Emergece Project (Chi)
Learning about Emergence and Heat Transfer
Michelene T. H. Chi
|PIs||Michelene T. H. Chi|
|Other Contributers|| David Yaron (Department of Chemistry, CMU))
|Study Start Date||May 2008|
|Study End Date||October 2008|
|Number of Students|
|Total Participant Hours||TBD|
The majority of learning studied within PSLC has been of the enrichment kind (Carey, 1991). Enrichment occurs when a student has missing knowledge components, and learning consists of adding knowledge components. Many PSLC studies focus on optimal methods and conditions for adding knowledge components that result in robust learning. However, there is a completely different kind of learning that few PSLC studies have addressed -— learning of the conceptual change kind. Conceptual change occurs when a student’s existing ideas conflict with to-be-learned material, usually in science domains (Vosniadou, 2004). It is customary to assume that the existing knowledge is incorrect or misconceived and that the to-be-learned information is correct by some normative standard. Although conceptual change can perhaps be modeled using canonical PSLC mechanisms of knowledge component construction, strengthening, and feature refinement, such a model would fail to explain why these mechanisms do not work when existing knowledge conflicts with new knowledge. Nor would it suggest how to alter the situation so that these mechanisms will work. The failure to achieve conceptual change in the face of robust misconceptions has been a problem for decades. The proposed study would extend the learning framework of PSLC by incorporating new information about conceptual change.
The purpose of the proposed study is to test an intervention designed to promote deep understanding of science processes about which many students hold robust misconceptions. The intervention, based on our previous research, involves introducing an unfamiliar type of process, called emergent, and differentiating it from a familiar type of process, called direct. This introduction and differentiation, called schema training, will occur in the context of a diffusion lesson in a college-level chemistry course. Our study will determine whether students who receive the schema training understand subsequent instruction on heat transfer better than students who do not receive the training.
Background & Significance
For years, researchers have been investigating student ideas about science concepts and how they interfere with learning. Some of these ideas, often referred to as misconceptions, are extremely robust, stable and resistant to instruction. Thousands of studies have documented specific misconceptions about a variety of science concepts and phenomena, beginning with a book by Novak (1977) and a review by Driver and Easley (1978), both published almost three decades ago. By 2004, there were over 6000 publications describing student ideas and instructional attempts to change them (Duit, 2004). Obviously, we have not yet solved the daunting problem of how to build conceptual understanding in the presence of robust misconceptions, a process sometimes referred to as radical conceptual change (Carey, 1985).
Our lab has been investigating the underlying structure of a class of science process concepts that are frequently misconceived across various science domains. Based on insights from that analysis, we hypothesize that emergence is the key structure that underlies this class of process concepts, and that misconceptions arise because most students know nothing about emergence. To test our hypothesis, we developed a brief, self-contained domain-general instructional module that we refer to as emergent schema training. It focuses on the idea of emergence, which is related to notions of complex dynamic systems (Goldstone, 2006), to see if it can help students achieve greater learning, deeper understanding, and more successful transfer to other concepts. Our domain-general approach is relevant to a variety of concepts across science domains as well as to concepts in domains other than science, such as economics, political science, and history. Thus, the structure of emergence is an ideal test bed for robust learning not only in terms of long-term retention, but also for accelerated learning of new concepts.
The long-term objective of our emergent work is to demonstrate that schema training can help students achieve radical conceptual change and, more importantly, transfer deep understanding to a variety of important concepts. Toward that end, we developed an emergent schema-training module for middle- and high-school level students. In 2005, we tested the module with eighth- and ninth graders, obtaining promising results. (See Prior Work below for details.)
The objective of the research proposed in this project plan is to replicate the 2005 study with college-level students in vivo, for reasons that are especially germane to the conception of LearnLabs. That is, we achieved only modest far transfer in our 2005 study primarily because the students did not possess the prerequisite knowledge needed for learning the transfer concept. A LearnLab context will completely avoid this problem. Our study will parallel and collaborate with an NSF-funded study being conducted with engineering students at the Colorado School of Mines, Purdue University and three other sites, by Ron Miller (a chemical engineer), Ruth Streveler (an educational psychologist), and Jim Slotta (an educational technologist). They have obtained NSF funding to test a modified version of our schema training method for helping college students develop accurate mental models of engineering processes. Our proposed project will develop materials for college students, and they will then use those materials to determine whether emergent schema training can help engineering students understand small-scale dynamic processes, such as microfluidics. A summary of the project is attached as an ancillary document. Taken together, these two studies will provide a rigorous test of our hypothesis about emergence and the value of schema training.
Through four years of funding from the Spencer Foundation, we conducted theoretical analyses of misconceptions and concluded that there are three types of conceptual change, based on the grain size of knowledge components: belief revision, mental model transformation, and categorical shift (Chi, in press). We contend that science misconceptions that are extremely robust require a categorical shift in order to build understanding that is equally robust. An example of a category shift is to reassign the concept whale from the category fish to the category mammal. This kind of shift is not difficult to achieve because most students are familiar with the category mammal. However, to understand many science processes, students need to shift their misconceived understanding from a familiar category, direct process, to an unfamiliar category, emergent process. Because the concept of an emergent process is often missing from student knowledge bases, we need to help them build understanding of the category before they can achieve a successful category shift. Once the emergence category is in place, appropriate science processes can be assimilated or added, using the PSLC mechanisms of construction, strengthening, and feature refinement.
Briefly, a direct (or indirect) process is one that usually has a causal agent or group that produces some outcome in a sequential and dependent sort of way. We will describe an everyday example, a less familiar example, and a scientific example.
Direct example 1. In the familiar process of a baseball game, the final outcome might be explained as being due to the excellent work of the pitcher (thus attributing the outcome to a single or multiple agents), or to some local events within the game (such as home runs). Moreover, the behavior of certain local events within the game corresponds to or aligns with the outcome. For example, a team with many home runs in a game is more likely to win. That is, home runs are positive local events and they align with the positive outcome of winning the game.
Direct example 2. A slightly less familiar example is seeing multiple airplanes flying in a V-formation. This V-pattern is intentional, created by the lead pilot (a single agent) telling the other pilots where to fly.
Direct example 3. A direct process from biology is cell division, which proceeds through a sequence of three stages. The first, interphase, is a period of cell growth. This is followed by mitosis, the division of the cell nucleus, and then cytokinesis, the division of the cytoplasm of a parent cell into two daughter cells. Such a process has a definite sequence, in which some events cannot occur until others are completed.
In contrast, emergent processes have neither a causal agent or agents nor an identifiable sequence of stages. Rather, the outcome results from the collective and simultaneous interactions of all local events. Let’s consider three examples here as well: an everyday example, a less familiar example, and a scientific example.
Emergent example 1. The process of a crowd forming a bottleneck at a door, as when the school bell rings and students hurry to get through the narrow classroom door, is an everyday example of an emergent process. Although there is an external trigger (the school bell), the outcome of forming the bottleneck cannot be attributed to any single agent or group of agents, and the process is not sequential. Instead, all the students simultaneously rush toward the door at about the same pace.
Emergent example 2. A slightly less familiar example is migrating geese flying in a V-formation. In contrast to the airplane example, the V-pattern is not intentionally formed, such as by the lead goose telling other geese where to fly. Instead, all the geese are doing the same thing, flying slightly behind another goose because instinctually they seek the area of least resistance. When all the geese do the same thing at the same time, a V-pattern emerges.
Emergent example 3. An example from biology is the diffusion of oxygen from the lungs to the blood vessels. This process is caused by all the oxygen and carbon dioxide molecules moving and colliding randomly. Because there are more oxygen molecules in the lungs than in the blood, more oxygen molecules tend to move from the lungs to the blood than from the blood to the lungs. The reverse is true for carbon dioxide molecules. Since all molecules move and collide randomly, both kinds of molecules move in both directions. Some oxygen molecules move from the blood to the lungs, and some carbon dioxide molecules move from the lungs to the blood. Despite local variations, however, the majority of oxygen molecules move from the lungs to the blood, and the majority of carbon dioxide molecules move in the opposite direction, creating an unintentional uni-directional flow.
These two sets of examples illustrate general differences between direct and emergent processes. A more detailed description that explicitly specifies a set of ten distinguishing features is provided in Chi (2005), and see also Table 2 at the end of this project plan. Explicit specification of these features is essential for purposes of instruction.
Our prior funding allowed us to develop two equal-length sets of instructional materials for middle school science, one for an experimental group and one for a control group. Each set consisted of one general module and two concept-specific modules. The general module for the experimental group introduced the emergence schema, describing everyday examples of emergent and direct processes and identifying explicit features for recognizing and explaining each type. The general module for the control group described the nature of science, using text taken directly from Science for All Americans, published by Project 2061 (AAAS, 1989). Pre- and post assessments of this module measured long-term retention because the posttest was administered several days after learning.
For both groups, the first concept-specific module was about diffusion. The diffusion modules for both the experimental and control groups included a text description and computer animations of diffusion. The text for the control group was a composite of passages taken from standard grade-appropriate textbooks. The same text was used for the experimental condition, with a small number of passages ablated and replaced by descriptions instantiating diffusion as an emergent process, so that they were about the same length.
The animations of the diffusion module consisted of dynamic simulations at both the macro (outwardly visible) and micro (molecular) levels. Although it has been found that in general, static diagrams are as supportive as animations when the to-be-learned materials can be visualized by students (Hegarty, Kriz, & Cate, 2004), in studying emergent processes, however, dynamic simulations are more helpful, since the emergence of the macro-level pattern from the micro-level interactions is not usually visualizable. Although both the experimental and control diffusion modules included the same simulations, the two modules differed in how the simulations were implemented. For the experimental group, the simulation was embedded with emergent “mapping prompts,” which are queries that focused on specific emergent properties depicted in the simulations. In essence, emergent mapping prompts helped students map diffusion to emergence. For the control group, the simulations were embedded with “attention prompts,” which were queries that focused student’s attention on similar aspects of the simulation, but without explicit mapping to emergence. The two modules contained the same number of questions. Pre- and post assessment of the diffusion module measured transfer.
The second concept-specific module, identical for both groups, was about heat transfer. Like the diffusion module, the heat module included animations of heat transfer at both the macro and micro levels. Descriptive text and questions were taken directly from a popular high school textbook, Conceptual Physical Science (Hewitt, 1987), without any modifications. Pre- and post assessment of the heat module measured preparation for future learning.
The modules were used in a lab study with 41 eighth and ninth graders, who were randomly assigned to either the experimental or control condition. The study was carried out over four consecutive days after classes in a private school. All instructional and assessment materials were delivered on-line, and students worked individually at their own laptops. Results from the diffusion assessment showed that not only did both groups gain significantly from pre- to posttest, the gains of the experimental group far exceeded those of the control group (ANCOVA, F(1,38) = 9.391, p = .004, with effect sizes of 1.4 for the control group and 1.6 for the experimental group). On the heat transfer assessment, experimental students gained significantly whereas control students did not. However, the gain of the emergent group did not significantly exceed the gain of the control group, when adjusted for pre-test. We believe that preparation for future learning was modest because students did not have sufficient prerequisite knowledge about concepts (such as energy) leading up to the study of heat transfer. Thus, the experimental group may have had difficulty mapping heat transfer to emergence. A LearnLab study can easily correct this shortcoming.
We hypothesize that emergence is the key structure that underlies this class of process concepts, and that misconceptions arise because most students know nothing about emergence. To test our hypothesis, we developed a brief, self-contained domain-general instructional module that we refer to as emergent schema training. It focuses on the idea of emergence, which is related to notions of complex dynamic systems (Goldstone, 2006), to see if it can help students achieve greater learning, deeper understanding, and more successful transfer to other concepts. Our domain-general approach is relevant to a variety of concepts across science domains as well as to concepts in domains other than science, such as economics, political science, and history. Thus, the structure of emergence is an ideal test bed for robust learning not only in terms of long-term retention, but also for accelerated learning of new concepts.
The gist of our project plan is to run a modified version of the 2005 study in the Chemistry LearnLab. With assistance from David Yaron, chair of the chemistry course committee, we will modify existing treatment materials (including all the assessment questions) to make them suitable for college-level instruction, and we will create new materials for the control conditions. We will pilot the materials in pullout studies with college students in the fall semester of 2007. The study itself will be run in Dr. Yaron’s Modern Chemistry II LearnLab course in the spring semester of 2008, and again in 2009, implementing modifications based on what we learn from the 2008 data. Each year the study will involve approximately 120 students, and they will use the materials during their two-hour recitation classes. The study design is modified to unconfound our prior design, although that prior confound was intentional because we were seeking a “proof of concept” of our materials and we did not have enough participants to unconfound the effect of the emergent module from the effects of mapping prompts.
Study Phases Experimental Map Control Control Pretest Emergence Diffusion Heat Transfer Training Instruction Emergence Diffusion w/Mapping Emergence Diffusion, no Mapping Diffusion, no Mapping Additional Diffusion Problems Training Posttest Emergence Diffusion Transfer Instruction Heat Transfer, no Mapping Transfer Posttest Heat Transfer Table 1. Design of proposed study.
The design consists of three groups. The Experimental group is comparable to our previous experimental group, in that the students will receive both the general emergent schema training as well as mapping of diffusion to emergence. The Map Control group will receive the emergence schema training but without any explicit mapping (for the simulation parts of the instruction, they will receive attention prompts). The Control group is comparable to our previous control group, in that they receive no training about emergence nor any mapping, but they will receive the same instruction about diffusion in both text and simulations, and they will be given additional diffusion problems to solve instead of reading about emergence, as suggested by David Yaron. This design will allow us to unconfound the benefit of being exposed to the emergent module from the benefit of mapping. Time on task in all three conditions will be equated, and all three groups will receive the same pre-test, the same training post-test, the same transfer instruction, and the same transfer post-test (see Table 1 above).
The pretest will consist of approximately 50-60 items, in both multiple-choice and constructed-response format addressing each of three topics: emergence, diffusion, and heat transfer. The pre-test items will be developed with inputs from the Colorado School of Mines, since they will administer a misconception exam to their students in order to identify common misconceptions about diffusion and heat transfer at the college level. The pre-test will assess all the attributes listed in Table 2 (that is, the micro-level, the macro, the and the interlevel attributes, as well as the aggregating mechanism, which we cannot discuss in this short project plan). We currently conceive of these attributes as knowledge components.
In contrast to our prior design, the training module for the Experimental condition will include materials on both the general emergent schema as well as the specific concept diffusion. The Experimental training module will begin by introducing the emergent schema, describing everyday examples of emergent and direct processes and identifying explicit features for recognizing and explaining each type (see Table 2 for the features). It will then describe diffusion, with mapping in the form of descriptive text and questions embedded in the text and the simulations. These mappings will be designed to help students recognize diffusion as an emergent process. The Map Control training module will be identical to the Experimental module, except that the diffusion portion will not include mapping. The Control module will omit the introduction to emergence, and the description of diffusion will not include mapping. Instead, the Control module will provide additional practice in solving quantitative diffusion problems. All three modules will include dynamic simulations of diffusion at both the macro (outwardly visible) and micro (molecular) levels.
The training posttest will consist of approximately the same items as the pre-test, with some additional new ones. Diffusion questions will include quantitative problems as well as items that assess qualitative understanding.
The transfer instruction and posttest will be provided by Dr. Yaron, who will teach the heat transfer portion of his Modern Chemistry II course according to his usual syllabus. Specific relevant topics include temperature changes, heat capacity, phase changes, and chemical reactions that absorb or release heat. We will work with Dr. Yaron to incorporate at least 15-20 items into his existing assessment system. These questions, which will include quantitative problems as well as qualitative items, will enable us to determine whether students who received schema training understood subsequent instruction on heat transfer better than students who did not receive the training, a test of accelerated future learning.
The exact number of questions in all our assessments may vary after our pilot testing in the lab to see how long they take. We will have to shorten them to fit a 2-hour recitation session, since all tests and materials of our intervention will be delivered on-line, during students’ recitation periods. Although final decisions regarding the instructional delivery platform will be made in consultation with David Yaron, our tentative plan is to develop the pretest, training modules, and training posttest using SAIL, an open-source curriculum development system based on the Berkeley WISE framework. The simulations will be developed in Molecular Workbench and incorporated into the SAIL module. The Molecular Workbench has the facility to allow simulations to be stopped, and such a static display will be a useful feature for our instructional needs.
All three groups will receive the same pretest one week prior to our intervention and two weeks prior to David Yaron’s class instruction. During the intervention phase, students will be randomly assigned to treatment condition, creating three groups of approximately 40 students each. The post-test on emergence and diffusion will be administered a week later, as a measure of long-term retention. Finally, the transfer post-test will be administered some time after Dave Yaron’s instruction (as determined by his class schedule), so that it will assess accelerated future learning. In short, we will have all three facets of robust learning, including long-term retention of knowledge components in the training materials, accelerated future learning of a new concept, and solving new problems involving those concepts.
American Association for the Advancement of Science (1989). Science for all Americans. New York: Oxford University Press.
Carey, S. (1985). Conceptual change in childhood. Cambridge, MA: MIT Press.
Carey, S. (1991). Knowledge acquisition: Enrichment or conceptual change? In S. Carey & R. Gelman (Eds.), The epigenesis of mind. (pp. 257-291). Hillsdale, NJ: Erlbaum.
Chi, M.T.H. (2005). Common sense conceptions of emergent processes: Why some misconceptions are robust. Journal of the Learning Sciences, 14, 161-199.
Chi, M.T.H. (in press). Three kinds of conceptual change: Belief revision, mental model transformation, and categorical shift. To appear in Vosniadou, S. (Ed.), Handbook of research on conceptual change. Hillsdale, NJ: Erlbaum.
Driver, R. & Easley, J. (1978). Pupils and paradigms: A review of literature related to concept development in adolescent science students. Studies in Science Education, 5, 61-84. Duit, R. (2004). Bibliography: Students’ and teachers’ conceptions and science education database. Kiel, Germany: University of Kiel.
Goldstone, R.L. (2006). The complex systems see-change in education. Journal of the Learning Sciences, 15, 35-43. Hegarty, M., Kriz, S., & Cate, C. (2004). The roles of mental animations and external animations in understanding mechanical systems. Cognition and Instruction, 21, 325-360.
Hewitt, P.G. (1987). Conceptual physics: A high school physics program. Menlo Park, CA: Addison-Wesley. Novak, J. (1977). A theory of education. Ithaca: Cornell University Press.
Vosniadou, S. (2004). Extending the conceptual change approach to mathematics learning and teaching. Learning and Instruction, 14, 445-451.
- To be added