Comment by thisiswater | Hacker News Reader

Comment by thisiswater | Hacker News Reader

I would like to contribute an anecdote which turned me from a "google is listening" skeptic to someone who turns their phone off as often as possible, especially during any kind of sensitive conversation. I'm one degree removed from this anecdote, but I have no reason to doubt its truth.

A group of friends were on a road trip, driving for several days cross-country. These are university-aged young males, 20-23. None have any interest in starting a family any time soon.

At the beginning of the trip, they agreed (verbally) to test if google was listening by discussing "nappies" intermittently, loudly and jokingly. They would only discuss nappies in person - everyone agreed not to mention this on any digital channel whatsoever, not to look up nappies for any reason, etc. That is, they deliberately excluded nappies from their online lives.

They chose nappies specifically because they're completely irrelevant. None had ever to their knowledge been delivered an ad for nappies. None had ever purchased nappies or any baby product.

Sure enough, end of the trip - ads for nappies.

I would love an explanation more plausible than their phones were listening.

>> I would love an explanation more plausible than their phones were listening.

One of the three deviated and searched for nappies online prior to the trip.

Rather than relying on external people to make judgement, you could test the hypothesis yourself: write on paper a phrase. Spend one week without ads and document all ads you see. Spend second week saying word N times per day. Also document all ads you see.

Ask N people to do this and report your findings, then you might provide statistical insight and validate your hypothesis.

Otherwise, who's to say one of those students didn't Google it before the trip?

Does anyone have any ways to improve the design of the above experiment?

"One of the three deviated and searched for nappies online prior to the trip."

This is a hypothetical which was deliberately excluded when they decided to run the experiment. These are intelligent young people acting in good faith to test if they're being spied on. Why would they have googled nappies?

I'm sorry but I don't find that explanation convincing at all.

I agree. While they may have made an effort to control for not searching about nappies, they could have implicitly biased themselves in other ways. For example, one may have googled something similar without realising the similarity, e.g., "Baby Shark" (a popular song a few years ago) or such. In other words, nappies may be related to many other terms that they may have (implicitly) searched for.

That's why running an experiment and capturing data (of searches amongst other things) could be more meaningful than anecdote from friends if a friend or such.

Yes, talking about nappies a lot probably has effects on other aspects of one's internet usage. With massive data, google may well pick up these statistical trends.

A similar performance to a perceptive fortune teller. Google can appear to be a mind reader.

Apparently, walmart used to send targetted ads when the pattern for pregnancy was detected - which created problems where not all parties knew about it, so walmart disabled it.

Another commentor's two week experiment might need to have you think about nappies for the first "control" week, to account for this bias.

Also, note that "nappies" wasn't randomly selected, but suggested by the group. This choice and the ads may have had a common cause. e.g, Young adults concerned about pregnancy.

It's true, there could definitely have been confounding factors and I wish there was a better record than their memories. I remain personally convinced due to knowing and trusting the involved parties, but agree that the anecdote doesn't (and shouldn't) carry much weight as objective evidence.

This seems like something which privacy researchers might be interested in, and wouldn't be too difficult to run - I wonder if anyone here can point us towards any more formal experiments that have been run on this?

Edit: The more I reflect, the more difficult the experiment seems to be. How does one recruit for an experiment, contact and instruct participants, etc., probably via surveilled email or messaging platforms, without creating data linking participants to one another?

Do you ever get irrelevant ads? I certainly do. Multiply that by billions. You're going to find someone getting ads irrelevant to them.

They could also have looked up something relevant to nappies, stayed in a house (and used internet there) where the owners have a baby, etc.

They were camping so I think it's unlikely they got confused with an ad ID which was interested in nappies.

For the sake of good faith I think we should exclude "they broke the rules". Frankly, my friends aren't that stupid.

Recently I started ordering groceries online. I found the site "organically" but searching for products in the space. I found one which satisfied me and stuck with it. Some time later I start getting Ads in Facebook Messenger for the website. Yes, not totally irrelevant, but it's like seeing a billboard for the supermarket you're on your way to.

That's just straight up ad personalization.

Did everyone stop using the internet and stick strictly to cash? If not, is it possible that someone made a purchase or search that may have related to what people who need nappies might buy or search for, but didn't realize it?

Reply on news.ycombinator.com