Social data miners need to become more explicit in discussing, and where possible addressing, the pitfalls and assumptions that underpin their data and methods. Doing so will force researchers to be cautious and stay within appropriate inferential boundaries, while also allowing consumers of insights derived from social data mining to be aware of their limitations.

The upside of the model organism, Twitter, is that it provides a common, shared laboratory of interest that cannot be ignored by scientists, researchers, or practitioners; however, as was hopefully made clear in this chapter, harnessing it should be done with caution. Only through careful, thorough contemplation of the nature and structure of social data will mining provide us with answers to today's pressing social and human questions.

The last note we wish to make is that all data has shortcomings. Data is produced from an imperfect world, and we would therefore expect it to be imperfect. Regardless of whether your data is perceived to be from an authoritative source or not, one should still think in the terms outlined earlier. Measurement, reliability, validity, and potentials and pitfalls are pervasive considerations of data whether that data is small or big, good or bad, or traditional or nontraditional.