學術論文

This article addresses the challenges in classifying textual data obtained from open online platforms, which are vulnerable to distortion. Most existing classification methods minimize the overall classification error and may yield an undesirably large Type I error (relevant textual messages are classified as irrelevant), particularly when available data exhibit an asymmetry between relevant and irrelevant information. Data distortion exacerbates this situation and often leads to fallacious prediction. To deal with inestimable data distortion, we propose the use of the Neyman–Pearson (NP) classification paradigm, which minimizes Type II error under a user-specified Type I error constraint. Theoretically, we show that the NP oracle is unaffected by data distortion when the class conditional distributions remain the same. Empirically, we study a case of classifying posts about worker strikes obtained from a leading Chinese microblogging platform, which are frequently prone to extensive, unpredictable and inestimable censorship. We demonstrate that, even though the training and test data are susceptible to different distortion and therefore potentially follow different distributions, our proposed NP methods control the Type I error on test data at the targeted level. The methods and implementation pipeline proposed in our case study are applicable to many other problems involving data distortion. Supplementary materials for this article, including a standardized description of the materials available for reproducing the work, are available as an online supplement.

From employees’ point of view, changes in ethical leadership perceptions can signal important changes in the nature of the employment relationship. Guided by social exchange theory, this study proposes that changes in ethical leadership perceptions shape how employees appraise their exchange relationship with the organization and affect their pride in or contempt for the organization. Changes in these associative/dissociative emotions, in turn, precipitate changes in behaviors that serve or hurt the organization, notably voice and turnover. Experimental data collected from 900 subjects (Study 1) and field data collected from 470 employees across 4 waves over 14 months (Study 2) converged to show that changes in ethical leadership perceptions were related to same-direction changes in employees’ pride in the organization and to opposite-direction changes in their contempt for the organization above and beyond the effect of the present ethical leadership level. Changes in pride were in turn related to same-direction changes in functional voice, whereas changes in contempt were related to same-direction changes in dysfunctional voice. The field study also provided evidence that when pride increased (decreased), employees were less (more) likely to leave the organization 6 months after. These results suggest that changes in ethical leadership perceptions are meaningful on their own, that they may alter employees’ organization-targeted behaviors, and that changes in associative/dissociative emotions are the mediating mechanism.