How Scientists Build Models: InVivo Science as a Window on the Scientific Mind

How do scientists think, reason and generate new models and theories? How do scientists represent their knowledge? Answers to these questions are of paramount importance not only in understanding what science is, but also in assessing different theories of science. Surprisingly, we know little about the basic processes that are involved in current-day scientific thinking. The goal of my research is to look at what scientists actually do in their research, what types of thinking and reasoning strategies they use, and how they change their knowledge. Over the past decade, I have been investigating scientific thinking by scientists in their own labs, reasoning about their research and by conducting experiments on scientific thinking and model building in my own laboratory. The labs that I have been investigating are molecular biology and immunology laboratories in the U.S., Canada, and Italy. I have found that one place where much reasoning and new discoveries are made is at weekly lab meetings. We have performed extensive cognitive analyses of these meetings and have identified some of the key components of contemporary scientific thinking that are important in generating new models, modifying old models and solving difficult problems. In this paper I will outline four activities that are important in model building: Analogical Reasoning, Attention to Unexpected findings, Experimental Design, and Distributed Reasoning. Introduction: A new way of investigating science-The InVivo Approach Kevin Dunbar, McGill University What scientists do and how they think has been a surprisingly difficult issue to investigate. Researchers have used notebooks, interviews, diaries, historical reconstructions and colleagues' accounts to obtain a picture of the cognitive processes that underlie scientific thinking and model building. Using these approaches, researchers have built many important accounts of what happens in science. However, over the past decade we have been pursuing a different approach to investigating the ways that scientists think and reason. We have been investigating the thinking and reasoning that occurs "live" at laboratory meetings in molecular biology and immunology laboratories (Dunbar, 1995, 1997, 1999). What we have done is to videotape and audiotape scientists as they think and reason in their own labs. We then analyze, sentence by sentence, the types of thinking and reasoning that the scientists use when formulating theories, analyzing data, designing experiments, and building models. We have investigated leading laboratories in The United States, Canada, and Italy. Each laboratory we have followed from three months to a year, taping the weekly laboratory meetings and supplementing the meetings with interviews and other documents such as grant proposals, drafts of papers, and one-on-one meetings. Using this approach we have been able to build detailed models of the cognitive underpinnings of the scientific mind. In particular, we have been investigating the roles of Analogy, Unexpected findings, and Group Reasoning in contemporary science. Why use the lab meeting as a source of data on scientists? I found that the most representative crosssection of the ways scientists think and reason is the weekly laboratory meeting. It is in the weekly lab meeting that scientists present their data, reason about findings and propose new experiments and theories. Thus, by analyzing the reasoning that scientists use at their lab meetings it was possible to obtain a clear vista into how scientists really reason. Furthermore, much of what happens at meetings consists of spontaneous reasoning, and sometimes scientific discoveries and breakthroughs are made at these meetings. Using this method it is possible to directly monitor thinking and reasoning rather than uncovering reasoning through after-the-fact interviews, questionnaires or think aloud protocols. The scientists externalize much of their thinking through interactions with other scientists in the lab. Thus by recording laboratory meetings it is possible to gain access to “online” thinking and reasoning. This approach allows us to see science as it unfolds rather than the selective information that the scientists put in their notebooks or reconstruct when attempting to remember what happened in their lab. The lab meetings therefore provide a unique opportunity to see science "live." Borrowing terminology from biology I have called this the "InVivo" cognitive approach (Dunbar 1993, 1995, 1997, 1999a). Another approach that we have used is the "InVitro" Cognitive approach where we conduct experiments on scientific thinking in our own cognitive laboratory. By using a combination of InVivo and InVitro approaches it is possible gain a clearer picture of the cognitive processes involved in science. A typical lab meeting The laboratories that we have been investigating consist of a senior scientist, three or four post-doctoral fellows, five or six graduate students, and one or two technicians. Each week one of the members of the lab will present the results of their latest experiments. The presentations consist of a brief rationale, many data slides, and what they are going to do next. While this is the basic plan of a meeting, some tend to focus more on theory, others on method, others on data, and many on future experiments. Thus, the meetings provide a nice cross section of the things that scientists do and provide a window into the workings of the scientific mind. While the lab meeting does have a presenter, the format of the meetings is very informal. Much of the lab meeting consists of many scientists reasoning about all aspects of the enterprise, models are discussed, diagrams drawn, inductions made, deductions given, competing models discussed, experiments designed and dissected. In addition, the possibility of alternate models, methodological errors, and feasibility of various approaches is discussed. From the point of view of a cognitive scientist, this type of data is unrivalled in its richness. Here we can see all the basic reasoning processes at work and how different cognition really works in science. These meetings are supplemented by interviews both before and after the meetings that provide detailed background information. Using this InVivo approach we have uncovered a number of basic features of the way scientists think, reason and build models. In the next sections I will provide an overview of four important aspects of building scientific models –Analogy, Unexpected findings, Distributed Reasoning, and the cognitive processes subserving experimental design. While each of these different aspects of model building can be discussed separately, it is important to note that scientists freely move between analogy, unexpected findings, distributed reasoning, and designing experiments. Thus, while I have four different sections on these aspects of scientific model building, I do not mean to imply that each process works alone, or moves along in a serial manner.