Science of Science
A few more findings on how we discover and learn (in case you don’t have a dog as I assumed in the post about Discovery and Being Self Aware ). Computational Approaches section discusses the use of artificial inteligence to help scientists make discoveries:
Scientific thinking as problem solving
“In a similar vein, Klahr and Dunbar (1988) characterized scientific thinking as a search in two problem spaces, an hypothesis space and an experiment space.” (hypothesis space is highly related to emergence)
Scientific thinking as hypothesis testing
“Using this approach, researchers have found that subjects usually try to confirm their hypotheses rather than disconfirm their hypotheses. That is, subjects will conduct an experiment that will generate a result that is predicted by their hypothesis. This is known as confirmation bias. Many researchers have shown that it is very difficult to overcome this type of bias. Mynatt, Doherty, and Tweney (1977) devised a task in which subjects had to conduct experiments in an artificial universe and found that subjects attempt to confirm their hypotheses. Dunbar (1993) has found that while subjects do try to confirm hypotheses, their hypotheses will change in the face of inconsistent findings. Klayman has argued that people possess a positive test bias – people attempt to conduct experiments that will yield a result that is predicted by their current hypothesis, and that under certain circumstances, this is a good strategy to use (Klayman & Ha, 1988).”
Experimental approaches to the development of scientific thinking
“Many researchers have noted that children are like scientists; they have theories, conduct experiments and revise their theories. Thus, while most researchers agree that scientists and adults have much more complex knowledge structures than children, the developmental question has been whether there are differences between children and adults abilities to formulate theories and test hypotheses.”
“Overall, recent research on the development of scientific reasoning indicates that, once amount of knowledge is held constant, few radical differences between children and adults abilities to test hypotheses and design experiments.”
Computational Approaches
“Early computational work consisted of taking a scientific discovery and building computational models of the reasoning processes involved in the discovery. Langley, Simon, Bradshaw, and Zytkow (1985) built a series of programs that simulated discoveries such as those of Copernicus and Stahl. These programs have various inductive reasoning algorithms built into them and when given the data that the scientists used, were able to propose the same rules. Computational models since the mid 1980’s have had more knowledge of scientific domains built in to the programs. For example, Kulkarni and Simon (1988) built a program, with much knowledge of biology and experimental techniques, and simulated Krebs’ discovery of the urea cycle. The incorporation of scientific knowledge into the computer programs has resulted in a shift in emphasis from using programs to simulate discoveries to building programs that are used to help scientists make discoveries. A number of these computer programs have made novel discoveries. For example, Valdes- Perez’s (1994) has built systems for discoveries in chemistry, and Fajtlowicz has done this in mathematics (Erdos, Fajtlowicz & Staton, 1991). See Darden (1997) for a summary of work on computational models of scientific discovery.”
Real-World Investigations of Science
“He has found that much of the scientists’ reasoning is concerned with interpreting unexpected findings. In fact over 50% of the findings that the scientists obtained were unexpected. As a consequence, scientists have developed specific strategies for dealing with unexpected findings that are very different from the strategies seen in the hypothesis testing literature. Dunbar has also found that scientists use analogies from similar – rather than dissimilar- domains in proposing new hypotheses . Furthermore the scientists distribute reasoning among members of a laboratory. For example, one scientsist may add one fact to an induction, another scientist add another fact, and yet a third scientist might make a generalization over the two facts. This type of research on real-world science is now making it possible to see what aspects of scientific thinking are important. By fusing together findings from real-world science with the results of the more standard experimental methods, it should be possible to build detailed models of scientific thinking that, when implemented, can be used by scientists to help make discoveries.”
from http://www.utsc.utoronto.ca/~dunbarlab/pubpdfs/DunbarMITECS.pdf
Related posts:
