One of the major issues that arises when examining the connections between science and policy are the ways information is framed. You can say that the rate of skin cancer caused by a particular phenomenon has increased from one in ten million cases to one in a million cases. You can say that the rate has increased tenfold, or that it has gone up by 1000%. Finally, you could say that an individual’s chances of getting skin cancer from this source have gone up from one tiny figure to a larger, but still tiny seeming, figure. People seem to perceive the risks involved in each presentation differently, and people pushing for one policy or another can manipulate that. This can be especially true when the situations being described are of not comparably rare: having your chances of being killed through domestic violence reduced 1% is a much greater absolute reduction than having your chances of dying in a terrorist attack reduced by 90%.
When talking about presentation of information, graphs are an important case. Normally, they are a great boon to understanding. A row of figures means very little to most people, but a graph provides a wealth of comprehensible information. You can see if there is a trend, what direction it is in, and approximately how strong it is. The right sort of graph, properly presented, can immediately illuminate the meaning of a dataset. Likewise, it can provide a compelling argument: at least, between those who disagree more about what is going on than how it would be appropriate to respond to different situations.
People see patterns intuitively, though sometimes they see order in chaos (the man on the moon, images of the Virgin Mary in cheese sandwiches). Even better, they have an automatic grasp of calculus. People who couldn’t tell you a thing about concavity and the second derivative can immediately see when a slope is upwards and growing ever steeper: likewise, one where something is increasing or decreasing, but at a decreasing rate. They can see what trends will level off, and which ones will explode off the scale. My post on global warming damage curves illustrates this.
Naturally, it is possible to use graphs in a manipulative way. You can tweak the scale, use a broken scale, or use a logarithmic scale without making clear what that means. You can position pie charts so that one part or another is emphasized, as well as abuse colour and three dimensional effects. That said, the advantages of graphs clearly outweigh the risks.
It is interesting to note how central a role one graph seems to have played in the debate about CFCs and ozone: the one of the concentration of chlorine in the stratosphere. Since that is what CFCs break down to produce, and that is what causes the breakdown of ozone, the concentration is clearly important. The graph clearly showing that concentrations would continue to rise, even under the original Montreal Protocol, seems to have had a big impact on the two rounds of further tightening. Perhaps the graph used so prominently in Al Gore in An Inconvenient Truth (the trends on display literally dwarfing him) will eventually have a similar effect.
Stats in recent personal experience
My six-month old Etymotic ER6i headphones are being returned to manufacturer tomorrow, because of the problems with the connector I reported earlier. Really not something you expect for such a premium product, but I suppose there are always going to be some defects that arise in a manufacturing process. Of course, being without good noise isolating headphones for the time it will take them to be shipped to the US, repaired or replaced, and returned means that reading in coffee shops is not a possibility. Their advantage over libraries only exists when you are capable of excluding the great majority of outside noise and of drowning the rest in suitable music.
Speaking of trends, I do wonder why so many of my electronics seem to run into problems. I think this is due to a host of selection effects. I (a) have more electronics than most people (b) use them a great deal (c) know how they are meant to work (d) know what sort of warranties they have and for how long (e) treat them so carefully that manufacturers can never claim they were abused (f) maintain a willingness to return defective products, as many times as is necessary and possible under the warranty. Given all that, it is not surprising that my own experience with electronics failing and being replaced under warranty is a lot greater than what you might estimate the background rate of such activity to be.
Two other considerations are also relevant. It is cheaper for manufacturers to rely upon consumers to test whether a particular item is defective, especially since some consumers will lose the item, abuse it, or simply not bother to return it even if defective. Secondly, it is almost always cheaper to simply replace consumer electronics to fix them, because of the economies of scale involved in either activity. From one perspective, it seems wasteful. From another, it seems the more frugal option. A bit of a paradox, really.
[14 March 2007] My replacement Etymotic headphones arrived today. Reading in coffee shops is possible again, and none too soon.