Pronoun Usage in Online Sexual Predation

Introduction

There is a growing prevalence of Internet-mediated sexual offenses, as indicated by increases in the frequency of prosecution and clinical referrals for these crimes. ¹ One common form of these offenses is sexual solicitation, in which adults have online interactions with minors for sexual purposes. ² Briggs et al. report evidence for a distinction between “fantasy-driven” and “contact-driven” solicitation offenders. ³ The former engage minors in online interactions they find gratifying in and of themselves, and thus are unlikely to attempt direct sexual contact. In contrast, the latter group engage in online interactions with minors primarily to arrange offline meetings.

Contact-driven solicitation offenders (CDSOs hereafter) are the focus of the present study. Although there has been extensive research on sexual predation, pronoun usage has not been exclusively investigated in this context. Several recent studies have investigated the linguistic patterns of CDSOs, but none of the studies have looked exclusively at pronouns.^4–6 The current study is one of the first studies to investigate specific pronoun usage in this context and compare these linguistic patterns to their counterparts in nondeviant romantic interactions.

Olsen et al. offer a general model of the “luring” process child molesters use to contact and control potential victims that we will use to frame the communication strategies used by CDSOs in particular. ⁷ According to the theory, luring begins with a “gaining access” phase, in which perpetrators strategically place themselves in positions to encounter targets. Once strategically placed, the perpetrator will initiate a “cycle of entrapment” by cultivating a deceptive sense of trust in a selected target by using flattery, humor, and promises of attention or gifts, which then escalates to grooming the target for sexual contact.

This is done through subtle communication strategies, such as discussing prior sexual experiences or showing the target pornographic images. Finally, the perpetrator attempts to isolate the target from parents, guardians, and others in the social network and ultimately approaches the target by explicitly arranging a meeting for sexual contact.

The present study explored chat transcripts via computerized text analysis, with a focus on CDSOs' language use in interactions with decoys as they attempt to cultivate a sense of trust. Although their messages are misleading and ultimately malicious in intent, it is plausible that their ostensible relationship-building interactions with decoys exhibit linguistic patterns similar to those observed in nondeviant romantic relationships.

Partners in developing romantic relationships “leak” relational beliefs and evaluations in implicit language forms. In their articulation of the “interactional” view of interpersonal communication, Watzlawick et al. drew a distinction between the explicit, literal meaning of a discourse message and an implicit meaning reflecting relational dynamics. ⁸ Interpersonal scholars have since identified a variety of linguistic devices that encode these implicit meanings.

One such device received significant attention over the years. Some researchers have argued that first person plural pronouns (we, us, our) are markers of shared identity, affiliation, and interdependence. ⁹ Although increased use of we has been found in written descriptions of highly committed romantic relationships, ¹⁰ other research has found that spoken we usage frequency has no association with relationship satisfaction^11–13 and only a marginal association with relational interdependence. ¹⁴

Findings regarding usage of first person singular pronouns (I, me, my) have also been mixed. Sillars et al. discovered couples who used fewer first-person singular pronouns tended to have higher relationship satisfaction than those who used them more frequently. ¹¹ However, Simmons et al. found that I usage to be marginally positively associated with relationship satisfaction and me to be positively associated with negative behaviors during problem-solving discussions. ¹² Use of I, these researchers argue, reflects self-disclosure and perspective taking, whereas use of me reflects feeling of passivity and victimization.

Finally, usage of second-person pronouns (you, your) by couples in problem-solving discussions has been shown to be negatively correlated with relationship satisfaction ¹¹ and positively correlated with negative relationship behaviors,^12,13 a pattern that reflects excessive other-focused attention, but in an accusatory manner. ¹⁵ Although CDSOs are not building upon established relationships the way couples are, these findings provide evidence that pronoun usage in a “romantic” setting (not that there is anything romantic about adults trying to sexually solicit minors) can reflect relational dynamics.

The romantic partners participating in the studies reviewed above are different in many important respects from CDSOs attempting to lure perceived targets into criminal sexual contact. However, both groups put considerable effort into cultivating trust and intimacy with desired others, although for deceptive and malicious purposes in the case of CDSOs. A comparison of CDSOs' pronoun usage with romantic couples engaged in relationship-building conversations may prove useful in providing insight into the underlying linguistic patterns of child sex offenders and how those patterns differ from nonoffenders.

The association in pronoun use between partners in an adult romantic relationship and CDSOs could take several forms. Although the empirical evidence has been somewhat inconsistent, CDSOs may have similar I usage to romantic couples because they want to focus more on the child than themselves, similar to how satisfied partners focus more on each other rather than themselves. On the contrary, romantic couples have an already established relationship and may be more comfortable talking about themselves, which means CDSOs may use fewer singular first-person pronouns as they do not want to shift the conversation away from the target.

A similar argument could be made for the use of second-person pronouns, where CDSOs' excessive focus on the target may manifest in higher usage of you compared with others because his goal may be to flatter the target by focusing on attractive characteristics. However, a second possibility is that the CDSO will attempt to create a sense of belonging, which will result in few second-person pronouns so as not to make the child feel isolated, which could be related to why negative relational outcomes are positively associated with second-person pronoun usage. ¹¹

If a strategy of CDSOs is to build intimacy, then it can be accomplished by focusing on the cohesiveness of the relationship. This cohesiveness can be achieved by combining the two individuals into a single entity (e.g., you and me → we). The use of we can have an effect on the perceived closeness between individuals ¹⁶ and also relates to relational interdependence. ¹⁴ So when a CDSO is interacting with a child, the intimacy-building they engage in might be reflected in their use of plural first-person pronouns.

Unfortunately, there have only been a handful of studies done on the linguistic patterns of CDSOs, so it is unclear how their language will be compared with that of romantic couples. As a result, the current project attempts to better understand the strategies CDSOs use by looking more closely at their pronoun use and comparing it with that of romantic couples, a group that frequently engages in intimacy and relationship building. However, because the pronoun literature regarding relational intimacy is inconsistent at times, we attempt to answer the following question:

RQ1: How does CDSOs' pronoun usage (i.e., first-person singular/plural and second-person singular) differ from the pronoun usage of romantic couples?

There is, however, another group of individuals who need to be considered in the current study. Decoys are adults that converse with the CDSOs by imitating adolescents, and because they are concerned with portraying themselves as authentic adolescents, they should be unconcerned with building intimacy. Furthermore, decoys should only be concerned with passively capturing CDSOs who are trying to engage in sexual relationships with children, not actively persuading individuals to meet them, else they run the risk of being accused of entrapment. The language of decoys has yet to be systematically studied. Therefore, we pose an additional question:

RQ2: How does decoys' pronoun usage (i.e., first-person singular/plural and second-person singular) differ from the pronoun usage of CDSOs?

Materials and Methods

The dataset

The current study analyzed online chats between convicted CDSOs and decoys. A publicly available archive of 623 chats (see Fig. 1 for an excerpt of the chats) is available on www.perverted-justice.com, which was compiled by the Perverted Justice Foundation Incorporated (PJF), a nonprofit organization that is dedicated to gathering evidence against sexual offenders via digital media (e.g., online chats, texts) and providing it to law enforcement. This organization has volunteers pose as children around the age of puberty (∼13 years old) and engage in online interaction in chat rooms with potential CDSOs. PJF then coordinates with law enforcement to arrest the CDSOs when/if they arrive for the meeting.

OPEN IN VIEWER

FIG. 1.

Chat transcript excerpt.

Importantly, they allow the CDSO to initiate contact with the decoy first to avoid police entrapment (the Department of Justice defines entrapment as the act of implanting a criminal act in an innocent person's mind and then inducing the act to later convict that individual of the crime), making the choice to continue the conversation completely up to the CDSO and ensure that the legal evidence against CDSOs is permissible in court.

All 622 chat transcripts on the website are between child decoys and male CDSOs who were convicted of attempting to harm a child. Extracted chats were divided into CDSO and decoy language segments using a script built in the Python programming language. The script was designed to remove nonchat text (commentary by the decoy, links, irrelevant webpage text, etc.) and to calculate the frequency of pronouns produced by each chat participant (Python syntax is available from the first author upon request). A specialized list of pronouns (Table 1) was built to account for all of the various ways individuals may choose to spell these words when engaged in text-based communication. In addition, it should be noted that some individuals produced text in a haphazard and ungrammatical form, such as accidentally typing the first letter of a word at the end of the previous word, which made it unrealistic to create a script that would capture all such mistakes accurately:

OPEN IN VIEWER

Table 1.

Pronoun Dictionary

First-person singular	i | me | ill | i'll | mine | my | myself | id | i'd | i'm | im | ive | i've | my | myself | myslf
Second-person	thee | thine | thou | thoust | thy | ya | yu | you | u | you've | u've | youve | uve | uv | ur | your | ru | you're | u're | ure | youre | ya'll | yall | you'll | youll | u'll | ull | ul | you'd | youd | u'd | ud
First-person plural	our | ours | ourselves | we | we're | we'll | us | we'd | weve | we've

Note: This pronoun dictionary was custom built to account for the many grammatical and spelling errors in the chat transcripts. The analysis was case-insensitive.

Decoy: hey

CDSO: wats up good looking

Some even chose not to include any spaces between words (e.g., “howareyoudoing?”), posing a problem for the script when counting pronouns that consist of letters that can be found within other words (e.g., “me” in “meeting”). Due to the size of the corpus and the length of each chat, having coders manually read through the chats and separate words was not feasible.

After all the chats were processed with the script, pronoun frequencies were calculated for each participant in each chat, total word counts were calculated for CDSOs and decoys, and the number of boy and girl targets was calculated. Gender counts were possible because decoys posted this information for most of the chats, but not all.

Results

Discussion

Overall, the results show that pronoun use tends to differ among CDSOs, decoys, and romantic couples. Specifically, CDSOs use more you words than decoys and romantic couples, while I and we differences were less consistent. CDSOs try to portray trust and warmth when they interact with children, ¹⁹ which can come in many forms, such as compliments. However, it is hard to say how CDSOs might choose to convey this “warmth” because one CDSO may choose to compliment the child's physical features, while another might choose to talk about how he will lavish the child with gifts.

It was hypothesized that this warmth would come across through the use of we as a result of the CDSO trying to portray closeness, but the results only provided partial support. What was clear was that CDSOs used significantly more you pronouns when talking to the decoys, suggesting that they were more focused on the decoy than themselves. This emphasis could be a symptom of trust building, which is an integral part of luring, ⁶ and a straightforward way of conveying interest and attention to the child.

Our results provide some evidence that CDSOs have similar pronoun use to romantic couples, but perhaps more interesting is how they are distinct. CDSOs tend to use more second-person pronouns than romantic partners or child decoys. This difference suggests that they are more other-oriented in their focus. They attempt to push the focus back onto the child either because they themselves want to focus more on the child or they want to avoid having the focus on them.

Focusing more on the child could aid the CDSO in portraying care or concern (e.g., “you are better than that” or “you are so cute”) and result in the child being more receptive of them. This pattern was also observed in Black et al.'s analysis when they found that almost all CDSOs used flirtation as a seduction device. ⁵ Similarly, CDSOs used fewer I than decoys and the romantic partners in Kurlak's dataset. This could be because CDSOs are trying to shift the focus off of themselves to the child. If the conversation is more focused on the CDSO, children may begin to distance themselves because they feel that the CDSO is not paying attention to them.

Although CDSOs use you consistently more than all of the groups in both datasets, their use of we was not as consistent. This can be accounted for by the designs of the two experiments, where Kurlak's participants were instructed to discuss their relationship, while Slatcher's participants merely provided their daily correspondence with their relational partner. Therefore, it is not surprising that couples in the Kurlak's study had significantly more we usage because they were actively discussing their relationships (e.g., “we need to do X, but I like how we do Y already”). Assuming that is the reason the data differed, it is interesting to see that CDSOs used a similar amount of first-person plural pronouns to Slatcher et al.'s couples, because this could point to CDSOs attempting to portray closeness, ¹⁸ just as romantic partners might. This pattern should be further tested with a sample of individuals actively engaged in flirting or seduction, which is similar to what CDSOs attempt to do.

There are two limitations that should be noted. First, the comparison data were only meant to test if CDSOs' language differed from romantic couples' language. Future comparisons should be made with adults actively engaged in courtship initiation (e.g., online dating) to get a clearer picture of how CDSOs' differ from nondeviant counterparts engaged in flirting. Second, our corpus comprised CDSOs caught by police, whose luring techniques may differ from those not convicted. However, currently there is no feasible way to obtain data from CDSOs who have not been caught.

The linguistic tendencies of CDSOs have recently become of interest to researchers because of advancements of computational analyses of language. This area of inquiry is important because in the past, researchers and practitioners analyzed the motivations and grooming strategies of sexual offenders through observation and clinical interactions. Nuances in their language use revealed by computational analyses can enable us to better understand these motivations and strategies. The current study contributes to this understanding by comparing CDSO pronoun use to nondeviant romantic interactions. Understanding that child sexual offenders speak to their targets in a more other-oriented tone compared with couples provides initial evidence that their language shows unique patterns. Understanding these patterns can help clinicians and practitioners understand these individuals and their motivations on a deeper psychological level, and future researchers can build upon this study in pursuit of accurate ways of identifying predatory language online.