Principal Investigator:
Ken Ford
Research Categories:
Expertise Studies
Knowledge Modeling and Sharing
Computer-Mediated Learning
Project Description:
The STORM-LK project was intended to illustrate the application of the principles of Human-Centered Computing (HCC). The domain of weather forecasting was selected for study, with the goal of creating a "cognitive prosthesis"–a computational system that leverages and extends human intellectual, perceptual, and collaborative capacities.

Another goal was to analyze a variety of methods of Cognitive Work Analysis: Workplace analysis, workpatterns analysis, the Critical Decision Method, protocol analysis, etc. (see Hoffman, Crandall, & Shadbolt, 1998; Hoffman, & Woods, 2000; Hoffman, Shadbolt, Burton, & Klein, 1995). The multimethod approach permitted the empirical comparison of the methods in terms of their relative efficiency in generating knowledge models and identifying possible leverage points for the application of HCC.

After identifying local knowledge (Gulf Region forecasting) as a good leverage point for the prototyping effort, the weather forecasting skill of experts at the Meteorology and Oceanography Training Facility at Naval Air Station-Pensacola was captured in models of reasoning and models of knowledge. These were then implemented in the form of Concept-Maps using the "CMap Tools" software.

The resulting prototype demonstrates the feasibility of using the CMap system to generate large-scale knowledge models (i.e., models containing dozens of Concept-Maps, thousands of propositions, and hundreds to thousands of multimedia resources) and then use those models to integrate and navigate through the resources, i.e., the various data types used in weather forecasting--satellite images, text, graphics, video, etc.

Because a Concept-Map is basically a graph, it is easy to look at STORM-LK and assume that it is a model of reasoning, or a decision-tree or information flow diagram. It is not. STORM-LK is a model of the knowledge of forecasters. STORM-LK does not impose a sequence of forecasting operations.

Previous attempts to model the reasoning of forecasters have been dissatisfying for two reasons: (1). Because the modeling has been at too general a level, and (2). Because the modeling has been decision-analytic. It assumes, for example, that sequences of operations culminate in a single thing called a "decision," but that is not how forecasters think. Instead, there are interacting cascades of processes including perception, judgment, hypotheses, counterfactual reasoning, etc., all of which are highly contextualized and situation-dependent. In other words, modeling of forecaster reasoning has not worked because the models have fallen at the wrong grain of analysis and have relied on incomplete theories of what actual forecaster reasoning is all about. It is possible to model forecaster reasoning, but such a process mandates the creation of a great many reasoning models, numbering in the dozens (e.g., summertime thunderstorms in the Gulf region, wintertime fog during a La Niña episode, etc.).

Please look at STORM-LK from a different perspective. The purposes of a Concept-Map project such as STORM-LK are:

a). To accelerate the acquisition of expertise by supporting the experienced forecaster's understanding of the dynamics of weather in a region with which the forecaster has had little or no experience. STORM-LK is not a tool for teaching freshmen, even though a Concept-Map project of this sort could be crafted for that purpose, and even though Concept-Mapping as a process is useful in instructional contexts–it forces one to achieve clarity (see Mintzes, Wandersee, & Novak, 1998).

b). To support distance learning–access to weather data for regional forecasting in the context of a set of Concept-Maps that explain the local dynamics. For example, the forecaster in Maine who is transferring to Tallahassee can use the on-line STORM-LK to "talk to" and learn from the local experts. You can "stand on the shoulders" of the expert.

c). To facilitate navigation through data and resources. When you are web browsing using a traditional web interface, you find yourself clicking "back" more than any other clickable, right? In a Concept-Map interface, you can get from anywhere to anywhere more easily, and always in a meaningful context.

d). To support the on-going process of representing and preserving organizational expertise. In STORM-LK, the Concept-Maps are "frozen," but using the C-Map Software Tools one can create new Concept-Maps, move or change links or nodes, etc. Forecasters can continually update and refine their knowledge Concept-Maps. STORM might be used to substitute a new type of "living e-document" for the traditional Local Forecasting Handbooks that are used by both the military and the National Weather Service.

Key Personnel:
Alberto Cañas
Roger Carff
Mary Jo Carnot
John Coffey
Paul Groth
Robert Hoffman
Joseph Novak
Alan Ordway
David Shamma
Jeff Yerkes

Collaborators:
METOC-NASP: Daniel J. Soper, Commander; LCDR Claudia S. Whitney; AGCS Frances N. Arrington; AG1 Trisha A. Bednarczk; AG2 Scott D. Belt; AG3 Dexter S. Berassa; Mr. Kristopher W. Blom; AG1 Terry L. Brightwell; AG2 Sandra L. Brown; AGAA Quincy L. Campbell; AGAA Randall D. Cook; AG1 Richard L. Corkhum; AG3 Deandre J. Ely; Mr. David M. Etheridge; Mr. Paul A. Flores; AG2 Kenneth D. Fowler; AGCM James E. Frodge; AGC Jeffery S. Fulson; AG3 Eric S. Glover; Mr. Howard E. Graham; Mr. Robert W. Hollenbach; FC2 Felix Y. Hotard; AG2 Kenneth E. Klee; LT Matthew P. Lesser; AG3 Naomi M. Liverman; AGAN Heather M. Mathews; AGCS Jerome J. McNulty; AGAA Electa G. Peartree; AG3 Brady M. Peecher; AG1 Gary M. Pelletier; AG2 Jonathan R. Pittman; AGAN Remigio M. Ramos; AG2 Mitzi L. Romero; ET1 Spencer A. Stuart; AG2 Alvin L. Tatum; AG1 Bryan W. Thomas; Ms Kay West; Ms Nancy A. Williams; AG2 Charles E. Wright

Sponsor:
National Technology Alliance