[User23]

> Pichai said “there is nothing in the algorithm that has to do with political ideology”

That’s an interesting non-denial: he never actually says the spam algorithm isn’t biased against conservatives. There could well be a bias, because many ML systems have unintended biases. Google could easily answer the question. It’d probably take an hour max to write and run the mapreduce or whatever they use now.

[nicoburns]

I'd be very surprised if democratic party emails don't also frequently get marked as spam. I somehow ended up on their list despite not even living in the US, and it was a complete nightmare to get off of it because the unsubscribe link only worked for a specific campaign. I've marked hundreds of their emails as spam by now, and it's entirely their own fault.

[saghm]

I've somehow gotten on mailing lists from candidates and politicians from both major parties over the past several years, and many of them did not seem to respect when I try to unsubscribe, which leads to me marking them as spam. Assuming that I'm not the only one suffering from not respecting my wishes to unsubscribe from lists I never signed up for, I would have to imagine that with enough frustrated recipients making the messages as spam, the algorithm infers that all of the messages from the list are spam. This wouldn't even require the algorithm to even recognize that the messages are political, let alone infer the ideology; I did the same thing with emails trying to sell me sunglasses.

[syshum]

What I find ironic about google's position is they believe implicit unconscious bias permeates all of society, except when it comes to their programmers who are clearly majority democrat....

It should not be ironic, because Google has a history of seeing itself above any flaws, and as the Arbiter of Truth

[Retric]

The algorithm wouldn’t have a bias vs conservatives because it doesn’t include semantic meaning, but it could have a bias vs methods used by this conservative. Which really just comes down on whoever is generating those emails.

[klyrs]

I don't know if it does or doesn't, but Republicans and Democrats use different keywords, not just semantics. For example, 'law and order', 'police reform', etc. It's certainly possible for ML to pick up a bias in the presence of such differences.

[Retric]

They also quote each other, without context it’s just word choice not the message. Considering messages are already tailored to different kinds of voters, spam algorithms are just another part of messaging.

[kfprt]

Politicians of both parties often use the same middlemen. The 'campaigning industry' isn't so much red or blue but green.

[therealcamino]

What organizations are you thinking of? I'm not an expert in the area, and ones I know of are ActBlue that works only with Democrats, and WinRed which is the Republican clone.

[Retric]

Often true, but this was a specific case and there are a lot of red or blue focused middlemen.

[QuadmasterXLII]

It's probably just detecting and deleting capslock