undefined | Better HN

story

0 pointsthisiswater5y ago0 comments

"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.

0 comments

angrais5y ago

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.

hyperpallium25y ago

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.

thisiswaterOP5y ago

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?

j / k navigate · click thread line to collapse