In Praise of Messy Data: Lessons from the Search for Alien Worlds

The Next-Generation Science Standards emphasize the importance of teaching the practices of science alongside content ideas and crosscutting concepts (NGSS Lead States 2013). Chief among these practices is the ability to gather, assess, analyze, and interpret data. Authentic inquiry near the leading-edge of science offers a wonderful opportunity for students to have ownership of their data sets and to personalize the learning experience. While many teachers welcome authentic inquiry in their classrooms, they know that real-world data are often messy, in contrast to the picture-perfect graphs and data typical of a textbook illustration. Do messy data confuse students and obscure the point of a lesson? [FIGURE 1 OMITTED] That was our initial concern when we developed the Laboratory for the Study of Exoplanets (ExoLab), an online astronomical laboratory designed to increase students' data literacy, while engaging them in the search for habitable worlds and life beyond Earth (Gould et al. 2012). What we found is that messy data help students think more deeply about data. The ExoLab is aimed at high-school classrooms in physics, astronomy, and Earth science and has been field-tested with several thousand students in more than a dozen states. The resource is freely available online (see On the web). In the project, students can detect up to two dozen of the known or suspected transiting exoplanets that periodically pass in front of their stars and eclipse some of their star's light. By looking for a dimming of the starlight, students can indirectly detect the planet. From the data they gather, they can determine the size of the planet, how close it comes to its star, and even whether its orbital plane is tilted as seen from Earth. Before using the telescopes accessible via the website, students are first asked to reason from an interactive, 3-D model and then to predict what a graph of a star's brightness would look like over the course of one of these eclipses. Their predictive model consists of three parts: the 3-D simulation; their graphical representation of an eclipse; and their written description of how a star's apparent brightness should vary over time (Figure 1). To make the observations, students instruct the ExoLab's automated telescopes to take many night-time images of their target star. Students select their target star and observing times from a preset menu; typically, one or more exoplanets are detectable each night. The telescopes--located at the Smithsonian's Whipple Observatory near Amado, Arizona--swing into action at the appointed times, pointing to the star, taking the requested images, and posting them to each student's online account for use the next morning. For each image, students carefully measure the brightness of the star. As the planet passes in front of the star, it blocks some of the star's light, producing an apparent dimming that lasts an hour or two. As students graph the star's brightness versus time, they look for the telltale dip that signifies a distant planet eclipsing its star. Just a few years ago, detecting an exoplanet would have been worthy of a Nobel Prize; today, students can detect these distant worlds as a homework assignment! Initially, we hoped that students' data would be clear and crisp, revealing with textbook clarity the dip in brightness that signifies a planet. But we have come to see that real learning comes when students confront messy data, for several reasons: Messy data invite students to become data detectives An example of one school's data is shown in Figure 2. Each point represents a student's measurement of the star's brightness in one image at one time of night. If the data were ideal, the dots would align in a regular pattern that dipped when the exoplanet passed between the Earth and the star, similar to the prediction shown in Figure 1. In fact, a dip seems to be visible, but the data points certainly don't fall on the nice line envisioned by students. …