Hacker News Reader

Comment by dybber | Hacker News Reader

The criticism was that Google have a dominant position on search market, Google selected 1% of their users to run the experiment on, but without informing them. That is users didn’t know that their search results were manipulated and articles they would otherwise have found didn’t show up.

So the argument presented by Danish authorities and media companies were that information should flow freely in a democracy and by doing a huge experiment like this without informing users is against the rights of Danish citizens.

View on HN

"Manipulated" is a loaded and meaningless term here. All results are generated by algorithm, so that means 99% see the output of algorithm A and 1% see the output of algorithm B. Neither is more "manipulated" than the other.

It is possible that neither are more manipuled tough it's impossible to tell. What seems clear from your example above is that both are manipulated, just in different ways and with google's incentive. It is understandable that countries came to the conclusion that this is posing a threat to their national security.

"Manipulated" has strong negative connotations, but it could just mean that the results are chosen and controlled by the search engine. In which case, it's a meaningless statement. The entire purpose of any search engine is to choose results for queries.

Or it can mean the results were altered from some ideal baseline algorithm that we consider unmanipulated. The only obvious candidate for this baseline would be the search engine's regular algorithm. But if you're saying that's not the baseline, then it's unclear what you consider to be the true baseline and therefore unclear what "manipulated" means.

I agree that countries may consider search engines, social media, or anything else that can affect flow of information to be a national security threat.

And what, exactly, is the national security threat here? If Google is manipulating results to favor its advertisers or the political positions of its owners, that's what all publishers do, and have always done and nobody ever called it a national security threat. The "national security threat" here seems to be that they are showing people content that the government doesn't want them to see.

Of course they wouldn't tell users if they're in the control group or experimental group. It would destroy the validity of the experiment.

You still should have to consent to be studied.

I’ve been part of multiple clinical trials and consent was always there. The control group exists. They know they’re in the study but they may not know they’re the control group.

I'm not commenting on the ethics of A/B testing without informing the customer.

Maybe I misunderstood the comment I was replying to. If they meant that the experiment's results were probably valid but conducting the experiment was unethical, then my response was unnecessary.

First it is not manipulation please read the terms of service and user consent on this issue. Second this is standard practice A/B testing is universal and companies do a holdout experiment all the time it is also called Withholding test.

Reply on news.ycombinator.com