Understanding Disclosure Motivations and Their Linguistic Correlates

Overview

This report outlines key findings and ongoing research from CREST regarding the motivations behind individuals disclosing information about people they know, a critical area for enhancing security research and practitioner understanding.

Executive summary

This report builds on earlier CREST research into why individuals choose to disclose sensitive information about people they know, aiming to deepen understanding of disclosure behaviour in security-relevant contexts.

Phase I of the research, led by Ryan Boyd and Sheryl Prentice, employed behavioural and NLP analysis on open-ended narrative data of over 763,000 words from 848 participants to identify five core motivations behind disclosure: Responsibility, Relationships, Rules, Roles, and Righteousness. Contrary to earlier models focused on self-serving motives such as revenge or financial gain, this phase highlighted primarily prosocial motivations – including moral duty, professional obligation, and concern for others.

Phase II extended this work through forensic and corpus linguistic methods, examining how linguistic features correlate with personality traits and how these shape the ways individuals discuss their motives for disclosure. Interestingly, the top two key semantic categories, Expectations and Ethics, were notably prominent in motivation narratives by participants scoring higher in manipulative or emotionally detached traits, potentially suggesting a strategic moral self-framing. Additionally, participants high in emotional vulnerability emphasised close interpersonal relationships particularly with their mothers, while those with higher manipulative scores were more likely to focus on impersonal service-based relationships, for example with drivers and medical practitioners.

Together, these findings reveal personality-linked disclosure styles and offer a richer picture of the psychological and linguistic mechanisms underpinning disclosure decisions. They demonstrate the forensic value of linguistic analysis, informing approaches to eliciting, interpreting, and evaluating disclosures in intelligence, law enforcement, and organisational contexts. These insights may support the design of disclosure strategies that are psychologically and linguistically tailored to individual profiles, as well as partnering interviewer personality profiles to particular interviewees to improve outcomes, particularly in sensitive or high-stakes settings.

Introduction

Understanding why individuals choose to disclose sensitive or damaging information about others is critical for advancing research in areas such as law enforcement, intelligence, and corporate security. While previous studies have largely focused on self-disclosure in therapeutic, corporate, or social media contexts, often assuming shared goals or ulterior motives, there remains a significant gap in understanding the real-world motivations behind disclosures that carry risk, harm, or moral complexity. This report addresses that gap by tracing the development of CREST’s multi-phase research programme into disclosure behaviour, with a particular focus on its linguistic and psychological underpinnings.

Phase I, led by Sheryl Prentice and Ryan Boyd, established a foundational framework for understanding disclosure motivations, identifying five prosocial drivers – Responsibility, Relationships, Rules, Roles, and Righteousness – and profiling typical disclosers. Phase II deepens this analysis through forensic and corpus linguistic methods, exploring how individuals narrate their motives, how these align with personality traits, and the linguistic traces those rationalisations leave behind. This report presents those findings, examining how different psychological profiles shape disclosure narratives and the ethical, relational, and strategic tensions that underpin the decision to share sensitive information.

Phase I: Uncovering Disclosure Motivations

Phase I of CREST’s disclosure research, led by Sheryl Prentice and Ryan Boyd, set out to identify who discloses sensitive information about people they know, and why. 848 participants were recruited via Prolific and completed a survey administered through Qualtrics. Each participant responded to a series of open-ended prompts concerning a time they had disclosed negative or damaging information about someone they knew, specifically to a third party. These prompts elicited a total of 763,269 words of narrative data and included questions such as:

• Describe a time you disclosed information about someone you know or knew.
• What motivated you to disclose that information?
• Did you experience any conflicting feelings before the disclosure? (optional)
• Why did you choose to disclose to that particular person?
• What were the consequences for you or for the individual concerned?
• How recently did the event occur?
• How do you now feel about the disclosure in retrospect? (optional)

Demographically, those most likely to disclose were identified as female, aged in their twenties to thirties, with working- or middle-class affiliations and a strong sense of social standing. These respondents formed the basis of a typology built not just on their behaviours, but also on their psychological profiles.

However, when it came to motivations, this work challenged prevailing models, such as Miller (2011), whose framework suggested that disclosures were largely driven by self-serving interests such as revenge, financial gain, or avoidance of punishment. In contrast, Phase I findings pointed towards a fundamentally different picture: individuals disclosing for prosocial reasons, motivated by ethical concern, relational obligations, or institutional responsibility.

To explore the connections between personality and disclosure, participants also completed several standardised psychological assessments, including the Big Five (Gosling et al., 2003), the Dark Triad (Jonason & Webster, 2010), Buss’s Manipulation Tactics Scale (1987), and Ward et al.’s (2006) measures of agency, communion, and emotional vulnerability. These results were then linked to the linguistic patterns found in the disclosure narratives, analysed using Linguistic Inquiry and Word Count (LIWC), a natural language processing tool capable of extracting psychologically meaningful patterns from large text datasets. From this analysis, five core motivational themes emerged, often overlapping within individual cases:

• Responsibility – A sense of duty to protect others, particularly vulnerable individuals.
• Relationships – Motivations rooted in relational ties, such as family or close friends.
• Rules – Concerns about violations of formal or informal social norms.
• Roles – Obligations tied to a professional or institutional position.
• Righteousness – Moral imperatives and a desire to correct perceived wrongdoing.

These motivations were not necessarily discrete. For instance, a healthcare professional reporting clinical negligence might do so out of concern for patient safety (Responsibility), adherence to workplace standards (Roles), personal investment in a colleague or patient (Relationships), a desire to uphold procedural or legal expectations (Rules), and moral outrage at the misconduct (Righteousness).
Importantly, the study also found linguistic markers associated with these motivations. For example, disclosers often used action-oriented verbs (wanted, hoped, stop) and emotional terms (angry, hurt, upset) in their narratives, suggesting not only intent but affective investment. By combining linguistic evidence with personality and demographic profiling, the researchers were able to develop a scalable, data-driven framework that moves beyond the anecdotal or speculative models found in earlier research.

In sum, Phase I provided a solid empirical foundation for understanding disclosure as a complex, fraught, and/or prosocial act. Rather than necessarily being driven by malice or gain, individuals seemed to be disclosing because they felt they should. Whether this sense of obligation is internally held or externally performed, however, is a question that would be taken up in Phase II.

Phase II: Complicating Motivations

Building on the motivational framework established in Phase I, Phase II of the project sought to explore how those motivations were linguistically realised and shaped by personality. While Phase I identified what people say motivates them, Phase II examined how these motives are narrated, and how those linguistic choices correlated with underlying personality profiles.

Phase II used corpus linguistic tools to analyse participants’ responses, specifically focusing on their answers to the question: “What motivated you to disclose that information about that person?” Using Wmatrix, a semantic tagging and analysis tool, these motivation narratives were compared against the narrative responses to the rest of the questions in the survey. The goal was to identify semantic categories that were particularly overused in accounts of motivation, and to cross-reference these with participants’ personality profiles:

(Download Report to see Table 1: Top 10 Semantic Domains)

Two semantic domains stood out as especially prominent: Expectations and Ethics. These categories appeared markedly more frequently in participants’ motivation narratives than in their other responses. Mapping these linguistic patterns onto personality profiles revealed further interesting insights.

Participants who scored highly for traits associated with manipulation, entitlement, and emotional detachment – particularly within the Dark Triad and manipulation tactics scales – were more likely to invoke expectations and ethical reasoning in their justifications. In other words, individuals who self-identified as exploitative, coercive, threatening, or lacking empathy were paradoxically more likely to discuss their motivations with reference to some form of morality or social expectation. Terms such as ought, should, moral, fair, and principle occurred at elevated rates among participants who also indicated that they were not kind, not dependable, and indifferent to the moral implications of their behaviour.

This apparent contradiction raises an important interpretive possibility: that these individuals may be engaging in strategic moral self-framing. In a context where disclosing harmful information about someone else is socially fraught, strategic deployment of ethical language may serve a performative function – mitigating potential blame, preserving social standing, or rationalising self-interested action under a prosocial veneer.

Individuals who scored highest for emotional vulnerability (the “light” group) showed very different linguistic tendencies. Their narratives focused more on close interpersonal relationships with mothers, ex-partners, and male relatives. Their disclosures also invoked sites of social tension, such as schools and workplaces, and included negative actions such as drink driving and abuse. By contrast, the motivation narratives for those who scored highest for manipulative and coercive tendencies (the “dark” group) focussed more on impersonal, service-based relationships, for example with drivers and medical practitioners, as well as transactional matters such as places to spend money, and time off. They also used more utility-based reasoning such as “pointless” or “ineffective”.

Discussion: Light v Dark

The findings from Phase II offer both a validation and a complication of the motivational categories identified in Phase I. While the original framework – Responsibility, Relationships, Rules, Roles, and Righteousness – remains broadly supported, the linguistic evidence reveals new texture within each, particularly around how individuals narrate and rationalise their decisions to disclose.
For instance, the theme of Responsibility was echoed in the frequent invocation of expectations, a semantic domain significantly overused in motivation narratives. However, this alignment proved to be psychologically double-edged. Among individuals with high scores on exploitative or manipulative traits, expectations may be seen less as genuine obligations and more as rhetorical devices, potentially used to externally justify actions taken for internally self-serving reasons. This finding reframes the initial prosocial interpretation of disclosure: what may sound like duty might, in some cases, be strategic self-exoneration.

The motivational category of Righteousness received the strongest empirical support in Phase II, particularly through the semantic cluster of ethics. Yet again, however, the data revealed a split. Individuals who scored themselves low on empathy, kindness, and conscientiousness were paradoxically more likely to use moral language to justify their disclosures. This suggests that ethical reasoning, while central to many narratives, is not always an honest signal of internal values. It may be a kind of moral camouflage, deployed when reputational risk is high.
In contrast, emotionally vulnerable individuals – the “light” group – produced disclosure narratives that foregrounded close personal connections, sites of interpersonal tension, and behaviours carrying significant social and moral weight (abuse, drink driving).

Together, these findings suggest that individuals may adopt one of two broad rhetorical modes when explaining their decision to disclose:

• Rationalising mode – characterised by abstract moral language, appeals to duty, and strategic self-framing, cf. those with higher manipulative and coercive scores.
• Emotive mode – grounded in personal experience, interpersonal relationships, and proximal sensitive contexts, cf. those with higher vulnerability and empathy scores.

Further investigation will be necessary to explore the generalisability and wider implications of these findings. However, understanding the nature of disclosure motivations has practical relevance in forensic and intelligence contexts via informing interview protocols, decision-making schemata, narrative evaluation checklists, and so forth. When interpreting disclosures, it may be important not only to analyse what is said but how, and by whom. For instance, a polished moral justification might not necessarily signal ethical depth so much as reputational management.

Future Challenges

Despite the strengths of this dataset, limitations remain. Self-report always risks performativity, and even rich linguistic data is only a proxy for internal motivation. Additionally, while the participant sample was large by qualitative standards, it is far from representative of all disclosure contexts. Sub-setting the data also necessarily reduced the scope of generalisability. Future research should aim to test these patterns in real-world disclosure scenarios and further examine whether tailoring not only interview strategies, but also interviewer personality profiles to certain interviewees improves outcomes – particularly in sensitive or high-stakes settings.

Conclusion

The Phase II findings expand the story begun in Phase I. They demonstrate that disclosure is more than just a matter of motivation. It is also one of presentation: how people frame, justify, and linguistically construct their decisions. And crucially, those constructions may reveal not just what they did – but who they are.

Download the Report