts at 3:01 PM on February 29, 2020 [13 favorites]
Condescending much? If you want to dive into a deep and pointless theoretical side discussion, have at it, but maybe actually use words to say what you think is so important instead of vaguely tut-tutting about others not being leftist enough or "interested" in The Discourse.
Meanwhile, the actual topic of this post is a scaremongering article that is insinuating that in the system we currently have, a system in which literally every major campaign has paid staff to handle what is effectively marketing, that there's some nefarious algorithmic bias towards one candidate or another, as opposed to the reality which is that you can do a good and bad job at marketing.
Is the system we have the best one possible? Not by a long shot. But, nobody at Google is putting their thumb on the scale for Buttigieg, and the author of this article shows basically zero understanding in how email filtering actually works (because knowing anything would be counter to the point of clickbait anyways).
posted by tocts at 3:01 PM on February 29, 2020 [13 favorites]
I'm fairly well entrenched in the burn-all-political-spam-to-the-ground camp, snofoam — Canada has slightly toothy spam laws but of course all political e-mails are exempt because Canadians are just so precious.
posted by scruss at 3:22 PM on February 29, 2020
posted by scruss at 3:22 PM on February 29, 2020
I like the promotions tab and it pretty much works as intended for me - every day or so, I scroll through and click on a couple of things I find interesting, then delete everything in the tab and move on.
What I found weird was that even though I had clicked on multiple emails from Julián Castro, and I have never clicked on an email from Spotify, several times gmail prompted me to unsubscribe from emails from Maya Rupert, Castro's campaign manager. "You've never opened an email from this sender." It's the only time I've ever received a prompt like that, but there are a lot of senders I've never clicked on. Meanwhile, spotify was making it to the inbox with their 99c deals and songs I might like.
I'm sure there is a wonderful algorithmic explanation for it but that kind of stuff gets creepy.
posted by Emmy Rae at 3:22 PM on February 29, 2020
What I found weird was that even though I had clicked on multiple emails from Julián Castro, and I have never clicked on an email from Spotify, several times gmail prompted me to unsubscribe from emails from Maya Rupert, Castro's campaign manager. "You've never opened an email from this sender." It's the only time I've ever received a prompt like that, but there are a lot of senders I've never clicked on. Meanwhile, spotify was making it to the inbox with their 99c deals and songs I might like.
I'm sure there is a wonderful algorithmic explanation for it but that kind of stuff gets creepy.
posted by Emmy Rae at 3:22 PM on February 29, 2020
But I think Gmail has this turned on by default these days, so for most people, these emails they might want to see are being diverted from their main inbox that they look at every day.
The article said there's no clear data on this but that 1. they're enabled by default and 2. 34% of users in a 2016 survey were still using them. Which is actually sort of incredible to me, the idea that a majority of people (possibly) opted out pf a default setting. I, too, wish they'd done some content analysis here. I lol'd at Bernie's people saying "we are not concerned about our ability to communicate with supporters"
posted by jessamyn at 5:55 PM on February 29, 2020
The article said there's no clear data on this but that 1. they're enabled by default and 2. 34% of users in a 2016 survey were still using them. Which is actually sort of incredible to me, the idea that a majority of people (possibly) opted out pf a default setting. I, too, wish they'd done some content analysis here. I lol'd at Bernie's people saying "we are not concerned about our ability to communicate with supporters"
posted by jessamyn at 5:55 PM on February 29, 2020
These are opt-in e-mails that people have explicitly requested. They want to see them, but they are not.
How would Google (or any email provider) know this - are they supposed to read your mind? Email doesn't work that way and wishing it was so doesn't change it. It's an ancient federated system built on trust and Google (or anyone else running an email server, which I have done in the past) has no way of knowing what messages are real and trustworthy or not. That was not a consideration way back in the day when they invented it. Hence the need for spam filtering - it does it's best guess based on known spam messages vs. unknown new messages.
You can certainly argue that email is outdated and we should all switch to something better - and I'd agree - but that system wouldn't be email.
that smacks of self-regulation and so is something Google would never do.
You seem to be under the impression Google owns or controls the email protocol. They don't, which is a good thing. Anyone can set up an email server, I did it for a long time (before the hassle of dealing with spam became too great). Google has no authority to start instituting some sort of spam verification system any more than I do, and the way email is structured currently there is no way to implement this on a global level. You would have to replace email entirely - which you can already do! Go switch to Whatsapp or Telegram or whatever it is you prefer that isn't email.
and that's a failing of the technology, not the users at either end of the email chain.
No it's a failure of the OP, which is so ill informed and draws such wild conclusions from an obviously flawed understanding of both the technology and 'data' they collected that it's basically just scare mongering clickbait bullshit. And lots of people in this thread who don't really understand how email works are getting all worked up over nothing.
posted by bradbane at 7:22 PM on February 29, 2020 [7 favorites]
How would Google (or any email provider) know this - are they supposed to read your mind? Email doesn't work that way and wishing it was so doesn't change it. It's an ancient federated system built on trust and Google (or anyone else running an email server, which I have done in the past) has no way of knowing what messages are real and trustworthy or not. That was not a consideration way back in the day when they invented it. Hence the need for spam filtering - it does it's best guess based on known spam messages vs. unknown new messages.
You can certainly argue that email is outdated and we should all switch to something better - and I'd agree - but that system wouldn't be email.
that smacks of self-regulation and so is something Google would never do.
You seem to be under the impression Google owns or controls the email protocol. They don't, which is a good thing. Anyone can set up an email server, I did it for a long time (before the hassle of dealing with spam became too great). Google has no authority to start instituting some sort of spam verification system any more than I do, and the way email is structured currently there is no way to implement this on a global level. You would have to replace email entirely - which you can already do! Go switch to Whatsapp or Telegram or whatever it is you prefer that isn't email.
and that's a failing of the technology, not the users at either end of the email chain.
No it's a failure of the OP, which is so ill informed and draws such wild conclusions from an obviously flawed understanding of both the technology and 'data' they collected that it's basically just scare mongering clickbait bullshit. And lots of people in this thread who don't really understand how email works are getting all worked up over nothing.
posted by bradbane at 7:22 PM on February 29, 2020 [7 favorites]
Seeing the way that everyone had to accept DKIM when the mail services insisted on it, some form of consent-based token registration with legit bulk mailers isn't impossible.
You may want to revisit your understanding of DKIM.
posted by aramaic at 7:41 PM on February 29, 2020 [1 favorite]
You may want to revisit your understanding of DKIM.
posted by aramaic at 7:41 PM on February 29, 2020 [1 favorite]
The law, in its majestic equality, affords to Google and Joe's Computer Shack alike the right to filter emails directed to others while providing zero transparency whatsoever.
posted by Not A Thing at 8:10 PM on February 29, 2020 [1 favorite]
posted by Not A Thing at 8:10 PM on February 29, 2020 [1 favorite]
while providing zero transparency whatsoever.
Again, you are perfectly within your rights to create and run your own mailserver. There are, in fact, innumerable tools developed specifically to help you do that. Got a problem with "transparency" from commercial providers?
Then run your own. I do.
...but be warned, it's not as fun as getting worked up over nothing. It's actually a huge pain in the ass, and I get roughly 1000x more spam on my own mailservers than I do on my GMail accounts. Y'know, what with GMail being better at filtering spam than all of the Bayesian crap I've got set up.
posted by aramaic at 8:22 PM on February 29, 2020 [2 favorites]
Again, you are perfectly within your rights to create and run your own mailserver. There are, in fact, innumerable tools developed specifically to help you do that. Got a problem with "transparency" from commercial providers?
Then run your own. I do.
...but be warned, it's not as fun as getting worked up over nothing. It's actually a huge pain in the ass, and I get roughly 1000x more spam on my own mailservers than I do on my GMail accounts. Y'know, what with GMail being better at filtering spam than all of the Bayesian crap I've got set up.
posted by aramaic at 8:22 PM on February 29, 2020 [2 favorites]
Feel free to run your own email server and do no filtering. All of the software is free and open source and you can run it on your very own personal computer. Again, no one controls email, it's a federated system of independent servers. You can be a part of that network too.
But even diehard privacy nerds [generally!] don't do this in 2020, and if you want to find out why and learn something new, I suggest you try it!
posted by bradbane at 8:25 PM on February 29, 2020
But even diehard privacy nerds [generally!] don't do this in 2020, and if you want to find out why and learn something new, I suggest you try it!
posted by bradbane at 8:25 PM on February 29, 2020
This is one of the strangest MetaFilter threads I've seen in some time. I have no idea how all these people spent 54 comments talking directly past each other, or why.
posted by Not A Thing at 8:53 PM on February 29, 2020 [3 favorites]
posted by Not A Thing at 8:53 PM on February 29, 2020 [3 favorites]
Well, that's sorta what happens when people decide to be outraged by something that doesn't actually exist.
posted by aramaic at 8:58 PM on February 29, 2020 [6 favorites]
posted by aramaic at 8:58 PM on February 29, 2020 [6 favorites]
Hm. Well, to me the dynamic in this thread seems approximately as follows:
A: Maybe it hasn't been the greatest idea to give a giant unaccountable corporation using an opaque algorithm de facto control over essential political communication infrastructure.
B: What ill-informed nonsense! They're simply a giant unaccountable corporation using an opaque algorithm to exercise de facto control over essential political communication infrastructure! Any of us could do the same [if we had hundreds of millions of subscribers]!
(repeat until end of thread is reached)
posted by Not A Thing at 9:07 PM on February 29, 2020 [5 favorites]
A: Maybe it hasn't been the greatest idea to give a giant unaccountable corporation using an opaque algorithm de facto control over essential political communication infrastructure.
B: What ill-informed nonsense! They're simply a giant unaccountable corporation using an opaque algorithm to exercise de facto control over essential political communication infrastructure! Any of us could do the same [if we had hundreds of millions of subscribers]!
(repeat until end of thread is reached)
posted by Not A Thing at 9:07 PM on February 29, 2020 [5 favorites]
C: Maybe email is not the best tool for political outreach, suffering as it does from decades of technical debt and near-universal abuse by bad actors; maybe viewing email as essential political communication infrastructure is an idea well past its use-by date; and maybe ill-informed clickbait thinly disguised as investigative journalism is not the best foundation for a healthy discussion of that and related issues.
posted by flabdablet at 9:26 PM on February 29, 2020 [8 favorites]
posted by flabdablet at 9:26 PM on February 29, 2020 [8 favorites]
How would Google (or any email provider) know this - are they supposed to read your mind?
No. But they're not supposed to read my emails either.
posted by Cardinal Fang at 11:18 PM on February 29, 2020 [1 favorite]
No. But they're not supposed to read my emails either.
posted by Cardinal Fang at 11:18 PM on February 29, 2020 [1 favorite]
But, nobody at Google is putting their thumb on the scale for Buttigieg
No, but what they have done is left weights scattered around willy-nilly because there was no one to tell them not to. And it is possible, and I would say likely, that they have a systemic impact on political discourse. That this bias isn't nefariously chosen explicitly by a human, but rather what might as well be at random by an idiot machine is not comforting.
Regardless of any hyperbole or misunderstandings in the article, "you didn't take X into account when training this algorithm, did you?" is a claim that takes hardly any evidence to actually show; it is the sort of thing that, when pointed out, is trivially a problem.
Computers only do what you tell them. Machine learning algorithms only judge their output by the mathematical constraints you impose on them. They do not take into account anything else that would be common sense knowledge to humans. If you want them to do something, in this case not privilege one political party or candidates communication over another, you need to explicitly add constraints when you are training the algorithm to ensure that it happens.
Adding constraints like this is not some impossible task. I have no illusions about it being an easy task, but it is not a categorically different problem than the filtering gmail's spam filter manages anyway. But Google needs an incentive to put this problem in front of their engineers so they can figure out what constraints do lead to a fairly unbiased treatment of political communication. And the standard, boring way to get them to do that is government regulation.
Fair handling of political communication is a good thing.
An algorithm that, for better or worse, controls what a large number of people see does not ensure political communication is handled fairly.
It should do that.
Throwing up one's hands and saying "Oh well, every real political campaign has hired someone to tell them that their sheep pictures need to be on grass so they get past the filter" or the like is madness to me. It is letting tools control you rather than the other way around. Algorithms like gmail's spam filter are tools built by humans. They aren't forces of nature. We can build better ones. We can build ones that do what they should more exactly. But that's not going to happen if it's never made anyone's responsibility.
posted by Zalzidrax at 1:05 AM on March 1, 2020 [4 favorites]
No, but what they have done is left weights scattered around willy-nilly because there was no one to tell them not to. And it is possible, and I would say likely, that they have a systemic impact on political discourse. That this bias isn't nefariously chosen explicitly by a human, but rather what might as well be at random by an idiot machine is not comforting.
Regardless of any hyperbole or misunderstandings in the article, "you didn't take X into account when training this algorithm, did you?" is a claim that takes hardly any evidence to actually show; it is the sort of thing that, when pointed out, is trivially a problem.
Computers only do what you tell them. Machine learning algorithms only judge their output by the mathematical constraints you impose on them. They do not take into account anything else that would be common sense knowledge to humans. If you want them to do something, in this case not privilege one political party or candidates communication over another, you need to explicitly add constraints when you are training the algorithm to ensure that it happens.
Adding constraints like this is not some impossible task. I have no illusions about it being an easy task, but it is not a categorically different problem than the filtering gmail's spam filter manages anyway. But Google needs an incentive to put this problem in front of their engineers so they can figure out what constraints do lead to a fairly unbiased treatment of political communication. And the standard, boring way to get them to do that is government regulation.
Fair handling of political communication is a good thing.
An algorithm that, for better or worse, controls what a large number of people see does not ensure political communication is handled fairly.
It should do that.
Throwing up one's hands and saying "Oh well, every real political campaign has hired someone to tell them that their sheep pictures need to be on grass so they get past the filter" or the like is madness to me. It is letting tools control you rather than the other way around. Algorithms like gmail's spam filter are tools built by humans. They aren't forces of nature. We can build better ones. We can build ones that do what they should more exactly. But that's not going to happen if it's never made anyone's responsibility.
posted by Zalzidrax at 1:05 AM on March 1, 2020 [4 favorites]
they're not supposed to read my emails either
Since the very first implementation of email, it's been designed as more akin to a postcard delivery system than a private correspondence delivery system. Unless you're running end-to-end encryption, a complete and readable copy of every mail you ever exchange is going to exist for some while on every SMTP server on the delivery path between you and your correspondents. Email has always worked that way. Expecting email to remain unread by anybody other than its intended recipient is just not reasonable, ubiquitous legaloid disclaimers notwithstanding.
If you want private digital communication, use end-to-end encryption along with sound management of your encryption keys. If you don't want to do that, complaining about intermediaries reading your mails is pretty pointless. Especially if your email service of choice is one that explicitly makes a point of analyzing your messaging for the purpose of targeting advertising.
All of that said, handing out advice to individuals about how best to manage their communication reach and privacy doesn't address the legitimate issue raised (however hamfistedly) in TFA. Email is being used for political outreach because it's what people have, and there are many many ways for email delivery to fail.
However, I disagree that understanding a medium well enough to optimize the delivery of content through it amounts to "letting tools control you". It's not the communication tools per se that control what one-to-many senders need to do to be effective, it's the aggregate preferences of their intended recipients for tool selection and tool use.
Algorithms like gmail's spam filter are tools built by humans. They aren't forces of nature.
Indeed they're not, but treating them as if they were is more than likely the least-effort path for any individual organization to bend those tools to their own will. It's going to be a lot less costly to set up a more scientific version of TFA's delivery reach test and then optimize content against that than it would be to design and enforce the kind of regulation that would motivate a FAANG to do it server-side.
posted by flabdablet at 3:49 AM on March 1, 2020 [3 favorites]
Since the very first implementation of email, it's been designed as more akin to a postcard delivery system than a private correspondence delivery system. Unless you're running end-to-end encryption, a complete and readable copy of every mail you ever exchange is going to exist for some while on every SMTP server on the delivery path between you and your correspondents. Email has always worked that way. Expecting email to remain unread by anybody other than its intended recipient is just not reasonable, ubiquitous legaloid disclaimers notwithstanding.
If you want private digital communication, use end-to-end encryption along with sound management of your encryption keys. If you don't want to do that, complaining about intermediaries reading your mails is pretty pointless. Especially if your email service of choice is one that explicitly makes a point of analyzing your messaging for the purpose of targeting advertising.
All of that said, handing out advice to individuals about how best to manage their communication reach and privacy doesn't address the legitimate issue raised (however hamfistedly) in TFA. Email is being used for political outreach because it's what people have, and there are many many ways for email delivery to fail.
However, I disagree that understanding a medium well enough to optimize the delivery of content through it amounts to "letting tools control you". It's not the communication tools per se that control what one-to-many senders need to do to be effective, it's the aggregate preferences of their intended recipients for tool selection and tool use.
Algorithms like gmail's spam filter are tools built by humans. They aren't forces of nature.
Indeed they're not, but treating them as if they were is more than likely the least-effort path for any individual organization to bend those tools to their own will. It's going to be a lot less costly to set up a more scientific version of TFA's delivery reach test and then optimize content against that than it would be to design and enforce the kind of regulation that would motivate a FAANG to do it server-side.
posted by flabdablet at 3:49 AM on March 1, 2020 [3 favorites]
I guess some of us have spent what we feel is too much of our lives listening to people rant about how their marketing emails are the most critical thing evar and everybody else needs to change, not their 14 messages a day that seem almost intentionally crafted to trigger spam filters.
posted by wierdo at 3:57 AM on March 1, 2020 [2 favorites]
posted by wierdo at 3:57 AM on March 1, 2020 [2 favorites]
How would Google (or any email provider) know this - are they supposed to read your mind?
No. But they're not supposed to read my emails either.
so it's been a looooong time since gmail first came out, but IIRC the whole deal way back then (20 years ago now?) was that you get a FREE 1GB email storage with intelligent spam filtering and IN EXCHANGE you agree to the terms of service which state that google's algorithms will be reading your email so that they can more effectively target you with advertising?
So I'm just curious where this idea came from that they're not supposed to read your emails?
posted by some loser at 6:22 AM on March 1, 2020 [4 favorites]
Yeah, the whole point of Gmail from the beginning is that they read your emails.
posted by octothorpe at 6:31 AM on March 1, 2020 [3 favorites]
posted by octothorpe at 6:31 AM on March 1, 2020 [3 favorites]
Ugh... I know it's being hotly debated in an adjoining neighborhood, but isn't this OP yet another dubious conspiracy theory promotion?
My favorite politicians are being secretly marginalized by my least favorite big tech corporation. That warm feeling of prior beliefs being reinforced is sweet indeed. And it does follow a certain logic. Look at the first comment out the gate. The one with the most likes.
But wait, this is entirely explained by email spam filtering. But... there are more nuanced arguments to make! Sadly, look again at the first comment out the gate. The one with the most likes. Nuance, come back another day.
This is one of those rare threads I'm thankful for, where there are plenty of people who not only have an understanding of how the subject of the topic works, but some have direct experience and knowledge.
posted by 2N2222 at 9:15 AM on March 1, 2020 [4 favorites]
My favorite politicians are being secretly marginalized by my least favorite big tech corporation. That warm feeling of prior beliefs being reinforced is sweet indeed. And it does follow a certain logic. Look at the first comment out the gate. The one with the most likes.
But wait, this is entirely explained by email spam filtering. But... there are more nuanced arguments to make! Sadly, look again at the first comment out the gate. The one with the most likes. Nuance, come back another day.
This is one of those rare threads I'm thankful for, where there are plenty of people who not only have an understanding of how the subject of the topic works, but some have direct experience and knowledge.
posted by 2N2222 at 9:15 AM on March 1, 2020 [4 favorites]
Oh I just assumed this thread was blessed by the mods as a place for conspiracy venting
posted by some loser at 9:28 AM on March 1, 2020 [1 favorite]
posted by some loser at 9:28 AM on March 1, 2020 [1 favorite]
But wait, this is entirely explained by email spam filtering
It's even more absurd than that.
The entire premise of the linked article is that something being flagged as a promotion is equivalent to it being flagged as spam. This is a completely ludicrous premise.
When gmail decides something is spam, it puts it in a totally different folder. That folder is basically quarantined off, it doesn't show unless you go looking for it. Items in that folder won't even show up if you search for them, unless you explicitly search within that folder (vs. searching otherwise goes across all folders/labels). On top of this, if the recipient doesn't mark it as not spam within 30 days, it will auto-delete.
When gmail decides something is a promotion, it ... leaves it in your inbox. If you're one of the 1/3 of users who apparently use a feature to sort your inbox (not filter, sort), it will be sorted onto another tab. That tab is visible right in the UI, with a count of how many items are in it. It's one click away. It will show up in searches. It will never auto-delete.
Something being flagged as spam is orders of magnitude worse (in terms of it ever being seen) as something being identified as a promotion, but the authors of this article don't want you to think about that. They literally open it with:
The methodology underlying this article is absolute garbage.
posted by tocts at 9:47 AM on March 1, 2020 [9 favorites]
It's even more absurd than that.
The entire premise of the linked article is that something being flagged as a promotion is equivalent to it being flagged as spam. This is a completely ludicrous premise.
When gmail decides something is spam, it puts it in a totally different folder. That folder is basically quarantined off, it doesn't show unless you go looking for it. Items in that folder won't even show up if you search for them, unless you explicitly search within that folder (vs. searching otherwise goes across all folders/labels). On top of this, if the recipient doesn't mark it as not spam within 30 days, it will auto-delete.
When gmail decides something is a promotion, it ... leaves it in your inbox. If you're one of the 1/3 of users who apparently use a feature to sort your inbox (not filter, sort), it will be sorted onto another tab. That tab is visible right in the UI, with a count of how many items are in it. It's one click away. It will show up in searches. It will never auto-delete.
Something being flagged as spam is orders of magnitude worse (in terms of it ever being seen) as something being identified as a promotion, but the authors of this article don't want you to think about that. They literally open it with:
Pete Buttigieg is leading at 63 percent. Andrew Yang came in second at 46 percent. And Elizabeth Warren looks like she’s in trouble with 0 percent.But that's ridiculous. If you instead look at whose mails are avoiding getting flagged as actual spam and filtered (vs. optionally sorted), a more reasonable opening would be:
Michael Bloomberg leads at 100 percent. Bernie Sanders comes in second at 98 percent. And Joe Walsh looks like he's in trouble with 5 percentAnd Buttigieg, by the way, is in 9th place (72 percent) when looking at actual spam filtering, vs optional inbox sorting, way behind Warren (85%). But that fact is lost on the authors, since it would be inconvenient for their clickbait.
The methodology underlying this article is absolute garbage.
posted by tocts at 9:47 AM on March 1, 2020 [9 favorites]
this thread was blessed by the mods as a place for conspiracy venting
My take, as someone who sometimes-mods and also understands email, is that this thread is more or less for people who do and don't know/use gmail or other email services to talk about the weird way different candidate's email ls processed by gmail and why it might be that way. And also what kind of job the original article did at spelling that sort of stuff out.
My opinion: the article seems to be saying that there's a kinda-sorta gmail conspiracy for/against certain candidates. This thread, of many people who know better than that article's writers, think it's basically no such thing.
posted by jessamyn at 10:51 AM on March 1, 2020 [7 favorites]
My take, as someone who sometimes-mods and also understands email, is that this thread is more or less for people who do and don't know/use gmail or other email services to talk about the weird way different candidate's email ls processed by gmail and why it might be that way. And also what kind of job the original article did at spelling that sort of stuff out.
My opinion: the article seems to be saying that there's a kinda-sorta gmail conspiracy for/against certain candidates. This thread, of many people who know better than that article's writers, think it's basically no such thing.
posted by jessamyn at 10:51 AM on March 1, 2020 [7 favorites]
Google is actively and openly one of the biggest participants in one of the most flagrant, corrosive conspiracies against political power of the masses in the US (Citizens United).
We really don't need to make up histrionic clickbait about Gmail, and it might even detract from the ample legitimate criticism of the company.
posted by aspersioncast at 6:11 PM on March 1, 2020 [2 favorites]
We really don't need to make up histrionic clickbait about Gmail, and it might even detract from the ample legitimate criticism of the company.
posted by aspersioncast at 6:11 PM on March 1, 2020 [2 favorites]
So I've re-read TFA and I do also have some understanding of email and its vicissitudes (having even occasionally engaged in the rather pointless task of running my own SMTP server, not that it matters). While some of the methodological critiques are well-founded (e.g. maybe Gmail was "just" pushing Big Tech's favorite candidates because of the presumed biases of someone logging in from a Tor exit node), I still really don't grasp how anything in this article qualifies as "histrionic clickbait" or an accusation of a "conspiracy".
Representative quote:
I see absolutely no accusations of conspiracy in the article, or any claims that Gmail is deliberately putting a thumb on the scale for anyone. (OTOH, the implication that the people holding this particular form of power are somehow immune from the temptations of every other form of power is pretty extraordinary, and I'd be interested in seeing evidence for it.)
To me, this article seems to be highlighting a real and important concern about one aspect of Big Tech's outsized and unaccountable influence on the political process -- which the article correctly situates alongside existing concerns about newsfeed algorithms. I find the efforts to shut this discussion down through such exaggerated caricatures of the article's actual content quite concerning.
posted by Not A Thing at 8:31 PM on March 1, 2020 [2 favorites]
Representative quote:
“The fact that Gmail has so much control over our democracy and what happens and who raises money is frightening,” said Kenneth Pennington, a consultant who worked on Beto O’Rourke’s digital campaign.Hard to argue with any of that: they do have that power (whether they exercise it intentionally or not), and it is scary.
“It’s scary that if Gmail changes their algorithms,” he added, “they’d have the power to impact our election.”
I see absolutely no accusations of conspiracy in the article, or any claims that Gmail is deliberately putting a thumb on the scale for anyone. (OTOH, the implication that the people holding this particular form of power are somehow immune from the temptations of every other form of power is pretty extraordinary, and I'd be interested in seeing evidence for it.)
To me, this article seems to be highlighting a real and important concern about one aspect of Big Tech's outsized and unaccountable influence on the political process -- which the article correctly situates alongside existing concerns about newsfeed algorithms. I find the efforts to shut this discussion down through such exaggerated caricatures of the article's actual content quite concerning.
posted by Not A Thing at 8:31 PM on March 1, 2020 [2 favorites]
I see absolutely no accusations of conspiracy in the article
The article opens with a claim that some candidates are "leading" and another is "in trouble" because of the (bad) conclusions being pushed in the remainder of it. That is a fairly direct claim of harm by what Google is doing. This is not subtle.
OTOH, the implication that the people holding this particular form of power are somehow immune from the temptations of every other form of power is pretty extraordinary, and I'd be interested in seeing evidence for it.
Tell you what, I'll start proving negatives when you explain how such a thing is possible.
posted by tocts at 8:48 PM on March 1, 2020 [3 favorites]
The article opens with a claim that some candidates are "leading" and another is "in trouble" because of the (bad) conclusions being pushed in the remainder of it. That is a fairly direct claim of harm by what Google is doing. This is not subtle.
OTOH, the implication that the people holding this particular form of power are somehow immune from the temptations of every other form of power is pretty extraordinary, and I'd be interested in seeing evidence for it.
Tell you what, I'll start proving negatives when you explain how such a thing is possible.
posted by tocts at 8:48 PM on March 1, 2020 [3 favorites]
The article is a complete mess. In just the first few paragraphs it zips from 1) "hey, some candidates get more of their emails classified under the 'promotions' tab" -- which is primarily a function of the words and formatting in the emails that each campaign's own marketing folks are composing and sending -- to 2) "hey, because Google sells ads they have an incentive to get people to pay for top placement within the 'promotions' tab."
Conflating those very different topics, and opening with an incendiary lede that isn't supported by the facts of how emails end up in the 'promotions' tab, is why this article is clickbait.
posted by PhineasGage at 9:21 PM on March 1, 2020 [1 favorite]
Conflating those very different topics, and opening with an incendiary lede that isn't supported by the facts of how emails end up in the 'promotions' tab, is why this article is clickbait.
posted by PhineasGage at 9:21 PM on March 1, 2020 [1 favorite]
My opinion: the article seems to be saying that there's a kinda-sorta gmail conspiracy for/against certain candidates. This thread, of many people who know better than that article's writers, think it's basically no such thing.
There doesn't need to really be a "conspiracy" in modernity, all there really needs to be is machine learning, institutional incentives and pretending it's all automatic. The idea that Gmail is putting their thumb on the scale is missing the point. They built the scale. It's their responsibility to demonstrate that it's accurate, and it's baffling how many people in this thread are engaged in what amounts to algorithmic victim-blaming.
If I sign up a hundred million gmail accounts to a hundred million political mailing lists with a certain political bent, and then flag everything from those political lists as spam, the machine sees the trend and reinforces it and now I've successfully poisoned that algorithm, maybe for everyone.
Has that happened? It seems implausible, but it's certainly within the realm of extremely possible, and doesn't even need some nefarious single actor when there's already high-level alignment on rageclicks as the weapon of choice for both the opposing political party and the advertising company that hosts all this email. And that's before we even consider the notion that there really can be single actors putting their fingers on those scales, in-house at Google. I mean, James Damore was getting advice from noted billionaire-vampire Peter Thiel for a long time, and you've gotta know he wasn't alone there.
It would certainly put us at ease if this was demonstrably not true, but the problem is that there no way whatsoever to test or examine any of this, outside of this sort of black-box probing and no transparency from or accountability for the people operating that machine.
It's pretty nuts to me to see how many people in this thread are letting Google off the hook for the way their code works, as though Google is powerless, blameless and faultless. We hold bus drivers and forklift operators to higher standards than this.
posted by mhoye at 6:38 AM on March 2, 2020 [2 favorites]
There doesn't need to really be a "conspiracy" in modernity, all there really needs to be is machine learning, institutional incentives and pretending it's all automatic. The idea that Gmail is putting their thumb on the scale is missing the point. They built the scale. It's their responsibility to demonstrate that it's accurate, and it's baffling how many people in this thread are engaged in what amounts to algorithmic victim-blaming.
If I sign up a hundred million gmail accounts to a hundred million political mailing lists with a certain political bent, and then flag everything from those political lists as spam, the machine sees the trend and reinforces it and now I've successfully poisoned that algorithm, maybe for everyone.
Has that happened? It seems implausible, but it's certainly within the realm of extremely possible, and doesn't even need some nefarious single actor when there's already high-level alignment on rageclicks as the weapon of choice for both the opposing political party and the advertising company that hosts all this email. And that's before we even consider the notion that there really can be single actors putting their fingers on those scales, in-house at Google. I mean, James Damore was getting advice from noted billionaire-vampire Peter Thiel for a long time, and you've gotta know he wasn't alone there.
It would certainly put us at ease if this was demonstrably not true, but the problem is that there no way whatsoever to test or examine any of this, outside of this sort of black-box probing and no transparency from or accountability for the people operating that machine.
It's pretty nuts to me to see how many people in this thread are letting Google off the hook for the way their code works, as though Google is powerless, blameless and faultless. We hold bus drivers and forklift operators to higher standards than this.
posted by mhoye at 6:38 AM on March 2, 2020 [2 favorites]
Well, in return, it's pretty nuts to me that this weak shit is the thing you want to hold Google accountable for.
The irony of this whole thread is, I honestly have serious issues with things Google is doing with their products, including gmail. For example, they have rolled out (to some degree? unsure) features that identify things like flight confirmation emails and use that as a way to try to guide users into their Trips product, a product that competes with many other companies in the online travel space who of course can't do that. That's shitty, monopolistic behavior. That's a thing worth getting worked up about.
Instead, though, we're here talking about what amounts to fantasizing based on bad data or making up metrics to try to massage data into a pre-determined point.
posted by tocts at 6:51 AM on March 2, 2020 [3 favorites]
The irony of this whole thread is, I honestly have serious issues with things Google is doing with their products, including gmail. For example, they have rolled out (to some degree? unsure) features that identify things like flight confirmation emails and use that as a way to try to guide users into their Trips product, a product that competes with many other companies in the online travel space who of course can't do that. That's shitty, monopolistic behavior. That's a thing worth getting worked up about.
Instead, though, we're here talking about what amounts to fantasizing based on bad data or making up metrics to try to massage data into a pre-determined point.
posted by tocts at 6:51 AM on March 2, 2020 [3 favorites]
Well, in return, it's pretty nuts to me that this weak shit is the thing you want to hold Google accountable for.
It's nowhere near the only thing I want, but I feel like starting with the bit that puts somebody interested in regulating tech in public office is a good start.
posted by mhoye at 6:55 AM on March 2, 2020
It's nowhere near the only thing I want, but I feel like starting with the bit that puts somebody interested in regulating tech in public office is a good start.
posted by mhoye at 6:55 AM on March 2, 2020
In a thread where a one line joke got a four paragraph refutation, I feel like maybe some people here are a bit touchy about criticism of Google.
posted by howfar at 10:31 AM on March 2, 2020 [1 favorite]
posted by howfar at 10:31 AM on March 2, 2020 [1 favorite]
Why is the problem specifically the statistical curation, if we're choosing to focus on concentrations of power per se, and not the fact that Google is the email provider for $largefraction of users? I agree that it is a problem that a single actor controls so much dispatching, but it would be a problem even if Google did no algorithmic filtering at all. Say Google stops doing any filtering whatsoever, whether to sub-folders or the spam folder, tomorrow. Are you actually going to be meaningfully less concerned about their power as the mailserver (and a bunch of other gatekeeping) for $largenum people? I'm not.
To say that they haven't proved they aren't being abusive with that power is missing the point that regardless of what they are doing now, they could be arbitrarily abusive tomorrow. I would say that degree of centralization of control over message dispatching should not exist, period, but if it's going to exist then I don't really trust any publicly traded company with it regardless of what promises they make with what credibility. This is a high level symptom of dubious harm (there are jillions of structural biases built into the campaign processes that will harm some candidates and help others - why is this one especially bad that it's worth uniquely discussing), and focusing on it risks losing site of the root cause of the concentration of power in one company's data centers.
posted by PMdixon at 10:36 AM on March 2, 2020 [1 favorite]
To say that they haven't proved they aren't being abusive with that power is missing the point that regardless of what they are doing now, they could be arbitrarily abusive tomorrow. I would say that degree of centralization of control over message dispatching should not exist, period, but if it's going to exist then I don't really trust any publicly traded company with it regardless of what promises they make with what credibility. This is a high level symptom of dubious harm (there are jillions of structural biases built into the campaign processes that will harm some candidates and help others - why is this one especially bad that it's worth uniquely discussing), and focusing on it risks losing site of the root cause of the concentration of power in one company's data centers.
posted by PMdixon at 10:36 AM on March 2, 2020 [1 favorite]
How do you focus on an issue without focusing on the harms it does? How can you expect anyone to care about systemic causes if you think that discussing their effects is a waste of time?
Like...if there was a thread about prison labour, would you say "look, the real problem is structural racism and private prisons, so talking about this is sort of a distraction"? Because I don't really see the difference.
posted by howfar at 12:28 PM on March 2, 2020
Like...if there was a thread about prison labour, would you say "look, the real problem is structural racism and private prisons, so talking about this is sort of a distraction"? Because I don't really see the difference.
posted by howfar at 12:28 PM on March 2, 2020
I see a difference in that if we waved a magic wand tomorrow and abolished prison labor, actually existing human beings could definitely say "yes, my life is less bad now because I don't have to go and perform dangerous work for a pretense of pay on top of having lost all autonomy." What equivalent statement could who make if we waved a magic wand and prevented Google from moving emails into subfolders?
posted by PMdixon at 1:38 PM on March 2, 2020
posted by PMdixon at 1:38 PM on March 2, 2020
It's pretty nuts to me to see how many people in this thread are letting Google off the hook for the way their code works, as though Google is powerless, blameless and faultless. We hold bus drivers and forklift operators to higher standards than this.
Indeed we do, and the reason why we do is itself structural.
Google is a large company. Like all large companies, the services it offers are simply not capable of being offered by any randomly selected person trained for enough hours to become a competent bus driver or forklift operator.
The point is that any randomly selected person can easily get a pretty good idea of what a bus driver or forklift operator actually does, even while acknowledging that they do not have the skills to do it themselves. And based on that understanding, any randomly selected person is also going to be able to give a reasonably useful account of the kinds of things a bus driver or forklift operator should not be doing in order to do the job safely; don't do it drunk, for example.
But the chances of a randomly selected person being able to understand even the requirements for providing the services that a large company does is very, very low. The entire point of forming a large company is to organize the doing of things that no individual or small group could possibly do on their own. From most people's point of view, any given large company is a mysterious black box that stuff goes into and comes out of.
The nature of the political process means that regulation applied to large companies pretty much always has to be aimed at preventing activities that most people will immediately recognize as bad, like Du Pont releasing PFOA into the waterways and giving people and animals birth defects and cancer, or Westpac negligently facilitating millions of dollars in payments to commercial child rapists. The undesirability of these is as obvious as that of a drunken forklift operator dropping a pallet on the head of a co-worker.
But even in cases as clear as those, designing and implementing and enforcing the regulation required to rein in those activities takes almost inconceivable amounts of work. It's simply not feasible to regulate the internal activities of any large company in detail, simply because there are too many different kinds of activities that go on inside them; regulation has to be designed to constrain entire classes of activity across the whole competitive landscape in ways that hopefully discourage large actors from doing too much damage to the public interest.
Whenever a large company interacts directly with the public, there's a huge asymmetry of power built into that relationship. Anybody interested in politics will immediately recognize the inherent risks in that kind of asymmetry. But the sad fact is that most people are not interested in politics at any level beyond spectating as if it were a team sport; the degree to which most people interact with large companies like Google involves close to zero consideration of the inherent risks in doing so. Network effects and convenience, rather than political safety considerations, are almost completely dominant.
The riskiness of Google's ability to influence politics via its email service is a direct consequence of two factors: Google's dominant position as an email provider, and the inherent ability of email providers to inspect the contents of emails. Neither of those factors would be changed by regulation requiring email providers to apply special filtering standards to political communication.
There are plenty of other communication providers to whom neither of those factors apply. If you're seriously interested in mitigating this kind of risk, I think you'd be better advised to apply your efforts to advocacy of switching the preferred digital communication method from Gmail to e.g. Keybase than to pressing for content-sensitive regulation on email providers.
As an example of the kinds of outcomes to be expected from misapplied efforts to regulate the wrong thing one need look no further than the Australian Government, still in the process of attempting to declare pi equal to 3.2 in an attempt to regulate encryption. Something must be done, and this is something, so it's clearly up to us to do it.
posted by flabdablet at 7:50 PM on March 2, 2020 [2 favorites]
Indeed we do, and the reason why we do is itself structural.
Google is a large company. Like all large companies, the services it offers are simply not capable of being offered by any randomly selected person trained for enough hours to become a competent bus driver or forklift operator.
The point is that any randomly selected person can easily get a pretty good idea of what a bus driver or forklift operator actually does, even while acknowledging that they do not have the skills to do it themselves. And based on that understanding, any randomly selected person is also going to be able to give a reasonably useful account of the kinds of things a bus driver or forklift operator should not be doing in order to do the job safely; don't do it drunk, for example.
But the chances of a randomly selected person being able to understand even the requirements for providing the services that a large company does is very, very low. The entire point of forming a large company is to organize the doing of things that no individual or small group could possibly do on their own. From most people's point of view, any given large company is a mysterious black box that stuff goes into and comes out of.
The nature of the political process means that regulation applied to large companies pretty much always has to be aimed at preventing activities that most people will immediately recognize as bad, like Du Pont releasing PFOA into the waterways and giving people and animals birth defects and cancer, or Westpac negligently facilitating millions of dollars in payments to commercial child rapists. The undesirability of these is as obvious as that of a drunken forklift operator dropping a pallet on the head of a co-worker.
But even in cases as clear as those, designing and implementing and enforcing the regulation required to rein in those activities takes almost inconceivable amounts of work. It's simply not feasible to regulate the internal activities of any large company in detail, simply because there are too many different kinds of activities that go on inside them; regulation has to be designed to constrain entire classes of activity across the whole competitive landscape in ways that hopefully discourage large actors from doing too much damage to the public interest.
Whenever a large company interacts directly with the public, there's a huge asymmetry of power built into that relationship. Anybody interested in politics will immediately recognize the inherent risks in that kind of asymmetry. But the sad fact is that most people are not interested in politics at any level beyond spectating as if it were a team sport; the degree to which most people interact with large companies like Google involves close to zero consideration of the inherent risks in doing so. Network effects and convenience, rather than political safety considerations, are almost completely dominant.
The riskiness of Google's ability to influence politics via its email service is a direct consequence of two factors: Google's dominant position as an email provider, and the inherent ability of email providers to inspect the contents of emails. Neither of those factors would be changed by regulation requiring email providers to apply special filtering standards to political communication.
There are plenty of other communication providers to whom neither of those factors apply. If you're seriously interested in mitigating this kind of risk, I think you'd be better advised to apply your efforts to advocacy of switching the preferred digital communication method from Gmail to e.g. Keybase than to pressing for content-sensitive regulation on email providers.
As an example of the kinds of outcomes to be expected from misapplied efforts to regulate the wrong thing one need look no further than the Australian Government, still in the process of attempting to declare pi equal to 3.2 in an attempt to regulate encryption. Something must be done, and this is something, so it's clearly up to us to do it.
posted by flabdablet at 7:50 PM on March 2, 2020 [2 favorites]
How do you focus on an issue without focusing on the harms it does? How can you expect anyone to care about systemic causes if you think that discussing their effects is a waste of time?
I really think this comes back to TFA which has a really sort of facile and naive interpretation of a system that is actually kinda of complicated. And that's fine, a lot of people feel this way about how email works. However there are also a lot of people, many of whom are on MetaFilter and actually here in this thread, who do know how email works and know why it's a hard problem. And, more to the point, know that the way this article's experiment worked didn't really give us enough data. Things that would do that"
- making up a new account from scratch and using an existing account (or ten) and subscribe to ALL THE SAME email lists and see how they got sorted
- doing the same experiment as in TFA except use two accounts one that is set up using Tor and all the privacy and another that is closelylinked with an existing account, see if the results differ
- make a bunch of fake-o political candidate emails and see if you could affect how they got sorted in one (or ten) gmail accounts with differing profiles
- copy the emails from some of the candidates in this article word for word and see if they get sorted differently if they're coming from a different email server
Because right now, there are SO many variables in addition to the "Gmail is doing something shady" that it's kind of impossible to focus on the "Gmail is doing something shady." Because, honestly, I am pretty sure they are doing something shady, lots of somethings shady, and that's interesting to talk about, but it's not interesting at all because of the things this article points to (though I was happy to read it and happy for this conversation). Maybe I'm just jaded because I live in a state and work in a profession within that state where very few people are conversant in tech so I immediately jump to user error in terms of why email got sorted the way it did even if the other options are within the realm of possibility.
posted by jessamyn at 8:33 PM on March 2, 2020 [3 favorites]
I really think this comes back to TFA which has a really sort of facile and naive interpretation of a system that is actually kinda of complicated. And that's fine, a lot of people feel this way about how email works. However there are also a lot of people, many of whom are on MetaFilter and actually here in this thread, who do know how email works and know why it's a hard problem. And, more to the point, know that the way this article's experiment worked didn't really give us enough data. Things that would do that"
- making up a new account from scratch and using an existing account (or ten) and subscribe to ALL THE SAME email lists and see how they got sorted
- doing the same experiment as in TFA except use two accounts one that is set up using Tor and all the privacy and another that is closelylinked with an existing account, see if the results differ
- make a bunch of fake-o political candidate emails and see if you could affect how they got sorted in one (or ten) gmail accounts with differing profiles
- copy the emails from some of the candidates in this article word for word and see if they get sorted differently if they're coming from a different email server
Because right now, there are SO many variables in addition to the "Gmail is doing something shady" that it's kind of impossible to focus on the "Gmail is doing something shady." Because, honestly, I am pretty sure they are doing something shady, lots of somethings shady, and that's interesting to talk about, but it's not interesting at all because of the things this article points to (though I was happy to read it and happy for this conversation). Maybe I'm just jaded because I live in a state and work in a profession within that state where very few people are conversant in tech so I immediately jump to user error in terms of why email got sorted the way it did even if the other options are within the realm of possibility.
posted by jessamyn at 8:33 PM on March 2, 2020 [3 favorites]
I am pretty sure they are doing something shady, lots of somethings shady
...more shady things than could possibly be even identified in order to regulate, and most of which would be quite capable of being weaponized at a moment's notice to do irreversible amounts of harm to the political system before any regulatory response could conceivably be mounted.
It bears musing upon that the existence of the EPA did not stop DuPont dumping PFOA in the waterways, and that dumping PFOA is not, in and of itself, even the kind of shady activity that might contribute directly to a reduction in the strength of the regulator. Manipulation of an electoral process, on the other hand, absolutely is.
As always it's the de facto monopoly power of a private corporation that creates this whole class of public risk. And if anybody out there can think of a way to stop the body politic from keeping on granting de facto monopoly powers to private corporations via network effects, please do share. Because as hard as the problem of email filtering is, it's as nothing compared to the problem of our collective tendency to shoot ourselves in the foot in pursuit of a moment's convenience.
posted by flabdablet at 11:07 PM on March 2, 2020 [1 favorite]
...more shady things than could possibly be even identified in order to regulate, and most of which would be quite capable of being weaponized at a moment's notice to do irreversible amounts of harm to the political system before any regulatory response could conceivably be mounted.
It bears musing upon that the existence of the EPA did not stop DuPont dumping PFOA in the waterways, and that dumping PFOA is not, in and of itself, even the kind of shady activity that might contribute directly to a reduction in the strength of the regulator. Manipulation of an electoral process, on the other hand, absolutely is.
As always it's the de facto monopoly power of a private corporation that creates this whole class of public risk. And if anybody out there can think of a way to stop the body politic from keeping on granting de facto monopoly powers to private corporations via network effects, please do share. Because as hard as the problem of email filtering is, it's as nothing compared to the problem of our collective tendency to shoot ourselves in the foot in pursuit of a moment's convenience.
posted by flabdablet at 11:07 PM on March 2, 2020 [1 favorite]
Maybe I'm just jaded because I live in a state and work in a profession within that state where very few people are conversant in tech so I immediately jump to user error in terms of why email got sorted the way it did even if the other options are within the realm of possibility.
I feel like this comes back to how you take the article, and, in particular, one's views on how it might attract attention, why it might be useful that it do so, and who it might be useful to. I think that, for an audience that isn't conversant in how email works, pointing out that something shady could be going on, and illustrating this in specific terms that are meaningful to them, is going to be much more effective in inducing worry about the real structural issues than any number of clear and evenhanded explanations by people who feel prevented from suggesting possible examples.
I personally think the piece serves its purpose effectively, and I think that part of the weird miscommunication in this thread has probably been a strong divergence about what people think the point and audience of the article are.
posted by howfar at 9:24 AM on March 3, 2020
I feel like this comes back to how you take the article, and, in particular, one's views on how it might attract attention, why it might be useful that it do so, and who it might be useful to. I think that, for an audience that isn't conversant in how email works, pointing out that something shady could be going on, and illustrating this in specific terms that are meaningful to them, is going to be much more effective in inducing worry about the real structural issues than any number of clear and evenhanded explanations by people who feel prevented from suggesting possible examples.
I personally think the piece serves its purpose effectively, and I think that part of the weird miscommunication in this thread has probably been a strong divergence about what people think the point and audience of the article are.
posted by howfar at 9:24 AM on March 3, 2020
I personally think the piece serves its purpose effectively, and I think that part of the weird miscommunication in this thread has probably been a strong divergence about what people think the point and audience of the article are.
I understand the point of the article to be that Google is doing a specific bad thing that is disadvantaging left candidates specifically, and that in the absence of that specific bad thing everything would be OK - we just need the right bell on the cat, IOW.
What do you understand the point of the article to be?
posted by PMdixon at 9:54 AM on March 3, 2020 [1 favorite]
I understand the point of the article to be that Google is doing a specific bad thing that is disadvantaging left candidates specifically, and that in the absence of that specific bad thing everything would be OK - we just need the right bell on the cat, IOW.
What do you understand the point of the article to be?
posted by PMdixon at 9:54 AM on March 3, 2020 [1 favorite]
What it actually says.
"Some fear that, as a result, Gmail has the same conflict of interest that exists on social networks: If the platforms make it too easy to reach people for free, no one will buy ads."
You didn't read it, did you?
posted by howfar at 12:59 PM on March 3, 2020
"Some fear that, as a result, Gmail has the same conflict of interest that exists on social networks: If the platforms make it too easy to reach people for free, no one will buy ads."
You didn't read it, did you?
posted by howfar at 12:59 PM on March 3, 2020
OK. So Google is doing a bad thing (curating inboxes algorithmically), by the pull quote and the choice of people used to provide quotes I am led to believe that this is more of a problem for the left, and if Google was prevented from doing this then Change.org petitions that would otherwise peter out would be successful.
I don't usually find it useful to leap to the conclusion that if someone takes different things away from a text than me that they're just making shit up out of whole cloth.
posted by PMdixon at 2:41 PM on March 3, 2020
I don't usually find it useful to leap to the conclusion that if someone takes different things away from a text than me that they're just making shit up out of whole cloth.
posted by PMdixon at 2:41 PM on March 3, 2020
That different people read things differently was, of course, the point I was trying to make; it was accordingly unfortunate, in my view, that your response was couched in sarcastic terms, and that it either deliberately or out of ignorance neglected a number of substantive points in the article that reflected a reading that many other people, including myself, have already presented in this thread. It was accordingly difficult for me to believe that you were attempting to engage seriously with the article, or with my observation about it. I trust this is sufficient to explain my response.
posted by howfar at 8:16 AM on March 4, 2020
posted by howfar at 8:16 AM on March 4, 2020
« Older "We saw, every day, the effects of giving somebody... | "Improvisation is a characteristic of both music... Newer »
This thread has been archived and is closed to new comments
Surely that’s what this all is, right? A coincidence. Thank you, Machine Learning, for reassuring us that we as developers are powerless and blameless, that our hands are clean.
posted by mhoye at 8:41 AM on February 29, 2020 [40 favorites]