No abstract

Purpose The purpose of this paper is to examine the impact of internet censorship, which is represented by the Great Fire Wall, on Chinese internet users’ self-censorship. Design/methodology/approach A 3×2 factorial experiment (n=315) is designed. Different patterns of censorship (soft censorship, compared censorship, and hard censorship) and the justification of internet regulation are involved in the experiment as two factors. The dependent variable is self-censorship which is measured through the willingness to speak about sensitive issues and the behavior of refusing to sign petitions with true names. Findings The results show that perceived internet censorship significantly decreases the willingness to talk about sensitive issues and the likelihood of signing petitions with true names. The justification of censorship significantly decreases self-censorship on the behaviors of petition signing. Although there are different patterns of internet censorship that Chinese netizens may encounter, they do not differ from each other in causing different levels of self-censorship. Research limitations/implications The subjects are college students who were born in the early 1990s, and the characteristics of this generation may influence the results of the experiment. The measurement of self-censorship could be refined. Originality/value The study contributes to the body of literature about internet regulation because it identifies a causal relationship between the government’s internet censorship system and ordinary people’s reaction to the regulation in an authoritarian regime. Unpacking different patterns of censorship and different dimensions of self-censorship depicts the complexity of censoring and being censored.

No abstract

Ocean acoustic tomography can be used based on measurements of two-way travel-time differences between the nodes deployed on the perimeter of the surveying area to invert/map the ocean current inside the area. Data at different times can be related using a Kalman filter, and given an ocean circulation model, one can in principle now cast and even forecast current distribution given an initial distribution and/or the travel-time difference data on the boundary. However, an ocean circulation model requires many inputs (many of them often not available) and is unpractical for estimation of the current field. A simplified form of the discretized Navier-Stokes equation is used to show that the future velocity state is just a weighted spatial average of the current state. These weights could be obtained from an ocean circulation model, but here in a data driven approach, auto-regressive methods are used to obtain the time and space dependent weights from the data. It is shown, based on simulated data, that the current field tracked using a Kalman filter (with an arbitrary initial condition) is more accurate than that estimated by the standard methods where data at different times are treated independently. Real data are also examined.

The method of ocean acoustic tomography (OAT) can be used to invert/map the ocean current in a coastal area based on measurements of two-way travel time differences between the nodes deployed on the perimeter of the surveying area. Previous work has attempted to relate the different measurements in time using the Kalman filter. Now, if the ocean dynamics or model is known, one can also determine the current field given an initial distribution or the travel-time difference data on the boundary, and can even forecast the current changes. Based on the ocean dynamics, the current field is shown to be spatially and temporally correlated. We derive their relation and use that as the state model for the Kalman filter; the coefficients are estimated from data using an auto-regressive analysis. Armed with this model, it is shown based on simulated data that the current field can be tracked as a function of time using the Kalman filter (with an arbitrary initial condition) with a higher accuracy than that estimated by OAT. The reason of the improvement, the use of spatial-temporal state model (versus using only the temporal evolution), is studied. The method has also been applied to real data.