Take a look at the research being done over at Arizona State University. The concept is called "modeling instruction." Essentially students are faced with a typical physics situation or phenomenon but instead of being given a cookie-cutter laboratory packet to explore the phenomenon they are tasked to model the behavior of the thing by recording data and looking for mathematical patterns.Or, from ASU's own description:
For example, in a physics experiment students are asked to develop principles of motion for a pendulum. With the teacher as recorder, students brainstorm about properties of the pendulum that might affect its period. Teacher and students decide which properties should be investigated. In this case they decide to investigate how changes in mass of bob, length of string and amplitude of motion affect the period. Students then work in teams and determine their own procedure for collecting data. After collecting data, they plot it to look for relations among variables and then relate it to equations of motion. Then, in a technique called "whiteboarding", groups present results to the class. The class reaches consensus on an appropriate model to describe the behavior of the pendulum. They do this without being given the answer.Of course I like this approach because it replicates the nature of science in the classroom. Students are essentially forced to work through an actual scientific process and discover patterns on their own without being fed formulas. Further, the research team at ASU has shown this method to be highly successful and as a result many (if not most) of Phoenix area schools use this technique today to teach Physics.
For those interested in learning more, ASU offers summer workshops and a Master's degree program, both of which capture the essentials of modeling instruction.