Report A Visit With Daryl Bem, Who Found Precognition In the Men Who Stare at Porn

[The Helper]

You may not have met Dr. Daryl Bem yet, but if his findings are true, you may already feel his presence. The energetic emeritus Cornell parapsychologist has attempted to experimentally demonstrate that the human mind can “feel” future events. And in his latest article, he reports that people in the present are probably influenced by, and can predict, events that happen in the future.

At the very least, his research says, we are able to predict pornography. Really. But more on that in a moment.

Before you confuse Bem’s research with past parapsychology hoaxes such as Project Alpha, or the experiments in “The Men Who Stare at Goats,” keep in mind that the article (pdf) is set to appear this month in the Journal of Personality and Social Psychology, one of the field’s premier clearinghouses. “The article has passed review by four scientific experts,” Bem says.

His experiments test for precognition—the ability of people to perceive events in the future. In one of nine such experiments conducted on more than 1000 student subjects, a pornographic image would randomly appear on the left or right hand side of a computer screen. The results indicate that people could predict where the pornographic image would appear—even before the computer made its random decision.

Daryl Bem Proved ESP Is Real. Which Means Science Is Broken.

The scientist couldn't have foreseen the crisis his research would touch off.

slate.com

 

[sqrage]

And this is why I hate social psychology or any psychological fields that are not based on hard physical evidence.

Yea... the fact that they guessed 3% more than chance is really enough to say people can see into the future before the future is even generated by a randomizing algorithm. Right.

 

[Slapshot136]

And this is why I hate social psychology or an psychological fields that are not based on hard psychical evidence.

Yea... the fact that they guessed 3% more than chance is really enough to say people can see into the future before the future is even generated by a randomizing algorithm. Right.

I more or less agree with you, especially since it hasn't been replicated - I don't buy his explanation of why it wouldn't work when replicated as long as the control group never heard of it

 

[Sim]

Money has been wasted. Nothing to see here, move along.

 

[uberfoop]

And this is why I hate social psychology or an psychological fields that are not based on hard psychical evidence.

Yea... the fact that they guessed 3% more than chance is really enough to say people can see into the future before the future is even generated by a randomizing algorithm. Right.

In a well-controlled test with a massive sample size, a 3% offset could have extreme statistical significance.


Of course, if everyone has failed to replicate it, which would not be a particularly surprising development...

 

[sqrage]

>>In a well-controlled test with a massive sample size, a 3% offset could have extreme statistical significance.

I'd think of massive to be in the millions before I'd count it as significant. Not 1000 or less which is often the case with these experiments.

 

[uberfoop]

>>In a well-controlled test with a massive sample size, a 3% offset could have extreme statistical significance.

I'd think of massive to be in the millions before I'd count it as significant. Not 1000 or less which is often the case with these experiments.

I don't actually have my stats book on me right now and I'm kind of rusty, but I wouldn't be surprised if a few thousand students or so, particularly with a large amount of testing with each one, could assemble bounds of good confidence (I dunno, maybe 90% or so) around 53% that excludes 50%, if the trend exists in the data.
Statistics isn't really something where you take more data because your gut feeling says you should; there are rigorous ways to analyze confidence bounds. If the guy had the guts to send this in to a rigorous publisher, rigorous bounds analysis would be necessary, else they would have tossed the paper immediately.

Thus, since I very strongly doubt that the discovered correlation is real for logical reasons, I'd personally wager there's a systematic error with the experiment itself rather than a data analysis one, and the real issue is whether or not the results are explainable and/or reproduceable, both of which I strongly doubt.


edit: On the other hand, I may of course be completely wrong, and this sort of binomialish test actually results in substantial variance, and the sample size is absurdly small, and I just have too much confidence in people not being stupid. It's also possible that the situation due to something about the setup becomes mathematically complicated for confidence bound determining, and if this is the case, an article might make it a good ways before being called out for doing stats wrong. I guess we just don't have enough information unless one of us bothers looking into it.

 

[xPass]

What's algorithm? and what's replicate? lol I'm noob...

 

[UnknowVector]

An algorithm is a rigidly defined process for doing something.

"In mathematics, computer science, and related subjects, an algorithm (derived from the name of mathematician al-Khw?rizm? and transformed to match the Greek "arithmos"-number) is an effective method for solving a problem expressed as a finite sequence of steps." (http://en.wikipedia.org/wiki/Algorithm)

That definition seems pretty good to me. Though the term "algorithm" is often applied without a specific problem in mind. E.g., ray tracing encompasses a bunch of related algorithms, and they solve a bunch of related problems, but there isn't any _one_ problem that they solve.

To replicate is to repeat. In an experiment, the larger the sample size the more the results tell you, because chance plays a smaller role in determining the final outcome. Replicating an experiment is in part like increasing the sample size, but it also gives a chance for different scientists in different labs using different equipment to do the same thing. Maybe the original scientist didn't realize that the iron in his lab had been magnetized by the electrical engineering department, but the next guy who does the experiment might notice. Or maybe the next guy is in a completely different lab.

 

[phyrex1an]

I don't actually have my stats book on me right now and I'm kind of rusty, but I wouldn't be surprised if a few thousand students or so, particularly with a large amount of testing with each one, could assemble bounds of good confidence (I dunno, maybe 90% or so) around 53% that excludes 50%, if the trend exists in the data.
Statistics isn't really something where you take more data because your gut feeling says you should; there are rigorous ways to analyze confidence bounds. If the guy had the guts to send this in to a rigorous publisher, rigorous bounds analysis would be necessary, else they would have tossed the paper immediately.

Thus, since I very strongly doubt that the discovered correlation is real for logical reasons, I'd personally wager there's a systematic error with the experiment itself rather than a data analysis one, and the real issue is whether or not the results are explainable and/or reproduceable, both of which I strongly doubt.

Another thing to note here is that studies for "precognition" have been done to death. Statistically, it's not surprising if one of them gets a statistically significant result. But of course, stating that after the fact isn't very statistically valid