Social Sensing: A maximum likelihood estimation approach

Considering the heuristic nature of fact-finding techniques, the quantified source truthfulness and claim correctness from Bayesian interpretation presented in the previous chapter remains to be a linear approximation . Moreover, results of Bayesian interpretation are shown to be sensitive to the priors given to the algorithm. To overcome these limitations, this chapter reviews the first optimal solution to the reliable social sensing problem we introduced in Chapter 1 . Optimality, in the sense of maximum likelihood estimation (MLE), is attained through an expectation maximization approach that returns the best guess regarding the source reliability as well as the correctness of each claim. The algorithm jointly makes inferences regarding both source reliability and claim correctness by observing how sources corroborate the claims. The approach is shown to outperform the state-of-the-art fact-finding heuristics, as well as simple baselines such as majority voting under certain conditions.