The use of self-disclosure to build rapport with mock covert human intelligence sources (CHIS)

ABSTRACT
 The use of covert human intelligence sources for the protection of national security is an important but under-researched area. In light of increased globalisation, and technological developments, examining covert intelligence gathering online may have operational relevance. This study examined the use of similarities in self-disclosure to develop rapport with mock covert human intelligence sources online. A sample of 66 participants were asked to assume the role of a CHIS, listen to an audio recording of a meeting of the criminal network they are part of, and strike a balance between being forthcoming to their handler while at the same time not revealing all information they possessed. They were then interviewed online by a mock handler who opened the conversation in one of three ways: self-disclosing information that highlighted similarities between herself and the CHIS, self-disclosing information that highlighted dissimilarities, or control (no disclosure). Participants completed a follow-up rapport questionnaire based on their interaction with the handler. The study found that mock sources rated rapport higher in the similarities condition in comparison with other conditions; however, it did not translate into a significant difference in the intelligence reported. The findings of this research and its policy and practice implications are discussed.


Covert human intelligence sources (CHIS)
Human Intelligence (HUMINT) is the collection of information from human sources (Hartwig, Meissner, & Semel, 2014). The spectrum of human sources is relatively broad, however, only the use of covert human intelligence sources (also known as CHIS, 'sources' or 'informers') will be germane to this study. Inthe UK, CHIS are legislated and defined under the Regulation of Investigatory Powers Act (RIPA) 2000 s26(8) (Home Office, 2018) if: (a) they establish or maintain a personal or other relationship with a person for the covert purpose of facilitating the doing of anything falling within (b) or (c); (b) they covertly use such a relationship to obtain information or to provide access to any information to another person; or (c) they covertly disclose information obtained by the use of such a relationship or as a consequence of the existence of such a relationship. Whilst this legislation is an Act of the UK Parliament, the Scottish context is covered by the Regulation of Investigatory Powers (Scotland) Act 2000 which makes similar provisions under s7(2). RIPA 2000 also designates different roles for the police officers working in a Dedicated Source Unit (DSU; Innes, 2000). A source handler is typically a police officer at the rank of detective constable responsible for proactively recruiting, developing, handling, and tasking CHIS to gather intelligence, recording the information supplied by them, and directing their day-to-day activities in accordance with relevant legislation and policy (Home Office, 2018). In parallel, source handlers have also a duty of care towards their sources and regularly meet with their sources for welfare checks (Nunan, Stanier, Milne, Shawyer, & Walsh, 2020a, 2020b, 2020c. In light of the CHIS(Criminal Conduct) Act 2021, which grants CHIS permission to break the law whilst deployed in the interests of national security or to prevent serious crimes from being committed, public concerns have been raised about their necessity. However, there have been many positive outcomes based upon intelligence from CHIS. The 2021 CHIS Factsheet Bill reportedly outlined that CHIS activity has been crucial in thwarting 'many' terrorist plots and has led the National Crime Agency (NCA) to disrupt over 30 threats to life, safeguard 200 people and remove 3000kg of Class A drugs and 60 firearms off the street (Home Office, 2021). The factsheet also mentions that CHIS operations have been the linchpin in the recovery of over 100 firearms and 400 other weapons, seizure of over 400 kg of Class A drugs, and over £2.5m cash by the Metropolitan Police (Home Office, 2021).
Unlike criminal interviews which are retrospective, the goal of HUMINT interviews is to elicit intelligence about intentions, plans, and future events (Hartwig et al., 2014). Such intelligence can play a fundamental role in the prevention of crime and the preservation of national security (Home Office, 2018). Yet, despite its critical importance, research on the techniques for gathering HUMINT from CHIS is still nascent. This is in part due to an overreliance on biased, unethical, and untested interviewing techniques derived from investigators' personal experiences rather than sound evidence-based interviewing approaches (Hartwig et al., 2014). For intelligence investigators, the threat of false intelligence is grave and has the potential of wasting finite intelligence resources.
Within the small number of studies dedicated to the topic of CHIS, relatively few have focused on how to interview sources to maximise the amount of intelligence elicited. In a pioneering study, Potts (2005) recruited 40 Royal Military Police trainees to act as mock informers. They were shown a four-minute video clip of an 'environmental anarchist workshop' in which they were tasked to infiltrate and report to their handler everything they saw the activists doing. Then they were sent one hour away for lunch to experience the natural delay of a 'real life' CHIS operation. Following the delay, they were randomly assigned to either an Enhanced Cognitive Interview (ECI; Fisher, Geiselman, Raymond, & Jurkevich, 1987) or a standard interview condition, and were interviewed by experienced handlers in the Special Investigation Branch of the Royal Military Police. Potts (2005) found a 58% increase in correct intelligence reported using the ECI over the standard interview.

Rapport in intelligence-gathering settings
In investigative settings, rapport is synonymous with 'operational accord' which means that the interviewer creates a productive relationship based on cooperation and respect and in which the interviewee's anxiety is minimised (Abbe & Brandon, 2013; see also Gabbert et al., 2021). Efforts to build rapport have been found to facilitate the reporting of accurate and detailed information (Collins & Carthy, 2019;Collins, Lincoln, & Frank, 2002;Nash, Nash, Morris, & Smith, 2016) and reduce the misinformation effect (Vallano & Compo, 2011). A significant milestone within the rapport literature was the creation of a rapport coding framework developed through careful analysis of interviews with high-value targets, known as ORBIT (Observing Rapport-Based Interpersonal Techniques; Alison, Alison, Elntib, & Noone, 2012). Alison, Alison, Noone, Elntib, and Christiansen (2013) utilised the ORBIT framework to analyse the audio and video footage of 418 terrorist interviews with 29 convicted suspects in the UK and Ireland between 2004 and 2010. Their results showed that maladaptive techniques such as patronising ('come on mate, it's for the best'), sarcastic ('I find that very hard to believe, don't you?'), or insulting responses significantly decreased the suspects' cooperation and increased their disengagement, silence, and 'no comment' answers. In contrast, humane rapport-building strategies (e.g. respect, empathy, acceptance, autonomy) were found to reduce suspects' counter-interrogation tactics and increase their cooperation. US military investigators also report favouring an empathetic approach, personalising the interview, displaying patience, cultural awareness, and good listening skills, over coercive and accusatorial tactics to elicit actionable intelligence (Redlich, Kelly, & Miller, 2014;Russano, Narchet, Kleinman, & Meissner, 2014).

Rapport in CHIS settings
Whilst improvements on rapport building with suspects have been made in the overt areas of policing, the covert area has largely escaped the attention of academic and police practitioners, with a few notable exceptions. Birkett and Pike (2017) set the scene by examining five NCA male handlers' perceptions of different communication methods (traditional face-to-face meetings vs computer-based interactions) and their impact upon rapport. Handlers reported preferring traditional face-to-face meetings because they could better read sources' body language, use active listening skills, and achieve rapport with them than in computer-based interactions. All handlers stressed the importance of developing the relationship. They achieved this by discussing hobbies and interests, getting their sources football tickets in exchange for intelligence, having a bit of 'banter' with them, and creating a relationship built on trust, respect, and a 'team ethic'. One handler discussed using self-disclosure to match the reciprocity of personal details to achieve rapport. That is, if the CHIS disclosed that they had children, so would the handler, while being careful not to reveal any private or personal details.
More recently, Nunan et al. (2020aNunan et al. ( , 2020b conducted structured interviews with 24 source handlers from various counter-terrorism DSUs across England. The first study (Nunan et al., 2020a) explored handlers' perceptions of interviewing their sources and the techniques used. Most handlers recognised the importance of rapport-building strategies, alongside other best-practice techniques such as using open-ended questions, and cognitive techniques to facilitate retrieval. The second study (Nunan, Stanier, Milne, Shawyer, & Walsh, 2020b) focused solely on handlers' perceptions of the rapport-building behaviours employed. Rapport was regarded as essential for building and maintaining a CHIS relationship, but also in getting the sources to talk. Handlers reported displaying empathy, using motivational interviewing techniques, active listening, making eye contact, and mirroring sources' nonverbal behaviour where appropriate. Most importantly, the handlers compared their relationship to a 'friendship' built on establishing common ground, showing interest in the sources' life, remembering personal details about them, and showing empathy and genuine care. This mutual trust was believed to be associated with an increased willingness to reveal information. A third study by Nunan and colleagues (Nunan et al., 2020c) analysed the use of rapport behaviours (e.g. attention, positivity, and coordination) in 105 real-life telephone conversations between seven handlers and their informers in one English police force. Here, a positive relationship was found between rapport and the amount of intelligence elicited from sources.
A common theme voiced by practitioners in Nunan et al.'s work (see also Moffett, Oxburgh, Dresser, Watson, & Gabbert, 2021) is the lack of evidence-based handler training in interviewing and rapport-building techniques. Nunan et al. (2020aNunan et al. ( , 2020bNunan et al. ( , 2020c recommend the development and implementation of a rapport-building and intelligence interviewing training programme for handlers. However, empirical research with mock CHIS is needed to better understand which rapport behaviours are the most effective in developing new sources, and maximising the intelligence yielded from them. Based on a survey of 205 officers working in CHIS management and handling, Stanier and Nunan (2021a) noted that the COVID-19 outbreak restrictions have caused a shift in CHIS-handler interactions, with many now taking place over the telephone. This shift in communication coupled with the economic and social uncertainty the pandemic brought, has also allowed sources to reconnect with persons of interest and gather/ pass on intelligence more systematically and succinctly (Stanier & Nunan, 2021a). Some of the officers expressed hesitation to switch to phone recruitment due to staff shortages to process source referrals and assess sources' motivation and risk of operational deployment and due to the perception that this trial-and-error process could ruin 'months of hard surveillance work' (Stanier & Nunan, 2021a). Nevertheless, it can be argued that the this reluctance lies in the lack of evidence-based rapport-building training for handlers, and, consequently, an inability to know which techniques to use in different contexts. Non-verbal behaviours known to build rapport such as open body language, eye contact, and head-nodding cannot be used in audio contexts , so the most plausible technique remains self-disclosure. The current study seeks to contribute to this body of work by examining whether self-disclosure can be used to build rapport and thus help develop the handler-CHIS relationship.
Self-disclosure as an effective way to build rapport Self-disclosure appears to be the most consistent process by which people get to know each other (Evans, Meissner, Brandon, Russano, & Kleinman, 2010). Perhaps unsurprisingly then, a recent systematic review of studies examining the use of rapport in information gathering contexts found self-disclosure to be the second most used strategy to build rapport with interviewees . In police settings, it is recommended that officers engage in small talk to equalise the power dynamic (Ministry of Justice, 2011). Work in Conversation Analysis reveals that police officers use self-disclosure to alleviate the interviewee's anxiety and pursue admissions from suspects e.g. 'I've got a girlfriend myself' (Stokoe, 2009), 'If it makes you feel any better, I do feel a little bit nervous' (Childs & Walsh, 2017). In a therapeutic context, Henretty, Currier, Berman, and Levitt (2014) reviewed 53 studies that examined therapists' self-disclosure and found that self-disclosure was used to foster the therapeutic relationship, validate reality, equalise power, and counter clients' internalised hatred and shame. They also found that clients had a more favourable perception of therapists who disclosed information that revealed similarities between them and in return, they were more likely to reciprocate by disclosing personal information about themselves.
In the empirical literature, after exposing participants to a mock-crime video, Vallano and Compo (2011) assigned the witnesses to either a rapport or no rapport condition before asking them to recall the mock crime. In the rapport condition, the interviewers were trained on how to disclose personal information to match that of the interviewee. For example, once obtaining interviewee disclosure regarding their family, the interviewer would follow up by self-disclosing information about themselves (e.g. 'I have two brothers and sisters who live in California'). Whilst rapport did not have any influence on the number of correct details recalled, it decreased the amount of total inaccurate information reported. In a subsequent study, Kieckhaefer, Vallano, and Schreiber Compo (2014) used two different scripts to manipulate rapport (low vs high rapport). The low rapport script modelled the opening phase of a typical police interview and included questions such as 'What is your first name? Where do you live?' whereas the high rapport script encouraged participants to disclose personal information e.g. 'Tell me about your family' that was reciprocated by the interviewer sharing similar information. The use of the high rapport script reduced participants' anxiety, but again, did not result in an enhanced witness recall.
It is possible that not all self-disclosure is equal, and its effectiveness depends on the type of self-disclosure shared. Based on the therapeutic literature, Ruddle and Dilks (2015) proposed five categories of self-disclosure: (1) reactive (client asks) vs voluntary (therapist initiates), (2) positive vs negative, (3) with varying degrees of intimacy (e.g. 'I admire people like you' vs 'I admire you'), (4) with varying degrees of personal information (e.g. 'My cat didn't come home once and I felt anxious' vs. 'I lost my partner in a car accident') and (5) similar ('I also felt anxious when I was unemployed') vs. dissimilar ('I don't know what's like to be unemployed') to the client experience. Though empirical data is lacking, Ruddle and Dilks (2015) argue that the nature of self-disclosure may have different effects; whilst similar and positive disclosure may moderate intimacy, dissimilar or inappropriate disclosure (e.g. 'I'm off to the Bahamas for the holidays') may induce envy, distress, anxiety or blur the lines of professional boundaries.

Study rationale
The current study seeks to build on the literature summarised above to test empirically if the technique of self-disclosure can be successfully operationalised in an online interaction to build rapport with mock informers and elicit intelligence from them. Further, the study seeks to shine new light on which type of self-disclosure is the most effective highlighting similarities or dissimilarities between the handler and the source. Unlike in Nunan et al.'s (2020aNunan et al.'s ( , 2020bNunan et al.'s ( , 2020c studies where the ground truth of the intelligence provided could not be established, a bespoke paradigm was created. In this study, participants were asked to assume the role of a CHIS, listen to an audio recording of a meeting of the criminal network they are part of, and strike a balance between being forthcoming to their handler and not revealing all information they had as this could put their lives in danger. It is expected that: H1. Participants in the similarities condition will share more units of accurate intelligence than participants in the dissimilarities and control conditions. H2. Participants in the similarities condition will share more fine-grain details than participants in the dissimilarities and control conditions. H3. Participants in the similarities condition will rate the rapport with the handler higher than participants in the dissimilarities and control conditions.

Design and participants
The study utilised a between-subjects design with three conditions (use of disclosure to highlight similarities, use of disclosure to highlight dissimilarities, and control) to which participants were allocated randomly. The independent variable was the nature of disclosure. The dependent variables were the number of units of intelligence shared (including accurate fine-grain and coarse-grain details, and inaccurate details), and self-reported ratings of rapport with the handler.
A power analysis using G*Power with a large effect size (0.40), an alpha level of 0.05, and a power of 0.80 indicated that a sample of 66 participants were needed (Faul, Erdfelder, Lang, & Buchner, 2007). Participants were recruited through word of mouth, advertisements placed on social media (e.g. Facebook, LinkedIn, Call for Participants, and the Policing Insight Twitter page) and on an internal police system called DutySheet. A total of 68 participants took part in the study, however, two participants were excluded because they could not remember any details from the audio recording. Therefore, data from 66 participants between the ages of 19 and 58 (M = 31.45; SD = 9.68; 44 females, 21 males, and 1 other gender) were included for analysis. The ethnic composition reflected that of the general UK population, with 30 people identifying as White British, 24 as White Other, 5 as Asian British, 3 as Black British, and 4 as Mixed ethnic. It is argued in the literature that real-life sources often come from backgrounds with low educational attainment, so the highest educational level achieved by the current sample was also measured: 12 people had completed secondary education, 31 had a Bachelor's degree, 21 had a Master's degree, and 2 people had a PhD. Participants received a £10 Amazon voucher to thank them for their time and engagement.

Materials
Audio recording of a staged encounter of a criminal network The stimulus comprised a bespoke audio recording that depicted two Russian Federal Security Service (FSB) agents discussing the assassination of another Russian informer on British soil. The audio lasted 4 min 43 s and was recorded from the point of view of the CHIS (the participants of this study). This was achieved by the two actors introducing each character at the beginning of the recording e.g. 'Hi, I will be Alexei, please put yourself in my role for this task'; 'Hi, I am Dimitri, your Russian friend.' The audio was inspired by notorious events in the UK involving Russian agents, however, no real names, dates, or locations from those events were used. The audio begins with Dimitri informing Alexei of his knowledge of another Russian spy (Ivan) working for the British government and threatening Alexei to participate in the plot to kill him. It then proceeds with Dimitri telling Alexei how and when Ivan will be assassinated and giving him instructions to pick up the assassin from the airport and help him blend in with the local public. The script was written by the researcher and the audio was performed and recorded by actors.
Self-disclosure questions Five questions from Aron, Melinat, Aron, Vallone, and Bator's (1997) intimacy study were used to encourage self-disclosure from the participants and to facilitate the manipulation of similarity or dissimilarity via reciprocal self-disclosure from the handler. The questions were as follows: What is your favourite snack or food and why?;'If you could have dinner with anybody in history, dead or alive, who would it be, and why?'; 'Imagine we are not living with COVID-19 restrictions, and you could travel anywhere in the world, where would you go and why?'; If you could learn one new personal skill, what would it be and why?; Have you regifted something that you received as a gift? What was it?

Rapport questionnaire
The rapport questionnaire used in this current study was adapted from the Rapport Scales for Investigative Interviews and Interrogations, Interviewee Version (RS3i) by Duke, Wood, Bollin, Scullin, and LaBianca (2018). The RS3i is a 21-item self-report questionnaire intended to measure interviewees' experience of rapport in intelligence interviews, perceptions of the interviewer, and willingness to communicate information. It comprises 6 Scales (Attentiveness, Trust/Respect, Expertise, Cultural Similarity, Connected Flow, and Commitment to Communication) and has good internal reliability and construct validity. Given the present study was conducted in an audio online setting, only 8 out of the 21 items were deemed relevant and applicable and were thus selected e.g. 'I think that the handler can generally be trusted to keep her word'; 'The handler and I have a lot in common'; 'Based on our conversation today, I would be willing to report again to my handler'. These items were scored on a Likert scale from 1 (strongly disagree) to 5 (strongly agree) and produced a total score between 8 and 40 where higher scores indicate greater rapport achieved. Although items from different scales were culled together, the adapted rapport scale showed high internal consistency (Cronbach's α = .86).

Procedure
The study took place online. Before taking part, participants were emailed a link to an online questionnaire that provided instructions and information about the study, along with a summary of participants' rights to withdraw, confidentiality, and anonymity of data. On providing informed consent to take part, participants were asked to respond to a series of demographic questions. Once they had filled out these questions, they were given more information about their CHIS role: 'For the purpose of this task please imagine that you are someone called Alexei (…) Since your recruitment as a CHIS, you have not met your handler yet. However, your instructions are to attend the regular Café Concertto meeting and then wait for your handler to contact you online, through a Zoom link. During the chat with your handler, you should be as forthcoming as possible, but at the same time, you need to stay loyal to your FSB friends, and not share all the information you have, as this could put your life and the lives of your family at risk!' This 'split loyalties' dilemma was further enforced by telling participants that if they adhered to the instructions, they would receive three entries to an Amazon prize draw whereas if they failed to adhere to the instructions, they would only receive one entry. In reality, everyone received only one entry. Following these instructions, participants were asked to listen to the audio recording and contact the researcher via email once they had finished. The researcher responded by sending an invitation to a Zoom meeting that was scheduled to take place after a 20-minute break. Participants were invited to distract themselves from the task by doing something enjoyable during this time, after which they would talk to their handler (the researcher).
Participants had been told beforehand that the interview conducted via Zoom would be audio recorded for subsequent analysis and were asked to ensure all auditory distractions in their environment were minimised for the duration. Participants were randomly allocated to each condition (highlighting similarities, highlighting dissimilarities, or control). A written interview guide was scripted for the Researcher to use to ensure consistency, including precise instructions on how to greet the source, thank and reassure them that they will be protected for the duration of their collaboration with the police e.g. '(…) I appreciate that this situation is new to you and you might be fearful for both your life and the lives of your loved ones, and I would like to assure you that we are going to do everything in our power to protect you as long you'll be working with us.' In the experimental conditions, the handler began by saying: 'Before we crack on today, I thought we'd start with a few icebreakers that would allow us to get to know each other.' In the similarities condition, the handler disclosed information about herself that matched sources' answers to the series of five prepared self-disclosure questions. For instance, if the source disclosed that they liked would like to travel to a city location, the handler would compliment their choice and tell them that she has been there in the past. In the dissimilarities condition, the handler purposefully disclosed information about herself that was dissimilar to sources' answers to the series of prepared disclosure questions. For example, if the source reported their preference for a city location, the handler would respond by saying that they personally preferred the outdoors and natural surroundings. Following the disclosure manipulation, the handler ended the introductory part of the interview by saying either:'That's the end of the ice-breaker task. It's good that we know each other a little more now. I see that we have quite a lot in common!' (Similarities Condition), or 'That's the end of the ice-breaker task. I see that we don't have a lot in common, but at least we know each other a little more now' (Dissimilarities Condition). In the control condition, no disclosure questions were asked.
Following the manipulation phase, participants were asked to report, in their own words, all the details they could remember from the conversation between Dimitri and Alexei. Occasional open-ended prompts were used if participants were reluctant to provide information, for example 'Is that everything you remember?'Across all conditions, the handler displayed similar rapport behaviours while the participants were providing information, such as the use of active listening and backchannel responses. In this case, active listening will be defined as attentively listening to the mock sources' accounts, whereas backchannel responses asthe use of expressions such as 'yeah ', 'uh-huh', 'hmm', 'right' to convey interest in the mock sources' accounts. At the end of the interview, participants were emailed the online Rapport Questionnaire to complete based on their interaction with the handler. On completion of this final task, participants were fully debriefed with the actual aims and objectives of the experiment and were thanked for their participation.

Coding and scoring
CHIS interviews were audio-recorded and later transcribed to enable the coding process. Data were transcribed using the artificial intelligence software programme Otter (https:// otter.ai/) and then coded individually. A coding template was first developed based on the audio stimulus script to include only the vital units of intelligence. The coding scheme contained 35 units of intelligence, each worth one point. The number of withheld/forgotten intelligence units, and the number of incorrect/lie intelligence units, were also coded. Similar to Brewer, Vagadia, Hope, and Gabbert's (2018) study, each of the 35 units of intelligence reported were further coded as either a fine-grain detail if it was exact (e.g. 10 am, Lambeth, 13th of March), or a coarse-grain detail if it was vague (e.g. at breakfast time, south of the river, middle of March). To assess inter-rater reliability, 25 randomly selected interviews were coded by one independent rater who was blind to the aims of the study. Pearson correlations were computed for each coding classification:

Results
To examine the research hypotheses, separate ANOVA analyses were conducted to examine the amount, accuracy, and quality (grain-size), of the shared units of intelligence in each condition. A further ANOVA examined differences in mean rapport ratings between conditions. To evaluate the assumptions of a one-way ANOVA, data were screened for normality and homogeneity of variance. The coarse-grain and rapport variables were within the assumption of normality. Assumptions were violated for the remaining variables, and Welch F test was used as an equivalent test. One outlier was identified; the analyses were run with and without the outlier identified, but it had no impact on the results, so the present results are based upon a complete dataset.

Units of intelligence shared
The first hypothesis was not supported;participants in the similarities condition were not found to share more units of accurate intelligence than participants in the dissimilarities and control conditions, [F(2, 63) = 1.84, p = .173, η 2 = .093]. The second hypothesis was also not supported; participants in the similarities condition were not found to share more fine-grain accurate details than participants in the dissimilarities and control conditions, [F (2, 37) = .77, p = .471, η 2 = .04]. Nor was there a significant difference between the groups in the number of coarse-grain accurate details reported, [F(2, 41) = 2.7, p = .079, η 2 = .116]. Further separate Welch's ANOVAs were conducted to see if there was a significant difference between the groups in the number of inaccurate [F(2, 41) = .84, p = .438, η 2 = .039] and withheld [F(2, 36) = 1.74, p = .191, η 2 = .088] number of intelligence units reported, but these also failed to reach statistical significance. Please see Table 1 for mean units of intelligence shared in each condition.

Rapport
The third hypothesis was supported; mean ratings of rapport with the handler were significantly higher for participants in the similarities condition than participants in the dissimilarities and control conditions, [F(2, 63) = 10.5, p < .001, η 2 = .249]. Post-hoc Bonferroni tests revealed that those in the similarities condition experienced greater rapport than those in the dissimilarities condition (p < .001) and those in the control condition (p = .002). No significant differences in rapport were found between participants in the dissimilarities and control conditions.

Summary of findings and discussion
The purpose of the current research was to examine whether the technique of self-disclosure can be successfully operationalised in an online context to build rapport with mock covert sources and maximise the amount of intelligence elicited from them. Further, it examined whether self-disclosure was an effective means to build rapport, and if so, which type of self-disclosure was most effective. Hypotheses 1 and 2 were not supported as no significant differences between conditions were found regarding the number of shared and fine-grain units of intelligence revealed by the mock informers. Hypothesis 3 was supported as participants rated the rapport experienced with the handler significantly higher in the similarities than in the dissimilarities and control conditions. This study has demonstrated, for the first time, that the nature of self-disclosure impacts sources' perceived rapport with the handler. Rapport was perceived as higher when the handler self-disclosed similarities with the mock source, than when the handler self-disclosed dissimilaritiesor did not self-disclose any information. This was further accentuated by some sporadic comments made by two sources in the similarities condition: 'You're easy to talk to'; 'I feel very comfortable speaking to you so even though I am quite afraid, I think that you're quite comfortable speaking to.' It seems that mock sources in the present study were positively influenced by the experimenter's attempt to showcase the mutual affinity and positivity of the interaction. In contrast, highlighting dissimilarities appeared to have hindered rapport development; whilst not attempting any conversational pleasantries in the control condition made some participants feel 'uncomfortable' and considered the interview 'abrupt'. These findings confirm Ruddle and Dilks (2015) suggestion that not all self-disclosure is equal and provide empirical evidence that its effectiveness in facilitating rapport depends on the type of self-disclosure shared.
The present investigation did not find any effects of the type of self-disclosure on the number of shared and fine-grain units of intelligence reported. This is in line with findings from similar studies (Kieckhaefer et al., 2014;Vallano & Compo, 2011). However, given the promising findings that different types of self-disclosure influenced feelings of rapport with the handler in the current study, it is perhaps too early to suggest that self-disclosure may not facilitate the reporting of intelligence without further research. For example, similarities and dissimilarities might emerge not just through discussing interests, values, and aspirations, but also by inadvertently self-disclosing personal information Antaki, Barnes, and Leudar (2005). One such example is through accent, vocabulary, and pronunciation patterns which reveal information about the speaker's geographical origin, socio-cultural, and educational status, which can be either similar or dissimilar to those of the recipient. This type of inadvertent self-disclosure might be an immediate and salient means of highlighting similarities and common ground versus dissimilarities and disparity, that might, in turn, influence the disclosure or withholding of information. It is unfortunate that in none of the exploratory studies on handlers' rapport perceptions (e.g. Birkett & Pike, 2017;Moffett et al., 2021;Nunan et al., 2020aNunan et al., , 2020b was noted which type of rapport-building strategies were conducive to an increased amount of intelligence yielded nor the quality of that intelligence.

Limitations and future research
A key limitation of the current study that needs to be acknowledged is the artificial nature and lack of ecological validity. Real-life informers are more likely to be motivated by more than civic-mindedness and interest. Several studies (e.g. Atkinson, 2019;Charters, 2013;Innes, 2000;Stanier & Nunan, 2021b) document that motivations to inform include turning Queen's evidence in exchange for a lower sentence, financial reward, ideology, revenge, and protection. Albeit the current research could not incorporate these motivations, it did implement a 'split loyalties' dilemma as an incentive to balance reporting/ withholding information in exchange for a monetary reward. This paradigm has been used effectively by other researchers in similar studies (see Granhag, Montecinos, & Oleszkiewicz, 2015a;2015b;2020;Oleszkiewicz, Granhag, & Cancino Montecinos, 2014;Oleszkiewicz, Granhag, & Kleinman, 2017).
Another limitation is the background of the researcher/mock handler. The mock handler was a non-British female Post-Graduate student, with expertise in Forensic Psychology at Bachelor's and Master's levels, and with training in evidence-based investigative interviewing. At the time of starting data collection, the mock handler had been attested as a Special Constable in the Metropolitan Police Service and had undergone basic police training. However, only theoretical knowledge,the mock handler had no direct experience as a handler. Aside from the very few participants recruited via Duty-Sheet, the remaining sample was unaware of the experimenter's policing training. Still, given the discussion above regarding inadvertently self-disclosing information that highlights similarities or dissimilarities, it is possible that the researcher's demographics influenced the mock sources' willingness to provide information. Most of the mock sources recruited in this study were female and had a high level of educational attainment, thereby highlighting similarities despite the experimenter's attempt to showcase dissimilarities. Real-life sources tend to be predominantly men from low socioeconomic backgrounds (Billingsley, 2000), who hold stereotypical beliefs about women's societal role (Nemitz, 2013). This could jeopardise their willingness to unveil information about their criminal network's plans to a female handler. The scant open-source data on this topic suggests that the number of female handlers is lower than the number of male handlers -3 female vs 22 male handlers in Birkett and Pike (2017) and 7 female vs 34 male handlers in Moffett et al. (2021). Furthermore, the first encounter between a source handler and a source typically takes place within police custody, during prisoner debrief meetings, or during port/airport stops (Nunan et al., 2020a) where the status and educational differences between the handler and the source are more likely to stand out (Billingsley, 2000). Further research is required to establish if the current findings can be replicated with a more representative sample of handlers (older males) and mock sources (drawn from the lower socioeconomic strata). Further, real source handlers may enhance the perceived authenticity of the interaction by participants in some way.

Implications for policy and practice
The current research has the potential to add to a small but growing body of research that can scientifically inform a national handler training programme, as it has shown that highlighting similarities between the mock handler and the mock sources makes the sources perceive the handler in a more positive light than highlighting dissimilarities. The successful implementation of this technique over a series of meetings has the potential to lead to reliable intelligence being shared. Previous research (e.g. Colwell, Hiscock, & Memon, 2002;Griffiths, 2008;Walsh & Bull, 2012) also highlights that maintaining rapport-building behaviours throughout the interview not only leads to higher intelligence gathered but also increases the possibility of detecting deception. In the covert arena, CHIS practitioners have identified the need for a psychologically research-based handler training programme on how to build rapport with (new) sources and what interviewing techniques to use (e.g. Billingsley, 2000;Moffett et al., 2021;Nunan et al., 2020a;2020b;2020c). At present, CHIS interviewing in the UK is not contained within the Police and Criminal Evidence Act 1984 (PACE 1984) provisions regarding the questioning of suspects, which is why handlers are left improvising, inappropriately interviewing sources (e.g. not using open-ended questions, rushing, and interrupting the sources' accounts), or believing rapport is an innate flair which cannot be trained (Nunan et al., 2020a(Nunan et al., , 2020b(Nunan et al., , 2020c. This is not to say that handlers do not receive any training at all, but this tends to focus primarily on the logistics of CHIS handling rather than incorporating communication and interpersonal skills and interviewing techniques (Nunan et al., 2020a). Albeit the training curriculum of source handlers is not openly available in the academic literature, their training does involve interactive scenarios using role players as mock sources in the training exercises. Source handlers also complete their Professionalising Investigation Programme Level 2 (PIP2) during which they receive training in the enhanced cognitive interview (ECI) and the PEACE model. Despite this, the handlers interviewed by Nunan et al. (2020aNunan et al. ( , 2020b rarely mentioned using the full ECI as they could not remember how to employ it properly from their outdated PEACE training. To address the current lack of CHIS-specific training, Nunan et al. (2020aNunan et al. ( , 2020bNunan et al. ( , 2020c call for a standardised classroom and field training programme to be created, with regular refresher sessions to ensure its benefits do not fade over time.

Conclusion
From the scarce openly available research on this topic, it can be deduced that real-life handlers employ rapport-building strategies with their sources although their efficiency and conduciveness to intelligence elicitation are hard to quantify. Self-disclosure appears to hold a great promise for being included in a national research-based handler training programme on how to build rapport with and develop (new) sources. The current study has demonstrated that mock sources perceive the handler more positively when similarities between the two are highlighted as opposed to dissimilarities, or in comparison with no self-disclosure at all. Previous research has not isolated the technique of self-disclosure to examine its relationship with building rapport, nor has it examined different types of self-disclosure. However, further research is required to see if these results hold when more independent variables are added into the equation, such as differences in ethnic and socio-educational backgrounds, that might be found to moderate the relationship.

Disclosure statement
No potential conflict of interest was reported by the author(s).