5 ms·

We have a chance to notice this when it's mediated by a browser. Native apps don't barf all over the console and thus escape such scrutiny. I first considered

We have a chance to notice this when it's mediated by a browser. Native apps don't barf all over the console and thus escape such scrutiny. I first considered this when a friend told me about a brand of lawnmower of which I had never heard let alone searched (mowing lawns is the least interesting activity I can imagine), and one minute later a podcast app had a big banner at the top by which I could purchase a lawnmower of that exact brand. I don't have important conversations in the vicinity of mobile phones anymore.

avsteele 5y ago
Has anyone seen any well done research showing these effects?
- goldenkey 5y ago
  I've heard it from more folks than I'd like to. And counting them all as crazy or paranoid is less believable than the alternative.
  - ricardobayes 5y ago
    I've only experienced it first-hand on youtube. After trying to learn spanish and talking spanish on the phone (but searched nothing on youtube in Spanish) Youtube offers recommendations in Spanish. Also after I sneezed a couple of times it recommends something related to hay fever. Lots more. I find these way more than circumstancial.
  - andybak 5y ago
    > And counting them all as crazy or paranoid Nobody is saying that. It's simply a quirk of human psychology. We are pattern matching machines with a poor intuitive grasp of probability.
  - pumnikol 5y ago
    A while ago, out of the blue, I've been diagnosed with a very rare medical condition. Online ads and recommended videos then started showing relevant and very specific drugs, treatments, therapies, self-help groups etc. for me. I had not performed any online or offline research about the topic, had not talked to anyone about it at all, except the MD, and had bought all drugs with cash. What are the odds? Then at work, in a web meeting, I made a completely spontaneous pun about having a divorce. Minutes later, my social media were completely plastered with ads for divorce lawyers, which they had never been before. I've never been married, don't want to be, and I don't even know anyone who has gotten a divorce in the last five years. And more, and more. If social media companies are selling health and other personal data to advertisers, what's stopping them from selling it to insurances and recruiters? Maybe I am all wrong about this, and maybe many others are also. Or maybe we'll start seeing a big wave of whisteblowing and revelation books once the current crop of software engineers, managers and directors retire and/ or are sufficiently strapped for cash.
- keanebean86 5y ago
  Not a study but I've heard explanations that big companies have so much data their models are just that good. For instance they know who you're friends with, both your search histories, and location data. Knowing you friend searched for lawnmowers recently and now y'all are physically close it's possible that brand was discussed. Although I wouldn't be shocked at all if mics were being used. I just feel like that would have been leaked by someone by now.
  - clairity 5y ago
    no, they have so much data because their models are so bad. the vast data is used to give the illusion of good models by sheer volume, so little random matches are made more likely. that’s why google and the like want to hoover up our data, they’re desperate to keep the gravy train rolling long enough to create the good models and keep it going some more.
SahAssar 5y ago
I've heard similar stories many times but every time it seems to be baader-meinhof effect/frequency illusion or that the linking is not via audio but something else. People have man-in-the-middle checked advertising traffic to see if they either stream audio or send spoken keywords to ad servers and they do not seem to do that. It's more likely that it saw your phone and your friends at similar locations and your friend had searched for the brand before, therefore linking your ad profile to the brand. Do you think the apps on your phone real-time stream all mic audio or that they run speech-to-text on your device?
- colordrops 5y ago
  Would it have to be audio? It could do TTS on the phone then send the text back.
  - SahAssar 5y ago
    I said "run speech-to-text on your device". Also TTS would be the other way around as it means text to speech.
    - colordrops 5y ago
      Sorry I meant STT
- eloisius 5y ago
  That's almost scarier. Seems like it could leak embarrassing information about what you've been searching or buying to your friends.
  - SahAssar 5y ago
    > it could leak embarrassing information about what you've been searching or buying to your friends. Yes. Say no to tracking, regardless of if it listens to your voice. Even if it does not leak this way it's probable that one of the major ad brokers will leak data in the future.
marcellus23 5y ago
But mobile apps are required to ask for mic/camera permissions.
- kzrdude 5y ago
  There are exploits that circumvent this, of course.
  - andybak 5y ago
    I don't think you'd waste an exploit like that to server a lawnmower ad.
  - marcellus23 5y ago
    Sure, but those exist for the web too…
guptaneil 5y ago
This is a common belief that phones must be listening to us because ads are so targeted, but the scary truth is they aren’t[0] because they don’t need to. They have far more effective ways of targeting ads. For example, the reason you probably saw the lawnmower ad is because your friend searched for lawnmowers, and google knows they are friends with you, so they showed you targeted ads too because you might recommend that brand if you talk or you might subconsciously register that brand and reaffirm their decision to buy with something like “oh yeah I’ve heard of X, they’re supposed to be the best” without remembering where you saw that. It’s even possible you got the ad before your chat but didn’t notice because you had no reason to pay attention to a lawnmower ad. This is why data privacy is so important even if you feel like you have nothing to hide. 0: as you note, they technically have the ability to do so and random apps could be but the amount of effort it would take to record, transcribe, and evaluate that much data just isn’t worth it when most users voluntarily give their info anyway. This is why I don’t use a phone, browser, or email service created by an ad tech company though and it boggles my mind how many people are ok with that.
- jessaustin 5y ago
  In this case, the conservative explanation seemed unlikely for several reasons. This friend would be more precisely described as the elderly friend of my elderly parents. Neither of us are on social media, she doesn't know what podcasts are, and neither of us had even used a web browser in the preceding several hours. Neither of us was in the market for a new riding mower nor had been in the preceding decade. I was on her (remote, inconsistent-cell-reception) farm to help with some cattle. I had never heard of this brand before, and now over a year later I can't think of it. Is it so unlikely that an ad network sketchy enough to pay its way onto random Android apps would also be sketchy enough to monitor conversations for keywords that can get it paid? I don't think that's unlikely at all.

avsteele 5y ago

Has anyone seen any well done research showing these effects?

goldenkey 5y ago
I've heard it from more folks than I'd like to. And counting them all as crazy or paranoid is less believable than the alternative.
- ricardobayes 5y ago
  I've only experienced it first-hand on youtube. After trying to learn spanish and talking spanish on the phone (but searched nothing on youtube in Spanish) Youtube offers recommendations in Spanish. Also after I sneezed a couple of times it recommends something related to hay fever. Lots more. I find these way more than circumstancial.
- andybak 5y ago
  > And counting them all as crazy or paranoid Nobody is saying that. It's simply a quirk of human psychology. We are pattern matching machines with a poor intuitive grasp of probability.
- pumnikol 5y ago
  A while ago, out of the blue, I've been diagnosed with a very rare medical condition. Online ads and recommended videos then started showing relevant and very specific drugs, treatments, therapies, self-help groups etc. for me. I had not performed any online or offline research about the topic, had not talked to anyone about it at all, except the MD, and had bought all drugs with cash. What are the odds? Then at work, in a web meeting, I made a completely spontaneous pun about having a divorce. Minutes later, my social media were completely plastered with ads for divorce lawyers, which they had never been before. I've never been married, don't want to be, and I don't even know anyone who has gotten a divorce in the last five years. And more, and more. If social media companies are selling health and other personal data to advertisers, what's stopping them from selling it to insurances and recruiters? Maybe I am all wrong about this, and maybe many others are also. Or maybe we'll start seeing a big wave of whisteblowing and revelation books once the current crop of software engineers, managers and directors retire and/ or are sufficiently strapped for cash.
keanebean86 5y ago
Not a study but I've heard explanations that big companies have so much data their models are just that good. For instance they know who you're friends with, both your search histories, and location data. Knowing you friend searched for lawnmowers recently and now y'all are physically close it's possible that brand was discussed. Although I wouldn't be shocked at all if mics were being used. I just feel like that would have been leaked by someone by now.
- clairity 5y ago
  no, they have so much data because their models are so bad. the vast data is used to give the illusion of good models by sheer volume, so little random matches are made more likely. that’s why google and the like want to hoover up our data, they’re desperate to keep the gravy train rolling long enough to create the good models and keep it going some more.

SahAssar 5y ago

I've heard similar stories many times but every time it seems to be baader-meinhof effect/frequency illusion or that the linking is not via audio but something else. People have man-in-the-middle checked advertising traffic to see if they either stream audio or send spoken keywords to ad servers and they do not seem to do that. It's more likely that it saw your phone and your friends at similar locations and your friend had searched for the brand before, therefore linking your ad profile to the brand. Do you think the apps on your phone real-time stream all mic audio or that they run speech-to-text on your device?

colordrops 5y ago
Would it have to be audio? It could do TTS on the phone then send the text back.
- SahAssar 5y ago
  I said "run speech-to-text on your device". Also TTS would be the other way around as it means text to speech.
  - colordrops 5y ago
    Sorry I meant STT
eloisius 5y ago
That's almost scarier. Seems like it could leak embarrassing information about what you've been searching or buying to your friends.
- SahAssar 5y ago
  > it could leak embarrassing information about what you've been searching or buying to your friends. Yes. Say no to tracking, regardless of if it listens to your voice. Even if it does not leak this way it's probable that one of the major ad brokers will leak data in the future.

marcellus23 5y ago

But mobile apps are required to ask for mic/camera permissions.

kzrdude 5y ago
There are exploits that circumvent this, of course.
- andybak 5y ago
  I don't think you'd waste an exploit like that to server a lawnmower ad.
- marcellus23 5y ago
  Sure, but those exist for the web too…

guptaneil 5y ago

This is a common belief that phones must be listening to us because ads are so targeted, but the scary truth is they aren’t[0] because they don’t need to. They have far more effective ways of targeting ads. For example, the reason you probably saw the lawnmower ad is because your friend searched for lawnmowers, and google knows they are friends with you, so they showed you targeted ads too because you might recommend that brand if you talk or you might subconsciously register that brand and reaffirm their decision to buy with something like “oh yeah I’ve heard of X, they’re supposed to be the best” without remembering where you saw that. It’s even possible you got the ad before your chat but didn’t notice because you had no reason to pay attention to a lawnmower ad. This is why data privacy is so important even if you feel like you have nothing to hide. 0: as you note, they technically have the ability to do so and random apps could be but the amount of effort it would take to record, transcribe, and evaluate that much data just isn’t worth it when most users voluntarily give their info anyway. This is why I don’t use a phone, browser, or email service created by an ad tech company though and it boggles my mind how many people are ok with that.

jessaustin 5y ago
In this case, the conservative explanation seemed unlikely for several reasons. This friend would be more precisely described as the elderly friend of my elderly parents. Neither of us are on social media, she doesn't know what podcasts are, and neither of us had even used a web browser in the preceding several hours. Neither of us was in the market for a new riding mower nor had been in the preceding decade. I was on her (remote, inconsistent-cell-reception) farm to help with some cattle. I had never heard of this brand before, and now over a year later I can't think of it. Is it so unlikely that an ad network sketchy enough to pay its way onto random Android apps would also be sketchy enough to monitor conversations for keywords that can get it paid? I don't think that's unlikely at all.