Epiphany comments on Open Thread, October 16-31, 2012 - Less Wrong

5 Post author: OpenThreadGuy 16 October 2012 10:43PM

You are viewing a comment permalink. View the original post to see all comments and the full post content.

Comments (271)

You are viewing a single comment's thread. Show more comments above.

Comment author: Epiphany 28 October 2012 08:27:36AM *  5 points [-]

I'm having a pretty intense reaction to reading certain articles and could use some support or a solution:

Here's what I read and my reactions:

  1. Feynman's Cargo Cult Science (Which is about how a lot of scientific studies are done badly, often due to researchers not being allowed to do the research correctly.)

  2. The PLOS Medicine article "Why Most Published Research Findings Are False"

  3. An article about how psychologists aren't usually using the treatments most supported by science which links to a document that contains a horrifying account:

"During many of my 20 years at Stanford University, Albert Bandura and I tried to hold on to a science-based clinical training program. The bizarre situation we faced there is of more than personal and historical interest: I suspect that many of the same conflicts still exist and motivate the efforts described by Baker and colleagues. Bandura and I, and our students and other colleagues, were discovering the remarkable discrepancies between what the scientific work was revealing and the requirements imposed by the pressures for maintaining accreditation. The professional accreditation requirements insisted on continuing practices whose value was contradicted by the empirical findings. Those requirements not only flew in the face of the data but also made enormous demands on faculty and student time in the clinical program."

http://www.psychologicalscience.org/journals/pspi/inpress/baker.pdf

I'm having a variety of reactions:

  1. What meaning is there in doing anything (being a doctor or a psychologist for instance... or any number of other professions) if we can't even trust the research or the schooling? How can I make a difference in the world or do anything useful with no real knowledge? How do you find meaning, LessWrong?

  2. Thank goodness I found this place. I am in love with the glimmers of sanity I see here. Before I found LessWrong I was just kind of... "WTF humanity is a mess." Now it's more like "WTF humanity is a mess but at least there's a group of people trying not to be." If anyone is up to describing this wonderful and horrible feeling in their own words, I could really use to feel related to about this.

  3. Do you know of a website where one can look up a piece of research to see what flaws it has? Is one planned? I need this because it would take a very long time for me to read enough on each relevant topic to discover whether a piece of research I want to use is flawed or not. For instance, Feynman explained about how lots of studies have been done with mazes and rats, but people didn't seem to realize that the rats were using methods to find the food that were unexpected and all sorts of stuff has to be controlled for ranging from the scent of food to the type of flooring in the maze. If you don't know that all of these things need to be controlled for, you won't know that the vast majority of studies done on putting rats into mazes are useless. It's simply not realistic to expect ourselves to be able to single handedly give every single study we read a thorough enough review to detect all the flaws. I love research, but I now feel that it's futile. Does anyone know a solution? I know that peer reviewed journals are supposed to address this type of problem, but I don't see the online studies that I find being rated or marked as flawed in an obvious way.

Comment author: Epiphany 03 November 2012 12:39:51AM *  0 points [-]

suddenly thinks of a coping strategy

Wikipedia addresses this... I was just reading the wiki on the Paleo diet and saw a bunch of stuff about repeatability and study relevance like:

Loren Cordain, a proponent of a low-carbohydrate Paleolithic diet, responded to the U.S. News ranking, stating that their "conclusions are erroneous and misleading" and pointing out that "five studies, four since 2007, have experimentally tested contemporary versions of ancestral human diets and have found them to be superior to Mediterranean diets, diabetic diets and typical western diets in regards to weight loss, cardiovascular disease risk factors and risk factors for type 2 diabetes."[27] The editors of U.S. News replied that their ranking included a review of all five studies which found that all of them were small and/or of short duration.

I realize Wikipedia isn't credible for citing or anything but I feel heartened because:

  • I bet they often link to a credible meta-analysis, making it easier to find them (I've been told by Gwern that one way of coping with this is to read a meta-analysis because it gives you a number of advantages over reading individual pieces of research).

  • It serves as a method for finding out about some of the flaws you need to look for when reading studies on the topic.

  • It often lists a collection of relevant research, which can save time.

  • It might be a good starting point for creating your own thorough reviews of studies because a lot of things will already have been hashed out, so it's just a matter of verifying that what's there is correct, which should save time if you build on it.

Hm...

Wikipedia is not a perfect solution but I think this will help me cope.

.oO I wonder if there are features that could be added to Wikipedia that would encourage the entries to transform into credible meta-analyses...

Comment author: gwern 03 November 2012 01:29:06AM 1 point [-]

A very good Wikipedia article will be equivalent to a review article, but such an article isn't a meta-analysis: it doesn't include only studies which can be boiled down to a few summary statistics like d. There's also little way of being sure that the article is comprehensive and unbiased - one reason meta-analyses usually make a point of how they did a big search on Pubmed and looked through hundreds of results etc.

I don't know what features could be added to deal with either problem. Any meta-analyses tucked into WP articles would be rightly considered Original Research.