|
Pages: [1] 2
|
 |
|
Author
|
Topic: Echo chambers generate intellectual diversification (Read 2033 times)
|
Zanthius
Enlightened
 Offline
Posts: 941
|
We often worry about how the confirmation bias is causing people on the Internet to live inside of echo chambers. But are echo chambers necessarily just bad?
Before the development of Internet, I didn't have access to Wikipedia, and not equally good access to scientific news, documentaries, and lectures.
|
|
|
|
|
Logged
|
|
|
|
Scalare
*Many bubbles*
Offline
Posts: 245
|
I honestly don't have the ability to judge the information I read on the internet to be true or false.
I was never taught how to read scientific research and how to interpret it. So I have to make do with news articles about scientific discoveries, and a lot of my knowledge gaining actually comes from youtube videos these days  .
But to judge those news articles and youtube videos on being true or false, I usually just go by my gut feeling. And when an article tells small lies I disregard it entirely. But when it does convince me of something, it could also just as easily convince me of the wrong things. For example scientific research in which the P<0.05 was almost met but the news article still presents it as a confirmed hypothesis. Or even worse, it makes wrong connections between research and possible applications (the most known example is that cannabis cures cancer, because it killed cancer cells in a petri dish, and now people think that it also works in vivo 
Also, I think that many people who have studied in their younger years stop learning when they get their first job.
But when you work in IT like I do that is simply impossible, I'm lucky to have the ability to learn new things reasonably fast but I wonder when my neuroplasticity will run out and that would mean that I would become less usefull in my job.
I am curious by nature, a day is only succesful for me if I've learned something new. But I still get proven wrong a lot of times by others  .
Anyway, about echo chambers and the point I was going to make (sorry for the distraction).
Echo chambers can be generated by facebook, twitter and google. The internet isn't like it used to be, there used to be a time when curious websites were discovered by you by accident or shared in your social circles..
You actively chose which websites you would visit. But now the internet adapts to what it thinks you want to see. And this can also be used maliciously, and as it has been proven in research, facebook and twitter posts can specifically target you to brainwash you into accepting a certain reality. With more and more news articles being written by bots it gets less and less labor intensive to do this, and a neural network can specifically adapt to what is neccesary to make you believe in certain things. Fabricating a 'casus belli' against a certain country will become peanuts, and way more believable than the 'Iraq has WMDs' hoax that the Bush administration played.
This is a very dangerous development and I expect it could lead to an orwellian dystopic future if we're not careful. For all we know it could already be here.
|
|
|
|
|
Logged
|
|
|
|
Death 999
Global Moderator
Enlightened
Offline
Gender: 
Posts: 3874
We did. You did. Yes we can. No.
|
> For example scientific research in which the P<0.05 was almost met but the news article still presents it as a confirmed hypothesis
But… that's not… look. If they have evidence which produces an odds ratio of 18:1 that there is an effect versus that there isn't (i.e. not quite p=0.05), and we had a prior of 1:10 against that there's an effect, then we should be tentatively convinced that there's an effect. If we were on the fence, we should be more convinced.
But if they have evidence which produces an odds ratio of 100:1 that there is an effect versus that there isn't (p = 0.01) and we had a prior of 1:1000000 against, we should merely perk our ears and wait for the followup study.
|
|
|
|
|
Logged
|
|
|
|
|
|
Death 999
Global Moderator
Enlightened
Offline
Gender:
Posts: 3874
We did. You did. Yes we can. No.
|
P<0.05 is a convention for significance, and its weaknesses are progressively being revealed, what with the replication crisis in biology, and especially in psychology.
Also, I kind of mis-used the P-value there. It's not exactly an odds ratio because it ONLY considers the odds that chance would produce the effect, not the chance that the competing model would produce the effect (though for a first approximation it's not far off)
|
|
|
|
|
Logged
|
|
|
|
|
|
Death 999
Global Moderator
Enlightened
Offline
Gender:
Posts: 3874
We did. You did. Yes we can. No.
|
Hmm, how to explain that…
For a long time, we used some general guidelines for science as to what is significant or not. As long as we were asking simple questions, they worked. And they still work, more or less, as long as you're asking simple questions. You do an experiment, and there's some randomness involved. If the experiment yields a 'significant' result, that means that if there was no real effect and what you saw was all just random, that would have less than a 5% chance of producing an effect at least as strong as what you saw. Pretty good. Like the chance of a fumble in D&D. Sometimes you get even better than that, and that's great. But they decided that 5% was as much a chance as they would tolerate. Any less than that and it wasn't considered a big deal.
But… what if you ask two questions? What if you ask… twenty? Well, then you expect to have one of those come up as looking 'significant' even just by pure chance. So you need to reduce the tolerance for that error, by that factor (two, or twenty, respectively). Harder to get a result that strong, but it's arithmetically straightforward. But what if you have this experimental procedure where you base the second part of the experiment off the first? You COULD end up asking one of a billion questions, and which one you ask is determined by the first part of the experiment, and the questions overlap considerably? That isn't so simple. You certainly don't need P<1/(20 billion), but it's not 1/20 either. And what data are allowed to be used for the purposes of the analysis is a tricky question too (though if you have the money and time to just do a separate followup experiment it gets simple again)
As a consequence of this getting more complicated, along with some other things like people not publishing/not being able to publish results which did not reach the 'signficance' threshold, results in those parts of science which use significance tests have been skewed. Results which don't reach that magic P < 0.05 are ignored - not even published - even though they contain useful information (especially when that information is, 'this effect definitely does not exist') (this is called the 'file drawer effect'), and results which do meet 'significance', but not by enough to really justify attention given how a priori likely that the effect exists, are published. So, there has been a problem with some kinds of results not ending up being real upon further checking.
Fortunately, we've noticed recently and are trying to change publishing standards so as to better deal with this.
|
|
|
|
|
Logged
|
|
|
|
Scalare
*Many bubbles*
Offline
Posts: 245
|
Thanks Death999. That gives me some more munition in convincing people of certain stances in science.
Are there sites and magazines already working with this new method? If so, which ones? I don't want ot be spoiled by scientific magazines and sites that are more likely to publish only confirmed research.
|
|
|
|
|
Logged
|
|
|
|
|
|
Scalare
*Many bubbles*
Offline
Posts: 245
|
Thanks Death999. That gives me some more munition in convincing people of certain stances in science.
Are there sites and magazines already working with this new method? If so, which ones? I don't want ot be spoiled by scientific magazines and sites that are more likely to publish only confirmed research.
There is no such thing as "confirmed science", only broad scientific consensuses. You are more likely to find science with a broad consensus in university lectures, and/or in Internet courses; such as https://www.khanacademy.org/, https://www.coursera.org/ or https://www.thegreatcourses.com/. The Cochrane Collaboration publish meta-studies that are a bit more trustworthy: http://www.cochrane.org/
Thanks, I'll take a look at those when I get the time
|
|
|
|
|
Logged
|
|
|
|
Death 999
Global Moderator
Enlightened
Offline
Gender:
Posts: 3874
We did. You did. Yes we can. No.
|
There is confirmed science. It's not proven correct, but you can definitely have results which have been confirmed, and theories which have been stringently tested. Even if you get super-technical and philosophical about it, confirmation is a thing. It just can't be absolute confirmation.
|
|
|
|
|
Logged
|
|
|
|
Scalare
*Many bubbles*
Offline
Posts: 245
|
By the way, I think that many conspiracy theorists are just scientists looking for a better truth.
The most clear example I can think of is the flat earth society. They are just trolls. Scientist that troll, trying to provoke other scientists into exploring new theories about the earth and its form and its position in this solar system.
|
|
|
|
|
Logged
|
|
|
|
|
|
|
|
Zanthius
Enlightened
Offline
Posts: 941
|
About intellectual diversification. I also think the generation that is growing up with smartphones now, will have a huge clash with older generations in many third world countries. When we discuss things today, we can immediately check if it is true or not with our smartphones. For example, I was recently discussing mushrooms with 2 people when we had lunch together. They both believed that mushrooms have chloroplasts, and didn't believe me when I said that they don't. However, when I showed them a web page with my smartphone which confirmed that they don't have chloroplasts, they started to believe me. If this had happened before the advent of smartphones, they might still believed that mushrooms have chloroplasts. If people don't believe you, it is much easier to show them proof now.
Also, the older generations have lots of false beliefs, especially in third world countries. Many of these false beliefs can be easily debunked by googling it with a smartphone, and it doesn't necessarily need to be googled by every teenager. It is enough if a few google it and show it to their friends. Things are pretty bad in USA now. Progressive democrats can barely speak to and understand Trump supporters. It is almost as if they are speaking different languages. But the clash with older generations is probably even worse in many third world countries, since the older generations there often are even more conservative and misinformed than in USA.
|
|
|
|
« Last Edit: October 09, 2017, 05:28:23 pm by Zanthius »
|
Logged
|
|
|
|
|
Pages: [1] 2
|
|
|
|
|
