Thursday, February 27, 2014

Logical Fallacies - Cherry Picking

Next up on Uraniborg's Logical Fallacy Series: Cherry Picking.  Cherry picking is related to confirmation bias, which I wrote about earlier, but it is more likely to be done intentionally.  The idea is that you conveniently focus on evidence that supports your hypothesis and hide or ignore the rest.  Confirmation bias tends to rely more on anecdotal evidence while cherry picking is useful when you have lots of concrete data; however, both scientists and non-scientists can fall prey to either one within and without of the lab setting.

Why is cherry picking dangerous in addition to being unethical?  Consider a bill passed in the North Carolina state senate back in June of 2012 mandating that all climate change data measuring the change in sea level had to come from historical data after the year 1900.  This demand throws out data from before the industrial revolution so we can't see what impact the introduction of fossil fuel based technology had on the environment.  The law also requires that predictions of future sea levels are based on linear extrapolations from the historical data.  No exponential extrapolations allowed even though they're more accurate.  The result is a legislated prediction of sea level rise set at 39.6 centimeters (15.6 inches) in the next century. Other groups like the Centre for Australian Weather and Climate Research predict a rise of 50 cm to 100 cm (19.7 in to 39.3 in) over the next 100 years.

North Carolina could spend its time preparing for an increase in sea levels, writing legislation that might slow down the change, but instead it deliberately chose to look only at data that is convenient for themselves. You can't willfully ignore reality.  Sea levels will change regardless of political and economic wishes.  Cherry picking ensures that the people getting their information from you will be misinformed and ill-prepared.  Don't be lazy - take the time to look at the whole picture.

If you're on the receiving side of the data, learning statistics can help you find out when you're being tricked. Look at the stats presented to you carefully and ask for relevant omitted information.  After all, 100% of people that have read this article as of  February 26, 2014 concur that a little bit of statistics can be a dangerous thing.