The IDRC Asia Office was kind enough to permit publication of a part of a concept paper I wrote for them.
Here is an excerpt:
At present, the principal methods for understanding users or demand are quantitative research (representative-sample surveys) and qualitative research. The former is used primarily for understanding “what” questions and the latter for “why” questions. Quantitative research is very costly. Its limitations include problems of recall and different forms of bias. Data analytics is likely to displace quantitative research in the long term. There is not much value in relying on recall of what people do when behavioral data provide a great deal more detail in an unobtrusive manner at low cost. Without the hypothesis testing structure of “old” social science, computational social science that is based on big data can only provide correlations not causal explanations. Big data cannot tell why humans do what they do. Without that element causal explanations are difficult to derive. Therefore, qualitative research will continue to play a complementary role.
In the short term, there may however be a need for additional quantitative research. This is to test and improve the findings of data analytics. Surveys can also be used to “train” the big data. For example, in one big data for development study, phone surveys were used to establish various demographic characteristics that were then coupled with call detail records (CDRs) provided by the mobile operators to train machine learning algorithms. For example, anonymized (and perhaps even non-anonymized) CDRs by themselves will give no information on gender. To factor in gender and socio-economic status one has to bring in other data, usually from surveys.