I Fooled Millions Into Thinking Chocolate Helps Weight Loss

Re:I Fooled Millions Into Thinking Chocolate Helps Weight Loss by bjourne on Sunday June 14, 2015 @02:56PM (#9715947)

It's not news that the peer-review system isn't safe from exploits. People have been using Markov chains to generate studies which have then been published. And journalists uncritically picks up everything they can. All functions in society are based on trust and are easily exploitable if a person is not behaving trustworthy.

So making a fake study about chocolate and getting lots of magazines to publish it I think is a little unethical. It's not news, and the reason most people don't do it is because they want to act trustworthy. For example, someone could, if they wanted, submit patches to some free software projects which would introduce awful security holes and get them merged or just write shit on Wikipedia.

[ Parent | on HN ]

Re:I Fooled Millions Into Thinking Chocolate Helps Weight Loss by thomasahle on Sunday June 14, 2015 @03:01PM (#9715969)

It wasn't actually a fake study. That's the point.

They used the same statistical methods as too many 'honest' groups out there, and got results thereafter.

[ Parent | on HN ]

Re:I Fooled Millions Into Thinking Chocolate Helps Weight Loss by bjourne on Sunday June 14, 2015 @06:07PM (#9716503)

I disagree. The study was fake because the intent was to deceive. He used flawed methods, but could have used 10 times as sophisticated methods and still would have been able to publish completely invalid results.

[ Parent | on HN ]

Re:I Fooled Millions Into Thinking Chocolate Helps Weight Loss by olifante on Monday June 15, 2015 @01:18PM (#9720686)

I wasn't a fake study, it was a dishonest one.

[ Parent | on HN ]

Re:I Fooled Millions Into Thinking Chocolate Helps Weight Loss by munro on Sunday June 14, 2015 @03:37PM (#9716062)

To continue this line of though, distributed computation is hard problem, and that's just computers. Here, people have to do real work to statically verify a paper's procedure, which is very cognitive overloading. Imagine reviewing a PR and deciding with certainty if it will work or not. Even then they should also "recompute" the experiment to completely verify it. Of course add human error on top of all this.

It seems there's a problem here, the article made it seem anyone can produce a bad study for lulz, and get everyone's attention. What is more scary is people doing it for vested interests. Maybe there's something that could be learned from projects like BIONC, Wikipedia, or Bitcoin, but better science is going to cost more money, and require more cultural attention. Not sure how to fix those things :)

[ Parent | on HN ]