Page:Sm all cc.pdf/167

 Few books in philosophy of science have attracted the interest of scientists; Kuhn’s [1970] book is an exception. His vision of scientific change -- continuous incremental science, plus rare revolution -- is fascinating to those in fields undergoing this punctuated illumination. The changes associated with the 1968 geological paradigm of plate tectonics appear to fit his model, as does the recent paradigm of chaos, described by Gleick [1987] as a “revolution in physical sciences”. Successful prediction always confirms a model more persuasively than does detection of apparent pattern within existing data.

“Yet when we see how shaky were the ostensible foundations on which Einstein built his theory [of general relativity], we can only marvel at the intuition that guided him to his masterpiece. Such intuition is the essence of genius. Were not the foundations of Newton’s theory also shaky? And does this lessen his achievement? And did not Maxwell build on a wild mechanical model that he himself found unbelievable? By a sort of divination genius knows from the start in a nebulous way the goal toward which it must strive. In the painful journey through uncharted country it bolsters its confidence by plausible arguments that serve a Freudian rather than a logical purpose. These arguments do not have to be sound so long as they serve the irrational, clairvoyant, subconscious drive that is really in command. Indeed, we should not expect them to be sound in the sterile logical sense, since a man creating a scientific revolution has to build on the very ideas that he is in the process of replacing.” [Hoffmann, 1972]

Pitfalls of Evidence Evaluation
Scientific progress under a guiding paradigm is exhilarating. Paradigm-driven science can, however, undermine the objectivity with which we evaluate hypotheses and evidence.

Hidden Influence of Prior Theory on Evidence Evaluation
Data evaluation should consist of three separate steps: (1) objective appraisal of the observations, (2) confirmation or refutation of a hypothesis by these data, and (3) overall evaluation of a hypothesis in the context of these and other observations. All too often, we allow our prior opinion of a hypothesis to influence the evaluation of new evidence (steps #1 & 2), without being aware of the bias. This hidden influence is a pitfall, whereas it is completely valid to weight prior evidence more than the new data (step #3). In both cases, the impact of evidence depends on the perceived strength of the hypothesis it affects. Evidence sufficient to uproot a weakly established hypothesis may fail to dislodge a well established one.

We value simplicity, and it is much simpler and more comfortable if new evidence confirms previous beliefs than if it creates conflict. Ideally, one (and only one) hypothesis is consistent with all observations. To obtain this ideal, we may subconsciously reject evidence that conflicts with the hypothesis, while overemphasizing evidence that supports it. We must beware this subconscious theory-based rejection of data.

Children and adults use similar strategies to cope with evidence that is inconsistent with their prior beliefs [Kuhn et al., 1988]: