Qualitative results show that some system-level identities were significantly correlated and/or predictive of censorship. Posts placed in the “m4w” section showed the highest correlation, while a post placed in the “m4m” section and a young age were predictive of censorship.
Quantitative Results
Gender/Sexual Orientation. Correlations were calculated between censorship and a number of gender and sexual orientation categories embedded into the technology (see Table 1). Correlations revealed that some identity categories were positively correlated with censorship. Posts placed in the “m4m” section (r=.063, p=.012), “w4w” section (r=.02, p=.02), and all posts authored by women (r=.054, p=.026), and all posts targeted to men (r=.081, p=.002) were more likely to result in censorship. Other identity categories were negatively correlated with censorship. Posts placed in the “m4w” section (r=-.125, p<.001), all posts targeting women (r=-.102, p<.0001), and those posts written by men (r=-.072, p=.005) were less likely to end up censored. The correlation between censorship and the “w4m” category was positively correlated, but not significant (r=.031, p=.132).
Age. A correlation was also calculated between censorship and age when posts authors included this optional value (N=1289). With a negative correlation between censorship and age (r=-.08, p<.002), posts submitted by younger authors were more likely to be censored.
Correlation Coefficients. Partial correlation coefficients were calculated across all system-level identity variables using censorship as a dependent variable (see Table 2). The “m4m” gender/sexual orientation category was positively predictive of censorship. Both age and male authorship were negatively predictive of censorship. The contradictory findings between the “m4m” category and male authorship suggest that variables of “homosexual male authorship” and “heterosexual male authorship” may reveal stronger effects.
Qualitative Findings
While the sheer variety of censored posts revealed a number of different uses for both craigslist and the flagging system, a close reading of posts not marked as SPAM revealed a number of reoccurring, and frequently overlapping themes. Seven of the most common are described below.
SPAM-like. A portion of the posts shared features that some readers may have considered SPAM-like. Malformed message that did not contain a comprehensible message, for example, were often removed from the site. These posts might were excluded from the SPAM coding system because they did not meet the requirement of a reoccurring or commercial post, or a post that linked to an external site.
Miscategorized. Posts that may have been removed for being submitted to the wrong section were common in the censored portion of the dataset. A number of censored posts would have been more appropriate for the “Rants and Raves”, “Casual Encounters”, or “Personals” section.
Confessionary. Some authors use Missed Connections as a type of confessionary. These posts frequently addressed individuals the author knows or knew at one time. Apologies, regret, and explanations were all common purposes in these censored posts. The mourning, apologetic, and sometimes angry posts represent some identity aspects that readers may have influenced their censorship behavior.
Complaints. Post written by an individual issuing some complaint was another behavior that resulted in censorship. Authors of these posts use craigslist to voice their complaints. Angry posts, particularly those that made complaints about the gender and/or sexual orientation section in which they were posted, were quickly removed. These complaints were directed at people and events from the physical world, as well as fellow craigslist users.
Too little identity vs. too much identity. Posts that did not contain legible identities were common among the censored group. Most common were posts that didn’t identify a target or a second party. Similarly, posts that contained too much information were often censored. Posts that eliminated the relative anonymity of the site by included names and identifying information were frequently removed.
Sexual Content. By far the most commonly censored identity type were those connected to some time of sexual content. These posts often included details that some reader might find deviant, or otherwise inappropriate for an public space. “I know you think it’s hot that I didn’t wear panties to work today” ends one censored post from the w4w section. This type of censorship was particularly common in the m4m section. This may be a result of the quantity of these types of posts in the m4m section (censored or not), or this may reflect a type of censorship behavior specific to this section.