Chris Anderson holds a pretty positive view towards big data. He believes that models become no longer criticle for modern sciences after the emergence of big data analysis technology, and big data become the “end of theory” and “makes the scientific method obsolete”. Reseachers can simply push data into a computer and wait for the computing clusters to work out a correlation or analysis of the data. Scientists don’t need to make any hypothesis before using the data to test it. The success of Google and Amazon seem to verify this point since they can rely on algorithms to make recommendation to consumers without actually knowing the exact information of what the consumer has bought or wish to buy.
Rob Kitchen, however, maintains the importance of model and human intervention in sciences. Relying completely on computers can indeed save time for formulating hypothesis and can avoid biases from human operation, but in some social science realm, such as art or analyzing the constitution of eographic concentration of different ethnic communities, there are many factors that can’t be quantified or processed simply by machine. Moreover, humans are likely to be irrational so that they don’t necessarily follow a general rule. Human societies are complex and human activities are tend to be influenced by contingencies, making human behaviors hard to predict. Models that are built based on insufficient data or to make prediction on unknown things, the computer may provide a parochial or inaccurate result. In addition, the way data are collected is also manipulated by human; people need to have the knowledge in certain field and need to understand the context based on pre-existed theories to be able to think carefully in order to draw a representitive sample of population or eliminate the existance of noise in the data.
Big data and data analysis technologies are definitely important to science development, but there are still challenges including the skills deficiencies in analyzing and understanding this type of data, and creating epistemological methods that support computational social science.