Keeping our wits about us: introducing a bespoke informant interview model for covert human intelligence source (CHIS) interactions

ABSTRACT

The covert use of civilian informants leaves law enforcement agencies open to accusations of unethical conduct. The use of a structured interview protocol is a recognised method of promoting ethical interactions between police and public citizens, however, there is no known interview model specifically designed to meet informant handler objectives. The current study adopts a holistic view of the interaction between ‘informant’ and ‘handler’ to develop a bespoke informant interview model (RWITS-US: Review and Research, Welfare, Information, Tasking, Security, Understanding Context, Sharing). This model is compared to the PEACE model of interviewing as part of a novel experimental paradigm using mock-informants (N = 19), measuring levels of motivation, rapport, cooperation and intelligence gain. Results indicate that the RWITS-US model generated significantly greater levels of self-reported rapport without having any detrimental effect on the other measured variables. Whilst the results are encouraging, we suggest that the RWITS-US model should be tested in handler training environments before being recommended for widespread use in the field.

KEYWORDS:

• CHIS
• HUMINT
• informant
• intelligence interview
• RWITS-US

The use of civilian informants to assist in the proactive prevention and detection of crime is a widespread tactic employed across many policing jurisdictions (Loftus, Bacon, & Skinns, 2022). An informant could be described as any person providing information to a government agency regarding the actions or intentions of someone other than themselves with the expectation of confidentiality (Moffett, Oxburgh, Dresser, Watson, & Gabbert, 2021). The United Kingdom (UK) government provide a legal definition of an informant as a Covert Human Intelligence Source ([CHIS] Home Office, 2021), however, Loftus et al. (2022) argue that the covert status of an informant may provide a permissive environment for informant exploitation and unethical practices (see also Perry, 2021). For example, police officers often exploit a potential informants fear and naivety when recruiting them (Dabney & Tewksbury, 2016; Miller, 2011), whilst Dodge (2006) argues that a failure to fully understand the motivating factors of informants often results in a failure to protect them. Dodge (2006) provides several real-life examples of juvenile informants who have been murdered because of their status as a police informant, and it can be envisaged that many more go unreported.

The previous United Nations (U.N.) Special Rapporteur on Torture, Professor Juan Mendez recognised the importance of effective interviewing for ensuring ethical interactions between representatives of the state and their citizens, and was consequently inspired to compile a series of Principles on Effective Interviewing for Investigations and Information Gathering (hereinafter referred to as ‘The Mendez Principles’; Association for the Prevention of Torture [APT], 2021). The scope of The Mendez Principles are wide ranging, clearly stating that they are aimed at both law enforcement and intelligence officials, and defines an interview as, ‘ … a structured conversation where one person (the ‘interviewer’) seeks to gather information from another (the ‘interviewee’) as part of any investigation or intelligence operation' (APT, 2021, p. 1). This includes an interview with an informant.

Thus, The Mendez Principles maintain that an ethical interview ought to be a structured interview. Indeed, there is evidence from places such as Guantanamo Bay and Abu Ghraib that, when faced with the pressure of having to gather high-stakes intelligence without the benefit of a structured interview protocol, there is a risk that intelligence gatherers will revert to unethical practices (Alison & Alison, 2017; Brandon, 2014; Kleinman, 2006). Additionally, interviewing is recognised as a, ‘ … complex adaptive process' (APT, 2021, p. 13), which requires a considered and strategic approach based upon empirical research if interviewers are to achieve their overall objectives (APT, 2021; Fisher & Geiselman, 1992). Consequently, without a structured and strategic approach to interviewing, not only might interviewers revert to unethical practices, but they may also fail to maximise intelligence gathering opportunities. This is demonstrated by Hope et al. (2019) who found that without a structured approach to assisted recall (in this instance, the Timeline Technique), even participants who had been briefed to adopt the role of undercover police officers and were fully aware that they would be subsequently interviewed about the event they witnessed, failed to provide a full and detailed account. There are other interview objectives such as detecting deception (Moffett et al., 2021) that are also likely to benefit from a more structured approach to the interview process.

The Mendez Principles were developed with the assistance of a 15-strong Steering Committee comprising of global academic experts in the field of psychology, criminology, law and human rights. However, despite their strong support for structured interviewing, evidence suggests that practitioners do not employ a structured or strategic approach when interviewing informants (Dabney & Tewksbury, 2016; Nunan, Stanier, Milne, Shawyer, & Walsh, 2020c; Potts, 2009). In recognition of this, researchers have previously suggested that intelligence practitioners should use the existing PEACE model of interviewing (Evans et al., 2013; Nunan, Stanier, Milne, Shawyer, & Walsh, 2020a, 2020b). However, the current authors argue that PEACE may not be entirely suitable when interviewing informants, and we propose the use of a new (and bespoke) informant interview model, using the acronym RWITS-US (Review and Research, Welfare, Information, Tasking, Security, Understanding Context, Sharing), which can be remembered by the axiom: Keeping our wits about us.

The PEACE model

PEACE is a mnemonic acronym for the five stages of an investigative interview model developed in the early 1990s (Planning and preparation, Engage and explain, Account, clarify and challenge, Closure and Evaluation [CPTU, 1992a, 1992b]), each of which is outlined by the College of Policing (CoP) for England and Wales (CoP, 2020). The planning and preparation phase is designed to take account of all available information and to consider all key issues and objectives, including the use of a questioning strategy. The engage and explain phase is used to: (i) commence the establishment of rapport with the interviewee (see Gabbert et al., 2021 for a review); (ii) outline the purpose of the interview including objectives and how these will be met, and; (iii) outline any expectations the interviewer may have of the interviewee. The next phase is used to obtain the interviewee’s account using appropriate questions, the use of empathy and open mindedness (see Oxburgh, Myklebust, & Grant, 2010; Oxburgh, Ost, & Cherryman, 2012, 2014), plus the clarification of points raised. The interviewer will close the interview by summarising the account and providing the interviewee with a final opportunity to add anything further before explaining what will happen next. Once the interview has concluded, the interviewer should evaluate the interview to assess how the information gained fits into the wider investigation as well as assessing their own performance (CoP, 2020).

There is no doubt that the PEACE interview has considerable advantages, including flexibility and non-coerciveness, whilst utilising impartiality and an information-gathering approach, and is designed to obtain detailed information that can be used to direct criminal investigations (see Oxburgh & Hynes, 2016). Not only is it responsive to the information provided by the interviewee, but memory enhancing techniques can also be incorporated into the model (Dando, Wilcock, Behnkle, & Milne, 2011; Fisher & Geiselman, 1992; Ministry of Justice [MoJ], 2022). However, whilst the CoP (2020) recommend using PEACE when interviewing victims, witnesses and suspects of crime, they make no mention of informants, and it may be that the unique situational dilemma faced by informants, as individuals who have been tasked to obtain information through social interaction (Home Office, 2021), makes the PEACE model unsuitable for informant interviews. Furthermore, there is evidence that handlers must manage a range of interview objectives (Henry, Rajakaruna, Crous, & Buckley, 2019; Moffett et al., 2021) that may not be satisfactorily addressed by the PEACE model.

Interview objectives of an informant handler

Moffett et al. (2021) conducted a survey of informant handlers in England and Wales and found that they were required to consider a number of interview objectives, including understanding the infomants motivation, building and maintaining rapport, gaining informant cooperation, obtaining information and detecting deception. Importantly, these objectives are interconnected (Moffett et al., 2021), and therefore cannot be considered in isolation. Consequently, any suitable informant interview protocol must be able to maximise benefits across a range of requirements.

Informant motivation

Within the extant literature-base, informants are generally considered to be motivated by external factors, that is, factors external to the relationship itself (see e.g. Billingsley, 2001; Miller, 2011). Within the United States of America (USA), many informants appear to be negatively motivated, seeking to avoid the undesirable consequences of prosecution (Dabney & Tewksbury, 2016; Dodge, 2006; Miller, 2011). However, in the more legislated environment of the UK, the majority of informants appear to be positively motivated, seeking to gain something (i.e. money, revenge etc.) from their status (Billingsley, 2001). Stanier and Nunan (2021b) identify a range of potential informant motivations, which can be remembered by the mnemonic FIREPLACES (Financial; Ideological; Revenge; Excitement; Protection; Lifestyle; Access; Coercion; Ego; Sentence; Stanier & Nunan, 2021a). The inclusion of motivating factors such as ideology and ego indicate that, despite the dominance of external factors, informants can also be motivated internally; that is, the relationship itself can become a motivating factor for some informants (Billingsley, 2001; Birkett & Pike, 2017; Dabney & Tewksbury, 2016). Consequently, apart from simply seeking to understand informant motivation, handlers may also seek to exert a positive influence on their informants, to enhance their motivation through a positive encounter.

Building rapport

Rapport can be understood as the component of an interaction that allows an engagement to progress smoothly, minimising conflict and maximising shared understanding (Neuman & Salinas-Serrano, 2006). The PEACE model proposes that rapport ought to be established within the engage and explain phase of the interview (CoP, 2020). However, this presents a rather functional interpretation of rapport, in that shared understanding is promoted, and conflict minimised, by defining roles and interview objectives. In other words, rapport is envisaged as a means of achieving the function of the interview itself. Rapport is contextual though, and what constitutes sufficient and appropriate rapport in one context is not necessarily sufficient, or even appropriate, in another. This is recognised by informant handlers, who adopt a more long-term and relational attitude towards rapport (Henry et al., 2019; Nunan et al., 2020a). Rather than the functional rapport encapsulated within the PEACE model, this kind of relational rapport needs to be transposable across multiple meetings for the duration of the informant-handler relationship and is often expressed by handlers as a genuine concern for the informant’s wellbeing and welfare (Birkett & Pike, 2017; Moffett et al., 2021; Nunan et al., 2020a; Stanier & Nunan, 2021b). Observational studies and anecdotal accounts provide evidence of this approach, whereby an investment in relational rapport (demonstrated by a concern for the informant's welfare) produces greater cooperation and information later in the relationship (Dabney & Tewksbury, 2016; Hess & Amir, 2002; Yousef & Brackin, 2010).

Gaining cooperation

Within the realm of investigative interviewing, cooperation could be inferred by the revelation of pertinent information (Alison, Alison, Noone, Elntib, & Christiansen, 2013, Alison et al., 2014). However, it is clear from the UK’s definition of a CHIS (Home Office, 2021), as well as anecdotal accounts (i.e. Storm, Lister, & Cruickshank, 2015; Yousef & Brackin, 2010), that an informant deployment will progress through a series of phases: (i) an informant will be tasked to obtain target information from or about new or existing associates; (ii) an informant will actively obtain target information through social interaction, and; (iii) the informant will be expected to disclose target information during a de-brief with their handler. Consequently, handlers will need to gain sufficient levels of cooperation from their informants to ensure the successful completion of each of these stages. Moffett et al. (2021) found that considered tasking and the use of covert tradecraft techniques were essential for practitioners seeking to gain their informant’s cooperation, however, the tasking and security of informants is not encompassed within the PEACE model.

Obtaining information

The PEACE model is an information-gathering interview protocol (CoP, 2020) that can also incorporate memory enhancing techniques (MoJ, 2022). However, the phases employed are primarily designed to elicit autobiographical memories of unexpected events (Fisher & Geiselman, 1992; Geiselman, Fisher, MacKinnon, & Holland, 1985). For an informant who has been tasked to obtain information through social interaction, such as a legally defined CHIS (Home Office, 2021), the target information will not have been acquired unexpectedly and, rather than being stored in autobiographical memory, it is more likely to be stored in social memory. Social memory can be described simply as, ‘ … recalling people we know' (Hills & Pachur, 2012, p. 218) and refers to memories regarding social and group dynamics (Olick & Robbins, 1998; Van Bavel & Cunningham, 2012).

Blank (2009) argues that stereotypes, bias and self-identity provide a cognitive schemata for the formation of social memory that is often ignored during effortful retrieval. There is a naturalistic method of communicating social memory though, through the use of gossip (Dunbar, 2004; Foster, 2004; McAndrew & Milenkovic, 2002). Gossip has been defined as an, ‘ … exchange of information about absent third parties' (Foster, 2004, p. 81), or information, ‘ … concerning the personal matters of a third person who is not present' (Wert & Salovey, 2004, p. 122), with such communications often done covertly (Foster, 2004). Despite the negative connotations associated with gossip, Feinberg, Willer, Stellar, and Keltner (2012) observed that gossip can often have a prosocial motivation, providing a means of regulating behaviour within communities. It is clear from such definitions that the academic conceptualisation of gossip reflects the situational dilemma of an informant providing information about criminal associates to their handler. Furthermore, gossip is a form of storytelling, and story schemas have been shown to assist memory retrieval, providing a familiar framework to reconstruct an event (Delgadillo & Escalas, 2004; Massa & Simeoni, 2014; McGregor & Holmes, 1999). Consequently, there may be mnemonic benefits to encouraging an informant to engage in gossip.

Detecting deception

Broadly speaking, academic researchers espouse the cognitive approach (increasing cognitive load on the interviewee) as the most effective approach when attempting to detect deception during an interview (Leins et al., 2012; Shaw, et al., 2014; Vrij, 2014; Vrij, Fisher, & Blank, 2017; Vrij & Granhag, 2014), however, this may not be the most suitable approach when interviewing an informant. For example, the cognitive approach is concerned with eliciting and observing presumed cues to deceit, but lies that do not concern a personal transgression rarely produce observable cues (DePaulo et al., 2003). Given that a deceitful informant will be providing an account about the transgressions of a person of interest (rather than themselves), employing cognitive load techniques designed to elicit observable cues may not be the most successful strategy. Furthermore, when examining the effects of embedded lies (i.e. a single lie within an otherwise truthful account), Strofer, Ufkes, Noordzij, and Giebels (2016) found no differences in cognitive load between these types of liars and truthtellers. More worryingly, there is evidence that cognitive load may be the result of informant task characteristics. For example, Taylor et al. (2013) found that participants undertaking the role of a deceptive insider (i.e. an employee secretly reporting on the activities of their organisation to an outside agency) displayed verbal indicators of cognitive load. However, whilst the insider was considered deceptive from the employers perspective, from the perspective of the outside agency the insider was successfully undertaking the role of a tasked informant. Consequently, indicators of cognitive load were associated with normal informant behaviour.

Whilst the cognitive approach may not be suitable with informants, Moffett, Oxburgh, Dresser, and Gabbert (2022) were able to manipulate deception within a simulated informant paradigm (incorporating the three stages of tasking, social interaction and de-brief) and found that the way in which informants presented their narrative role differed between truthful and deceptive participants, with deceptive content frequently co-occuring alongside content consistent with low-potency narrative roles. Although deception was not manipulated within the current study, the method adopted by Moffett et al. provides handlers with a non-interventionary approach to detecting informant deception based on narrative analysis, and can therefore be employed following the elicitation of a normal narrative account. Labov and Waletzky (1997) identified a normal narrative as consisting of six functional elements: (i) abstract; (ii) orientation; (iii) complicating action; (iv) evaluation; (v) resolution, and; (vi) coda (which is an opportunity to re-orient the conversation to present circumstances; see Reissman, 2008 for a full description). Therefore, handlers should seek to explore each of these narrative elements to increase their opportunity to detect deceit.

Introducing the RWITS-US model

The RWITS-US interview model was designed with consideration for both the unique operational context of an informant, and the range of interconnected interview objectives of an informant handler. Additionally, the RWITS-US model takes account of all aspects contained in, and is compliant with, The Mendez Principles (APT, 2021). Each stage of RWITS-US is outlined below:

Review and Research: This is similar to the planning and preparation phase of the PEACE model but instead, focusses on reviewing previous contacts with the informant and their access to target information (Atkinson, 2021; Moffett et al., 2021). Additionally, research should be conducted on the informant themselves. Informant motivation is often linked to their current circumstances (Billingsley, 2001) thus, an understanding of an informant’s circumstances will likely result in a greater understanding of their motivation. It is foreseeable that an appreciation of past behaviour, existing relationships and current circumstances will also better prepare the handler to detect any potential deceit (Schirman, 2014; Vrij, 2008).

Welfare: The word welfare is indicative of the relational rapport that needs to be established between handler and informant if the handler is to achieve their intelligence objectives (Henry et al., 2019). Interpersonal relationships are developed by engendering a sense of autonomy (Alison et al., 2013), expressing empathy (Gabbert et al., 2021) and demonstrating a willingness to address concerns and expectations (Hargie & Dickson, 2004). Handlers should therefore begin an informant interview with a welfare enquiry. This will provide handlers with an opportunity to employ a number of interpersonal skills, including active listening and, where appropriate, personal self-disclosures (Hargie & Dickson, 2004; Marin & Gabbert, 2022) thereby increasing relational rapport. At a more practical level, a welfare enquiry at the outset will ensure that there are no apparent obstacles to progressing the interview.

Information: Social information (i.e. about associates and social relations) is often stored and recalled as a gossip narrative (Blank, 2009; Brown, Kouri, & Hirst, 2012; Van Bavel & Cunningham, 2012). Previous research has found that individuals are more likely to gossip if the subject is of high interest, their behaviour is anti-social and there is a pro-social motive for gossip (Feinberg et al., 2012; Wert & Salovey, 2004). Consequently, handlers should encourage informants to produce a gossip narrative by emphasising these features in relation to any particular person of interest (PoI) and by exploring the six functional elements of a normal narrative.

Tasking: Tasking an informant will provide them with clear steps that they need to take prior to the next meeting. At the conclusion of the tasking process, the informant must be prepared to undertake their task, attend to and encode target information, and subsequently re-engage with their handler. To facilitate this, the handler should provide the informant with a meaningful context for their tasking objectives (Anderson, 2009). Given the importance of narrative and narrative identity to the formation of social memory, this elaborated tasking may benefit from being presented in the structure of a normal narrative as outlined by Labov and Waletzky (1997).

Security: Unlike most witnesses or victims of crime, informants are covert, and handlers are responsible for ensuring their safety and security (Henry et al., 2019; Home Office, 2021; Moffett et al., 2021). It is foreseeable that the implementation of suitable security protocols would increase overall informant cooperation, therefore, security measures should be explained to the informant before they leave the meeting. One documented technique available to handlers is to provide informants with a cover story for their actvities (Birkett & Pike, 2017; Nunan et al., 2020a, 2020b; Stanier & Nunan, 2021b).

Understanding Context: Where possible, the information provided should be verified by post-interview intelligence checks (Nahari, Vrij, & Fisher, 2014) and examined in the context of the wider intelligence picture, to identify points of corroboration and/or contradiction. Additionally, the context of informant relationships ought to be considered. It is possible that the informant is being completely truthful and accurate in their reporting, but the relationship with the PoI is such that the informant has themselves been deceived or misled. Finally, narrative analysis can be employed to assist in the detection of any possible deceit, with Moffett et al. (2022) demonstrating that deceptive informants often express a low potency narrative role, such as a victim or a tragic hero.

Sharing: A review of four successful terrorist attacks committed in the UK in 2017 found that ‘ … improvement in analysing and sharing data' is required along with a willingness to ‘ … allow intelligence-derived knowledge to be shared more widely' (Anderson, 2017, p. 42). However, two years later both the security service and counter-terrorism policing faced identical criticism following a further terrorist attack (Weaver, 2021). Following this attack a report commissioned by Her Majesty’s Inspectorate of Constabulary and Fire and Rescue Services (HMICFRS) found an ‘ … unstructured approach' (HMICFRS, 2020) to the dissemination of information. It is clear from these reports that the sharing of intelligence is capable of saving lives. As such it ought to be incorporated into the interview process and handlers, controllers and analysts receiving intelligence should all take responsibility for identifying any other relevant departments or agencies whose decisions and actions may be informed by that intelligence.

Current study

The aim of the current study was to experimentally test a bespoke informant interview model (RWITS-US) against the existing PEACE model of interviewing in a mock-informant role-play paradigm. Given the exploratory nature of this research, no hypotheses were proposed, however, both models were tested against outcomes that are known to be important to informant handlers (e.g. motivation, rapport, cooperation, information; Moffett et al., 2021). The study was designed to ascertain whether a bespoke informant interview model could perform as well as, or even better than, the existing PEACE model on these measures within a realistic informant paradigm.

Method

Design

A single factor (interview), within-subjects design consisting of two levels (PEACE versus RWITS-US) was utilised. Paired sample t-tests were used to determine differences between the two interview approaches. The dependent variables (DVs) were: (i) motivation; (ii) rapport; (iii) cooperativeness, and; (iv) information gain. The two levels were counter-balanced between mock-informant participants with approximately half of the participants (n = 10) being allocated to the PEACE interview first, whilst the remainder (n = 9) were allocated to the RWITS-US interview first.

Participants

There were two groups of voluntary participants in this study and no incentives were offered:

Group 1: mock-informants (N = 19)

An online snowball sampling method was used for the recruitment of participants to this group. The research team circulated a link via email to various contacts who were known to have a research or professional interest in gathering information from informants. This initial purposive sample consisted of both academics conducting relevant research within universities and practitioners with either a law enforcement, military or non-government organisation background. These contacts forwarded the link to students and colleagues who they believed may have an interest in taking part. Participants were also able to forward the link on to any other associates. Nineteen participants (n = 11 male; n = 8 female) with a mean age of 25.79 years (range = 19–51 years; SD = 7.69) took part as mock-informants. Whilst the personal experiences / interests of the eventual participants were not recorded, given that the purposive sample included both academics and practitioners, it can be envisaged that the final sample consisted of both researchers / students as well as individuals with a professional interest or experience in human source intelligence. Whilst there is scant research to determine the socio-economic background of actual informants, the diverse recruitment of mock-informant participants was used to represent a relatively broad spectrum of society.

Group 2: mock-handlers (N = 3)

This group were specifically invited to take part by the authors. Mock-handlers needed to be convincing in the role of intelligence officers representing a government agency, therefore, participants for this role were purposively sampled based on their experience as current or former employees of a government agency and previous intelligence experience. Specifically, this group included a former UK police detective and two military personnel who were all experienced in the collection and analysis of HUMINT, either domestically or overseas. This criterion was adopted to minimise the impact of individual differences between mock-handlers. The lead author participated in the role of mock Person of Interest (PoI) for each experimental scenario.

Materials

A fictional case-study was produced to provide the basis for the role-play scenario.¹ The scenario revolved around the activities of a fictional PoI who was the presumed leader of a radical separatist movement, code-named the 25th Brigade. The notional aim of the 25th Brigade was for their local town (the fictional town of Stockfield) to secede from the rest of the UK and they were prepared to commit criminal acts to highlight their political cause.

Two separate non-public websites were built to facilitate the experiment which was conducted online to accommodate restrictions imposed as a consequence of the COVID-19 pandemic (HMG, 2020). The first was a booking site that provided participants with information regarding the experiment and an opportunity to book a time for the research. The second website consisted of thirteen tabs or pages containing all the information and links required to participate in the study. Embedded links to an online video conferencing platform (Zoom) were used for meetings between the mock-informant and the mock-PoI, and the mock-informant and the mock-handler. Zoom was selected for its widespread and ease of use and because it has previously been used successfully in other online experiments (see Archibald, Ambagtsheer, Casey, & Lawless, 2019).

Background information, including target information, was prepared for the mock-PoI. This was not a prescriptive script, and the mock-PoI responded naturally and dynamically to the mock-informant (based on previous field experience) to maintain ecological validity and a sense of realism for mock-informants (as active collectors of information). However, the mock-PoI ensured that all the target pieces of information were revealed at the correct stage of the scenario.

Mock-handlers were provided with two interview scripts: (i) PEACE, and; (ii) RWITS-US. It was acknowledged that interviews are necessarily dynamic and occasionally reactive (CoP, 2020; Pearse, 2009), thus, interviewers were permitted to react to mock-informants in a naturally conversational manner and to answer any questions they might pose, whilst maintaining their role as a mock-handler, should this become necessary. However, interview questions and the overall structure and sequence that they were presented were scripted for consistency and in adherence to the two interview models.

The DVs of motivation, rapport and cooperation were measured using a questionnaire embedded within the experimental website (see Table 1). Each question could be answered on a 7-point Likert scale (with 1 being ‘not at all’ and 7 being ‘completely’). Scores for rapport and cooperation were obtained by working out the average score from three questions designed to address each variable. Motivation was scored from a single question (intelligence gain was scored by the amount of target information actually reported).

Table 1. Questions used to measure the dependent variables of motivation, rapport and cooperation.

Variable	Question
Motivation	Please rate how motivated you were to perform in your role as a CHIS.
Rapport	Please rate how likeable the handler was. Please rate how much you felt you had in common with the handler. Please rate how easy you felt it was for you to talk to the handler.
Cooperation	Please rate how likely you would be to meet with the handler again. Please rate how likely you would be to complete further tasks for the handler. Please rate how much of the information you obtained you disclosed to the handler.

Procedure

Prior to data collection, full ethical approval was provided from the first author’s institution (Ref: 3986/2020). Initially, a pilot study was conducted to familiarise mock-handlers with the interview techniques and the online procedure. Mock-handlers were provided with a PowerPoint briefing on the two interview models and were afforded the opportunity to ask questions prior to the pilot study taking place. Two participants from a military organisation within the UK performed the role of mock-informant and completed the full procedure twice – this allowed four mock-handler participants to practice the entire procedure, including both interview techniques. Of the four mock-handlers who completed the pilot study, three progressed to the live data collection phase. The pilot study also allowed the experimenters to validate use of the website and online tools prior to live data collection.

From the booking site, mock-informant participants were provided background information regarding the study (a role-play scenario where they would be required to adopt the role of a mock-informant; meet a mock-PoI and mock-handler on two separate occasions, and that the procedure would last approximately 45–60 min in total) and were able to select an available time-slot. On completion of this process, participants received an automated email (managed by an online application) containing the link for the online experiment. At their elected time, participants clicked on the link for the online experiment and landed on the home page which repeated the information provided on the booking page as well as listing email contact information for the lead author. Having registered their consent, participants were informed that to complete the experiment they needed to work through the procedure as outlined in Table 2.

Table 2. Online procedure.

Written Briefing (1)	You are a registered CHIS who has previously reported on political extremist activity. A PoI is holding a virtual rally, you are tasked to attend the rally to find out more about their political activities, ideology, past actions and future intentions. You will need to contact your handler as soon as you finish the meeting.
PoI Meeting (1)	PoI reveals 7 items of target information: Name – RED; Email address – [e-mail] Group name – The 25th Brigade; Group aim – to secede from the national government; Local activities – raise awareness and attract new members; Past criminality – graffiti of government buildings; Future criminality – if no-one will listen PoI is prepared to engage in more serious offences.
Handler De-brief (1)	PEACE / RWITS Interview (counter balanced)
Questionnaire	See Table 1.
Written Briefing (2)	PoI has invited you to join a further online meeting. You are tasked to attend the meeting and find out more about their political activities and future intentions. You will need to contact your handler as soon as you finish the meeting.
PoI Meet (2)	PoI reveals 7 items of target information: PoI wants you to rent a van; PoI has kidnapped someone; Victim – A local councillor called Xander Wright; Meeting place – car park of a disused pub; Address – THE CHIMNEY SWEEP SB11 4LY Plan – to make a video with councillor demanding secession; Aim – to inspire people to join the cause.
Handler De-brief (2)	PEACE / RWITS interview
Questionnaire	See Table 1.

Each interview was live monitored by the lead author who scored how many items of target information (a maximum of seven for each interview) the mock-informant revealed. Mock-handlers were also aware of the target information and scored this concurrently. At the conclusion of each procedure, the lead author conducted a verbal de-brief with the mock-handler. Any discrepancies in scoring were discussed and agreement reached as to whether the participant gave sufficiently clear information to justify a positive score. Interviews were recorded so that these could be referred to in the event of disagreement – ultimately, this did not prove necessary as discussions did not centre around what was said by mock-informants, but whether it was sufficient to justify a score. Because mock-informant participants were briefed to adopt the role of an established and cooperative informant, errors and confabulations were not coded or scored and did not feature in any further analysis.

Results

Mean scores and standard deviations were obtained for the DVs of: (i) motivation; (ii) rapport; (iii) cooperation, and; (iv) information. Paired t-tests found that, with an alpha level of p > .05, there was a statistically significant difference between conditions for rapport, with mock-informants reporting greater levels of rapport in the RWITS-US condition (M = 5.56; SD = 1.01) compared to PEACE (M = 4.98; SD = 1.44), t(18) = 2.274, p = .035. Although there was no statistically significant difference between conditions for motivation, cooperation and information, results did indicate that there was a slight advantage in terms of self-reported cooperation and information gain within the RWITS-US condition, whilst self-reported motivation was higher in the PEACE condition. Full results, including mean scores and standard deviations (from 7-point Likert scales), can be found in Table 3.

Table 3. Results of paired t-tests on dependent variables.

	PEACE M (SD)	RWITS-US M (SD)	t (df)	p	d	95% CI
Motivation	5.79 (1.27)	5.63 (.90)	.615 (18)	.546	0.15	.381–.697
Rapport	4.98 (1.44)	5.56 (1.01)	2.274 (18)	.035	0.47	1.11–.044
Cooperation	6.02 (1.00)	6.18 (.95)	.856 (18)	.403	0.16	.545–.23
Information	5.95 (1.51)	6.00 (1.25)	.152 (18)	.881	0.15	.78–.674

Discussion

The use of covert tactics by government agencies is coming under increasing public scrutiny (i.e. the ongoing Undercover Policing Inquiry [2021] and Operation Kenova [2021] in the UK, the latter of which investigates the historic use and conduct of an IRA informant) with concerns being raised about the ethical use of informants (Dodge, 2006; Loftus et al., 2022). State-enacted interviews are recognised as permissive environments for unethical practice (APT, 2021) and it is therefore foreseeable that informant handlers will seek reassurance and transparency by employing evidence-based solutions to meet their operational requirements (Palmer, Kirby, & Coleman, 2019; Sherman, 2013). The current study tested the efficacy of a bespoke informant interview model (RWITS-US) against the existing PEACE model of interviewing in a unique mock-informant paradigm.

The PEACE model is a rapport-based information-gathering interview model that was developed in collaboration between academics, police officers and legal professionals (CPTU, 1992a, 1992b) and has evolved over nearly 30-years of academic research and practitioner usage (Bull, Valentine, & Williamson, 2009; Clarke & Milne, 2001; Walsh & Bull, 2012). The model is now widely employed by policing, law enforcement and other governmental investigative agencies in many countries (CoP, 2020; Oxburgh, Walsh, & Milne, 2011; Walsh & Bull, 2012; Walsh & Milne, 2008) and has previoulsy been recommended for use with informants (Evans et al., 2013; Nunan et al., 2020a, 2020c). As such, it is a relevant comparator for any new model to be tested against. However, in our novel experimental paradigm designed to simulate the unique situational and operational dilemma of an informant (as an active participant in the collection and dissemination of information), the PEACE model did not perform significantly better than the previously untested RWITS-US interview model. Indeed, the RWITS-US model actually performed significantly better than PEACE in relation to rapport, with slight advantages in terms of cooperation and information gain.

The importance of rapport to informant handlers has been consistently re-affirmed by practitioner surveys (Birkett & Pike, 2017; Henry et al., 2019; Moffett et al., 2021; Nunan et al., 2020a; Stanier & Nunan, 2021b), however, it was recognised by the current study that the functional rapport, established as part of the existing PEACE model, may be inadequate for the kind of long-term relationship building required of informant handlers. Therefore, the RWITS-US model was designed to increase relational rapport between handler and informant. The welfare phase was designed to provide the informant with a sense of autonomy, whilst presenting the handler with an opportunity to demonstrate concern through the use of interpersonal skills, such as active listening and appropriate self-disclosures (Hargie & Dickson, 2004; Marin & Gabbert, 2022). The handlers concern for the informants welfare was re-iterated during the security phase at the close of the interview. Additionally, the use of narrative prompts during the information phase was designed to induce a gossip narrative, and the act of gossip is also believed to increase social bonding.

Previous research indicates that increased rapport results in greater intelligence yield (Alison et al., 2013; Gabbert et al., 2021; Nunan et al., 2020d), however, this was not the case in the current study, with there being no statistical difference in information gain between PEACE and RWITS-US interviews. There are several possible explanations for this. Firstly, it is possible that rapport was not actually established as part of the RWITS-US model. Rapport was measured by the use of a short questionnaire, and whilst the use of questionnaires for measuring rapport is not unusual, the Rapport Scales for Investigative Interviews and Interrogation, Interviewee version (RS3i; Duke, Wood, Bollin, Scullin, & LaBianca, 2018) is more commonly used (see also Gabbert et al., 2021). Unfortunately, the RS3i is a relatively long-winded tool (21 items) and the current exploratory research already involved a lengthy procedure that was designed to measure multiple responses (i.e. motivation, rapport, cooperativeness and information). Therefore, the possibility exists that the questionnaire employed was an inappropriate measure of rapport, and rapport was not actually established; consequently, there would be no discernable difference in information gain.

An alternative explanation could be that only a minimal amount of rapport may be required to promote disclosure; the PEACE model is a rapport-based interview model, and the levels of rapport generated by the PEACE model, although statistically lower than the RWITS-US model, may have been sufficient to maximise intelligence gain, therefore, any further increase in rapport did not enjoy a corresponding increase in disclosure. A further explanation would be that the relationship between rapport and disclosure of information is highly complex. It could be argued that the relationship is indirect, and that rapport improves cooperation which in turn prompts disclosure (Evans et al., 2014; Kelly, Miller, Redlich, & Kleinman, 2013). Whilst self-reported levels of both rapport and cooperation were higher in the RWITS-US model than the PEACE model, levels of cooperation were not significantly higher. Consequently, a greater increase in levels of cooperation may result in more intelligence gain. Nunan et al. (2020d) conducted a field study of handlers and informants examining the relationship between rapport and information, and whilst they found a direct relationship between rapport and intelligence yield there are a number of other potentially confounding variables (such as length of relationship, nature of tasking, value of intelligence) that were unreported and which might correlate to both rapport and information. Indeed, Marin and Gabbert (2022) found that handlers were able to increase rapport with mock-informants through the use of self-disclosure, but this did not translate into greater intelligence yield; it is perhaps noteworthy that in their study, Marin and Gabbert controlled for cooperation by informing participants to strike a balance between being forthcoming whilst remaining loyal to their criminal associates. As such, the influence of rapport on the amount of information gathered from an informant is unclear and remains an understudied area of research. Overall, whilst there may be relational benefits to building rapport with an active informant (Birkett & Pike, 2017), rapport may not be as essential to information gain as it perhaps is in other interview contexts (Gabbert et al., 2021).

Although the RWITS-US model performed slightly better in some areas than the PEACE model, there are several limitations, and caution should be exercised when interpreting the results.

Strengths and limitations of study

Like all experimental research of this nature, there are both strengths and limitations. A major strength is that this is the first known study that has tested a bespoke interview model for use solely with informants against a known investigative interviewing model, whilst at the same time addressing a range of handler objectives (see Moffett et al., 2021). The paradigm used is both novel and unique in that it has been specifically designed to replicate the situational and operational dilemma of an active informant. Crucial to the development of this paradigm was the inclusion of three clear phases of an informant deployment: (i) receipt of tasking instructions; (ii) interpersonal engagement with a PoI, and; (iii) handler de-brief. Rarely have any other studies examining intelligence interviews utilised a paradigm combining each of these features (but see Moffett et al., 2022).

Nonetheless, it remains the case that this was exploratory research conducted in a laboratory setting during the Covid-19 pandemic, as such, it is hoped that the current research can continue to evolve and that more naturalistic paradigms can be developed and employed. Once the RWITS-US (or any other interview) model has been established as effective within a laboratory setting, it can then be progressed into field-based research before being recommended for practitioner use.

One possible limitation of the study is the conducting of online interviews using video-conferencing software which some may argue does not accurately replicate face-to-face engagements. However, internet-based communications with informants have been recommended for greater use in the future (Birkett & Pike, 2017; Stanier & Nunan, 2021b). Consequently, results from this study may provide some reassurance to practitioners and policy-makers looking to shift their communications toward emerging technologies.

The sample size (N = 19) could be viewed as a further limitation, however, this was exploratory research testing the efficacy of a new informant interview model (RWITS-US) through the use of a novel experimental role-play paradigm during the COVID-19 pandemic (which brought about enforced new ways of working). In particular, our study was designed to inform practitioner application and when conducting research of this nature, it is often preferrable to utilise a smaller sample whilst prioritising ecological validity and replicability (see Ziliak, 2019). Ultimately, caution must be employed when drawing overall conclusions from a single study of this nature. That said, our results provide researchers with some confidence to attempt replication and may additionally encourage practitioners to engage in the development of the RWITS-US informant interview model.

Implications for practice and future directions

This is the first study to propose and test a new interview model for interviewing informants, consequently, before any conclusions can be drawn regarding its scientific efficacy and robustness, further research is required. However, our results suggest that the RWITS-US interview model produces significantly greater levels of rapport when compared to the PEACE model, with slight advantages in other areas too. Further research should seek to replicate results across various modes of communication, especially those that are more commonly employed, such as telephone and face-to-face encounters (Birkett & Pike, 2017; Stanier & Nunan, 2021b).

Informant interviewing might also benefit from research into other areas affecting informant performance more broadly. For example, whilst the average amount of target information correctly recalled in this study was high in both conditions, there were a number of individual mock-informant participants who failed to recall the majority of the target details presented to them, despite the fact that there was no delay between encoding and recall. This suggests that there may be benefit in understanding what information tasked informants attend to and why; understanding this may not only inform elicitation attempts but may also influence the tasking process. It is also worth remembering that the RWITS-US model was designed for use on established cooperative informants as this appears to be the dominant situation in practice (Kleinman, 2006; Nunan et al., 2020a, 2020b), however, the process involved in recruiting informants and establishing that cooperative relationship is under-researched (Dabney & Tewksbury, 2016) and a greater understanding of this process could inform the implementation of the RWITS-US (or any other) interview model.

Moffett et al. (2021) found an overriding concern with detecting deception amongst informant handlers, and whilst deception was not manipulated in the current study, previous research has found that narrative analysis of a mock-informant account can assist in the identification of deceit (Moffett et al., 2022). Given that the RWITS-US model is specifically designed to elicit a gossip narrative, the detection of deceit may also benefit from its use, and this would provide another opportunity for future research.

Crucially, field research is also required. Fleming and Rhodes (2017) comment that evidence-based practice needs to be informed by and responsive to practitioner experience, therefore, the inclusion of practitioner feedback throughout the testing and development process will be critical to the successful evolution of the RWITS-US model.

Conclusion

This study tested a new bespoke informant interview model (RWITS-US) in a novel experimental paradigm designed to simulate the unique situational and operational dilemma of an active informant. Whilst further research is undoubtedly required, statistical anlysis suggests that the well-established PEACE model does not perform significantly better than the previously untested RWITS-US model on any of our measures, and there may in fact be benefits to using RWITS-US, especially in terms of rapport. Although these benefits may only be slight, within the context of applied research, slight benefits may well be sufficient.

This study does not claim to present a complete and indisputable informant interview model that can be foisted upon practitioners on the basis of a single study; rather, this study provides a rationale for further debate and a potential direction for future research. The role-play paradigm employed was designed to simulate the unique situational dilemma of an active informant and the PEACE model was specifically chosen as a relevant comparator. It is hoped that the detailed presentation of the RWITS-US model will inspire future researchers to attempt replication in the future. Importantly, whilst the RWITS-US model is conceived as a suitable informant interview model worthy of future research and development, it is recognised that this should be progressed as much as possible in conjunction with (and responsive to) practitioner engagement.

Disclosure statement

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

Notes

1 The fictional case-study is available from the first author upon request.