Confidence bounds in social sensing

In the previous chapter, we reviewed a maximum likelihood estimator (MLE) to jointly estimate the reliability of sources and determine the correctness of facts concluded from the data. However, an important problem that remains unanswered from the MLE approach is: what is the confidence of the resulting source reliability estimation? Only by answering this question, can we completely characterize estimation performance, and hence source reliability in social sensing applications. This chapter reviews analytically-founded bounds that quantify the accuracy of such MLE in social sensing. It is shown that the estimation confidence can be quantified accurately based on both real and asymptotic Cramer-Rao lower bound (CRLB). Additionally, this chapter also reviews an estimator on the accuracy of claim classification without knowing the ground truth values of the claims.