An Opinion on Perception of Probabilistic Forecasts
Cécile Penland
Climate Diagnostics Center
University of Colorado.
Boulder, CO. USA
mcp@cdc.noaa.gov
Let's face it: black and white decisions are the easiest. That's why we like them. There's no judgement involved; we simply employ an "If....Then....Else...." command as we would in a computer program and our decisions are made. This is why we demand a strict definition of El Niño, with an "Is it? Or isn't it?", when there clearly exists a continuum of magnitudes and patterns of sea surface temperatures (SSTs). A bounded set of IndoPacific SSTs, be it ever so continuous, must have its maximum and minimum values, and we call these El Niño and La Niña, respectively, but the terms also apply to a range of values neither so warm nor so cold as the extremes. Where, then, does one draw the line when a natural delineation does not present itself? The same quandary occurs in considering the patterns of SSTs. We know a strong El Niño pattern when we see it, but there are variations of this pattern which may also be considered El Niño, variations which continuously meld into other patterns until we are not quite sure where to say that the El Niño pattern started or stopped. How do we handle this? We may declare that a strict definition is needed, but how does one apply "on-off" logic to a continuous variable? It seems that an equivalent question is as follows: how do we force a physical phenomenon, over which we have little or no control, to fit our methods of contemplation? When faced with the question put this way, it seems easier to change the way we think, particularly if we flatter ourselves that we do have some small control over our own thought processes. But there is a price: changing the way we think about a continuous, complex system requires us to admit the concepts of uncertainty and dynamical probability.
Many scientists claim their willingness to consider probabilistic description of the El Niño phenomenon. The demand for yes-no answers is blamed on the general public, who, it is claimed by scientists and science writers, are incapable of understanding or accepting complex systems or probabilistic forecasts. Is this claim accurate? Well, most people shrug off the impossibility of describing with certainty what any person will do, but they get married anyway. A volatile stock market is still a popular place to invest pension funds. One of the most highly-populated areas in the country is plagued with earthquakes. People seem willing to make generalizations based on statistics and to accept the impossibility of predicting complex systems with infinite accuracy, as long as they are convinced that the system is truly complex, and as long as the decision-making process isn't perceived as "science".
I believe that public perception of scientists as chosen cognoscenti, and scientists' willingness to accept that perception, has done more to hinder the communication between scientists and nonscientists than any other cause. A popular ad for a mortgage broker shows the stereotypical scientist -- white male, portly, fiftyish, turtleneck sweater, white hair and moustache, holding a pipe -- with the caption "You don't have to be a rocket scientist to understand our mortgage programs." Message: "You can't understand what scientists do, but you can understand all of the risks and legal ramifications of what may be the largest, most important financial transaction of your life." No wonder the public is willing to believe itself incapable of understanding any but the simplist scientific concepts! (Side comment on ads: It is interesting how little of the discussion of the hype about El Niño by "the media" separates the news media from the advertising media.)
But, given the evidence that nonscientists make probabilistic choices every day, why do scientists doubt the public ability to accept probability? It is my opinion that physical scientists themselves are uncomfortable with the role probability plays in modern scientific research, and that this discomfort is understandable. After all, most of our knowledge of classical physics comes from a reductionist philosophy of science coupled with carefully controlled experiments. The key word here is "control". An uncontrolled experiment implies an incompetent scientist in the traditional scientific culture, and this cultural wisdom has historically been correct in laboratory experiments where specified uncertainty does not indicate a lack, but rather the extent, of control. But how, for example, can we deterministically describe the collective behavior of molecules in a macroscopic container of gas? We can't, and that is how dynamical probability theory found its way into the physical sciences only one century ago. But the mere necessity of considering dynamical probability indicates a lack of control, somehow, and while most scientists can intellectually accept the necessity for probabilistic dynamical description to some extent, I fear that the emotional difficulties the concept of dynamical probability presents to the traditional scientist keeps him/her from whole-heartedly accepting the concept as a physical phenomenon. It is difficult to accept fully a concept which is emotionally repugnant, and I believe it is this difficulty which leads many physical scientists to consider dynamical probability "hard". If we ourselves have trouble dealing with it, we believe it must be hard for the general public to understand.
I submit that the general public has never been given an adequate opportunity to deal with probabilistic forecasts. Without even considering the effect of the absolutely-certain-sounding interference from the outside, would the probabilistic forecasts of El Niño issued in Australia have had a happier outcome if all forecasts, not just the El Niño forecasts, allowed for a sense of uncertainty? Imagine a television meteorologist (using the metric system): "Well, folks, highs in the low 30s plus or minus about 6 degrees and lows at night in the high teens plus or minus about 4 degrees, so it looks nice and warm. But remember, that's quite a bit of spread!" One cannot help but ask whether or not confused interpretations of probabilistic forecasts aren't a consequence of a simple lack of practice interpreting probabilistic forecasts by the general public, a lack of communication skills by scientists, or a combination of both. After all, there seems to have been little or no backlash in Florida against forecasters who frankly admitted that they weren't sure where or if Hurricane Georges was going to hit the southern coast but provided a range of possibilities (nonscientist friends in Florida, 1998, personal communication). I honestly don't think the communication problem lies with the public. I think the problem lies with the inability we scientists have to come to terms with the fact that there are physical reasons why we must sometimes be unsure, and that it's okay to let other people know it.