Which Parts of Scientific Explanations are Most Important?

Which Parts of Scientific Explanations are Most Important? Benjamin M. Rottman (benjamin.rottman@yale.edu) Frank C. Keil (frank.keil@yale.edu) Department of Psychology, Yale U., 2 Hillhouse Ave New Haven, CT 06520 Abstract more essential to the writer's purpose and is comprehensible independently of the first half, making it more important. In fact, Marcu found that his algorithm based on which components rely upon others for comprehension predicts peoples' judgments of which components are important for including in a summary. Extrapolating Marcu's (1999) hypothesis about text to explanations, it seems likely that people view components of explanations that are independently comprehensible to be more important than components that are incomprehensible without first knowing the independent component. For example, one must understand the basic concept of global warming (the average temperature of the earth is increasing) before understanding how greenhouse gasses contribute to global warming. Here we say that a component of an explanation, Y, “elaborates” on X when Y provides additional details about X and X must be understood first in order to understand Y. In our studies, we assess whether people identify important components of explanations as those that have the most elaboration – i.e., the most components that depend upon it in order to be understood. The current experiments investigated which components of large scientific explanations people think are most important. Instead of judging importance based on a textual explanation of the sort one might read in an encyclopedia, we were interested in situations when a person learns many facts and develops an explanation by combining the different pieces of knowledge. For example, a person might accumulate bits of related knowledge over time from multiple sources including testimony, direct experience, or even through induction and deduction. When devising an explanation, a person must sort through all this information and decide what is most important. Such a process may occur constantly in our daily experiences as we update our explanatory interpretations of the world, discarding unimportant information and retaining central concepts. Investigating how people identify important information in unstructured explanations is essential for two reasons. First, much of the previous work on importance has focused on intact text which contains many cues to importance (e.g,, topic sentences and conclusions and how sentences are ordered within paragraphs). Studying unstructured explanations can help us isolate structural and content factors that people use to identify import components from purely textual factors. Second, much of the previous research has focused on how people identify important parts of stories. However, narrative stories often follow one individual character on a temporal journey, and much of the work has focused on “story grammar” categories such as Given the depth and breadth of available information, determining which components of an explanation are most important is a crucial process for simplifying learning. Two experiments tested whether people believe that components of an explanation with more elaboration are more important. In Experiment 1, participants gave higher importance scores for components that they judged to be elaborated upon by many other components. In Experiment 2, the amount and type of elaboration was experimentally manipulated. Experiment 2 demonstrated that elaboration increases the importance of the elaborated information by providing insight into understanding the elaborated information; information that was too technical to provide insight into the elaborated component did not increase the importance of the elaborated component. While learning an explanation, people piece together the structure of elaboration relationships between components and use the insight provided by elaboration to identify important components. Keywords: Explanation; Science Education; Summarization Introduction A hallmark of modern culture is dramatically increased access to information. People are often confronted with an overwhelming depth and breadth of information for a given explanation. For example, consider how many pieces of information you have amassed from different sources about global warming. Is it most important to focus on carbon dioxide emissions from humans, sun spots, the ice-albedo positive feedback loop, methane release from peat bogs, or deforestation, etc? In order to limit our search for useful information and decide which information to try to understand and remember, we must focus on the most important components of explanations. Indeed, when learning a textual passage, people focus on the information they judge to be important, which leads to selectively better recall (e.g., Brown & Smiley, 1978; Johnson, 1970). What determines which components people think are most important? One common theoretical position is that information high up in a hierarchical structure of a text represents the most important information (e.g., Thorndyke, 1977). One modern approach from computational linguistics (Marcu, 1999) suggests that important components of a text can be identified by the discourse structure and the rhetorical relations in the text. For example, consider a text about Mars presented by Marcu; With its distant orbit … and slim atmospheric blanket, Mars experiences frigid weather conditions. Marcu argues that the first half of the sentence with... blanket is subordinate to the second half Mars ... conditions. The second half expresses what is