Twitter science - June 01, 2009
News of 'the world's first scientific experiment on Twitter' has received some media pickup today. The trial starts tomorrow, conducted by psychologist Richard Wiseman together with New Scientist.
Wiseman, who works at the University of Hertfordshire, UK, says the study will examine the psychic ability of 'remote viewing' - the idea that one can see an object without being shown or told what it is. He'll tweet from a location, then offer five photo options; tweeters have to guess which of the five he is actually in. The trial will be repeated four times using different locations (3pm, Tuesday to Friday this week).
Although rigorous scientific experiments have been conducted into remote viewing before (with no evidence found for the ability), this isn't one of them. The choice of Twitter pretty much guarantees that, as Wiseman readily concedes in conversation, though not in his quoted press comments.
He says that if the majority of people pick the right location, then the trial will count as a hit; "three or more hits in four trials will be seen as supporting the existence of extrasensory perception".
The odds of that success rate happening by chance, if each tweeter’s vote were independent of all the others, are about 1 in 37, or 2.7% [*]. So if such a thing happened, it would be unlikely enough to drop into the ballpark of what statisticians call a 'significant' result.
But (as Wiseman explains) tweeters' votes are not going to be independent. For example, they'll tend to cluster, or stack, on what others start tweeting (tweets can be viewed with twitterfall); or on common location ideas held in people's minds (such as a sunny grassy area, given the UK weather recently).
Given that psychologists often design experiments to test something other than the professed trial - some bloggers have wondered whether Wiseman is actually hoping to check out something else. Let the fun begin ... don't confuse it with a scientific experiment for remote sensing, though.
[*] Here's the maths. Chance of a 'hit' in one trial, the right location getting more votes than any other location, is 1/5 (with a large number of voters). Over four trials, chance of at least three 'hits' is 4*(1/5)^4 + (1/5)^4. (PA quoted Wiseman as stating that the odds were 1 in 125 - but he later tweeted that he gave that figure in answer to a different question).
