[strogonoff]

I observed a clean experiment that showed a friend’s Google Pixel phone listening to us and adjusting news stories on Google app’s home screen.

However:

— IIRC the phone was unlocked,

— this only affected the news feed, and

— this was 5–6 years ago.

We 1) noted how Google app shows some selection of news after opening, 2) talked clearly for a minute about a very random and conspicuous topic in presence of the unlocked phone, and 3) demonstrated that the Google app showing an article relevant to the topic within a few minutes. The article was a few days old, too, so it was clearly boosted out of more recent stories.

The only reason it could be something other than the phone microphone is if I was misled by my friend steering us towards a predefined topic. However, that would require some extensive preparation to rule out the story appearing in the first step and would be very atypical for that person.

I recall seeing an article about Google admitting this and changing their policy to stop, but can’t seem to find it now. I imagine it was bad publicity, though to my friend it was a feature to see personalized content.

[Aurornis]

This was a coincidence.

That’s why it’s something you observed one time 5-6 years ago, not something that happens repeatedly in a testable way.

[theoreticalmal]

Isn’t it more likely it’s not a coincidence though?

[Aurornis]

How often does someone look at their phone over 5-6 years?

Having one incidence where you’re talking about something and then you also see that something on your phone out of 2000 days of using a phone is definitely more likely to be coincidence.

[afiori]

How often do you think this person did experiments? It is a study with n=1 but the unrelated metric of how many times something else happens does not influence the likelihood of a false positive

[strogonoff]

Only did it once. The likelihood of coincidence is low, because the topic was specific and unusual.

Here’s something relevant in Google’s current support KB[0], where the combination of the following further supports that the experiment did not have be staged (emphasis mine):

> Web & App Activity saves your searches and activity from other Google services in your Google Account. You may get more personalized experiences, like: <…> Content recommendations

> When Web & App Activity is on, you can include audio recordings from your interactions with Google Search, Assistant, and Maps as part of your activity.

Let’s now go back to the experiment. Given the phone was unlocked, voice activity was enabled, and Google app or search widget was on Google Pixel’s screen (I am certain at least the latter was true) during the experiment, could talking near the phone be counted as “interaction”? If the answer is “yes” then it seems very reasonable for us[1] to expect, per that KB, that the app would listen more actively than what’s required for assistant activation, and that recorded snippets would count as your “activity” designed to affect content recommendations (including the article feed Google app showed to us on its app’s main screen).

No tinfoil hat required.

***

Note that it does not mention ads among personalized experiences[2], and we had not observed any change in the ads either. I didn’t see what exactly counts as “interaction” or whether this blazing-fast content personalization used to include ads previously, but in line with the “move fast” culture of mid-2010s Silicon Valley it could well have been much more lax at some point. If so, I do not envy all the people who have observed it only to be gaslit and mocked by peers and media.

***

As to the article I was vaguely remembering in my original comment, the above makes me think that it was merely about the change of the default to opt-in, which it is as of today:

> This voice and audio activity setting is off unless you choose to turn it on.

[0] https://support.google.com/websearch/answer/54068?hl=en&co=G...

[1] Us tech people; this might not at all align with the intuition of other people.

[2] I rather suspect that ToS and possibly some other KB article would indicate that your activity would, in fact, affect your interest profile and by extension ads, but probably in a much less obvious and more gradual fashion.