Strange bedfellows? - Meteorology and Social Science

I recently gave a talk at the National Weather Association's Annual Meeting in OKC on the role for social science in the weather forecasting business.  This is something I've long been saying is needed, as a result of my long friendship with the late Al Moller.  It was Al who first made it clear to me that just putting out a good forecast is only the start of a chain of events that must take place if the forecast is to be of any value to the users.  By the way, I should give credit to the late Allan Murphy for his insight that the value of a forecast is never determined by the forecaster, but by the user of the forecast.

Anyway, I've presented this concept in several talks, usually with a tweak or two based on my most recent thoughts regarding the topic.  For a forecast to be effective, it must be:

1.  Honest
2.  Accurate
3.  Received
4.  Understood
5.  Believed
6.  Helpful in making a decision

Only the first two items are under the control of forecasters.  Once a forecaster transmits a forecast, the rest of the steps depend on others.  Forecasters are paid to forecast, not to do all these other things.  Meteorology is what they are educated and trained to do.  We shouldn't expect them to become social scientists, as well!  Moreover, most forecasters know little or nothing about how to go about helping the users make good decisions.  Note that most users have to account for many other factors that go into making a decision besides the weather, and most forecasters know little or nothing about those other factors most of the time.  Weather information is just a part of the decision-making process.  The thing is that progress in forecasting is written in to the system, so while this blog is mostly about social sciences, the meteorology will be forging ahead, without doubt.

The existing watch-warning system, which began to take form in the mid-1960s, has saved many lives.  It's not a "broken" system, despite the fact that it likely is far from perfect.  I think I can justifiably assert that no one actually knows just how effective it is, but there are obvious declines in fatality and injury rates that seem to justify the existing system.  If we intend to change it, let's follow the famous dictum (often erroneously attributed to the Hippocratic Oath): First, do no harm!  Don't make stupid bureaucratic decisions in haste, just for the sake of doing something, without first having a clear picture of the shortcomings of the existing system and having a proposed change that has been given something like peer review - some sort of vetting process that includes participation by both forecasters and users, who after all are the ones with the most at stake in any changes.

We meteorologists need help from social scientists: experts in communications, psychology, economics, etc., where we have no expertise.  We need participation from them so that any studies that review the current situation and any proposals for change will take human factors properly into account.  We need surveys of what the public knows and actually does under the current system, over a very broad spectrum of users, since "the public" is pretty far from a monolithic block.  Ideas for changes need to be given thorough testing to see if they actual improve upon the existing system by helping users make their own decisions.  We do not necessarily need to be telling users what to do!!  What we really need is to learn how to make our products more effective at helping users make good decisions with the information we can realistically provide for them (including uncertainty information!)

Although the movement to get social science into meteorology has been percolating for quite some time, and has become something of an "in" thing to advocate, what's been absent is much real collaboration between meteorologists and social scientists to produce actionable results.  We don't need more conferences, workshops, and other "feel good" exercises.  We need folks to roll up their sleeves and start getting some useful results to provide a scientific basis upon which to move forward.  No more kumbaya songs around the campfire, please.  Let's make something real and substantive, not just endless palaver.  When we make changes, they need to be tested to make sure they're doing what we wanted them to do.  The process should be one of never-ending evaluation and revision (see below).

If social science is to have a role in weather forecasting, and I think it should, then what might it look like?  Does it make sense to have a token social scientist in every forecast office?  I think that makes little sense.  What about a Social Science Center, comparable to the Storm Prediction Center?  I think the NWS will have heartburn in setting up something like this.  What do they know about the skills needed to make it be effective?  How would they know to pick the right people and what resources to provide for them?  I think the best path for integrating social science into weather forecasting is for the NWS to have a budget that includes the funding specifically designated to support ad hoc collaborative efforts with social scientists, who would stay with their own institutions but form working partnerships with meteorologists  to answer specific questions.  This gives the most flexibility and avoids the creation of isolated "lone wolves" or some bureaucratic agency that would be a stranger in the strange land of weather forecasting.

Further, the NWS needs to understand this effort is not some sort of one-time project.  It must be a continuous process, because the social, cultural, technological, and meteorological landscapes are constantly changing.  We shouldn't have to wait decades for it to become painfully clear the system needs to evolve in the face of change.  However, it will take some time to gather data about what works and what doesn't work with the current system, and even more time to develop and test proposed changes.  Finally, before implementing anything, there should be an extensive public education campaign to help users understand the forthcoming changes.  Let's not make the same mistake we made when probability of precipitation (PoP) was implemented in the mid-1960s.  We're still paying the price in credibility for that one!

If this is done right, it will be a boon to weather forecasting and our whole society will reap the benefits.  Please, let's not screw it up this time!