Data-driven (DD) science is focused on the collection and use of a diverse archive of observations. Data analysis discovers generalizations in the archive, which are inductively extrapolated to the world. This leads to a search for empirical consequences and their corroborating (or not) data via field observations and experiments. New results are added to the archive and the process repeats. In this way general laws are discovered and justified.
Hypothesis-driven (HD) science is focused on generating and testing hypotheses. After a search for evidence or experimental testing, an hypothesis gains or loses support. The process repeats with further investigation of the same hypotheses or alternative ones. While the hypotheses are often based on data analysis, they may be based on intuition, dreams, ideology, or whatever. The data used for HD hypothesis testing is often not made available for others to use.
DD science is more conservative and slower to progress but has a stronger justification. It is how younger sciences work best, leading to well-grounded laws. It is how statistical sampling works, where random samples justify generalization to the whole population. For natural science, the greater the archive of observations, the more there is a sound basis for universal generalizations.
HD science is better at quickly accumulating minor advances of a mature science with well-established laws. However, it is open to faddishness and ideological bias. The hypotheses that are investigated tend to be ones that are trendy or ideologically correct, while alternatives are ignored. Data sources may be cherry-picked to ensure positive results. This may lead to research with support for contradictory hypotheses, a notable problem with medical research.
DD science works best when much data has been collected. If data collection is too expensive or impossible (e.g., the subject is in the past), the temptation to say something immediately leads researchers toward HD science prematurely. This happened with geology and biology in the 19th century. The accumulation of large scientific databases has led to more interest in DD science but HD science is strongly entrenched as the standard. Bioinformatics is providing an opportunity for DD science with its large archive for well-grounded research.
HD science is often called hypothetico-deductive though that is not the original meaning of Whewell’s term. It is a common misconception of Whewell’s nuanced “Philosophy of the Inductive Sciences”. Darwin was one of the first to make it: take a grand hypothesis and look for scraps of evidence to prop it up. That is not justified induction, i.e., science.
Creationists have adopted HD science with a biblical foundation instead of the dominant naturalistic one. While this may be a good apologetic approach, it has all the weaknesses of HD science, which leads to dueling hypotheses rather than consensus. DD science takes longer but leads to general laws with strong support and so is a better methodology when there is controversy. The strongest arguments creationists can make are based on DD science, not HD science.
August 2011