Increasing Perspective-taking in Civil-military Influence Interactions Aiden Hoyle
11117907
Department of Psychology, University of Amsterdam Research master’s thesis
Supervisors: Dr Bertjan Doosje, UvA Dr Helma van den Berg, TNO Second reader: Dr Marc Heerdink, UvA
Abstract
During deployment, soldiers are commonly tasked with trying to influence the behaviour of local civilians (e.g. increasing cooperation, increasing information flow). Perspective-taking is a construct that has been propagated as able to enhance influence skills, and consequently, a soldier’s success in influencing civilians. However, evidence for this rests upon research that uses general, dyadic influence paradigms. These paradigms are not representative of the unique goals and scenarios found in a military environment, or how military interactions typically use interpreters to communicate with influence targets. This investigation sought to validate the assumption that perspective-taking increases influence skill when these factors are considered, by extending research paradigms to a military context. Two studies, using situational judgement tests and a branching virtual interaction, assessed perspective-taking’s effect on influence skill in military scenarios, and the impact of an interpreter on perspective-taking ability and subsequent influence success. Across both experiments, watching a
perspective-taking manipulation did not significantly increase influence skill, compared to watching a control manipulation. Furthermore, there were no significant differences in influence skill between participants whose virtual interactions were mediated by an interpreter and those whose were not. These results are discussed, methodological improvements are elaborated on and recommendations for future research are provided.
Increasing Perspective-taking in Civil-military Influence Interactions
The term ‘social influence’ refers broadly to the processes individuals use to form or modify the attitudes or behaviours of others (Mugny & Pérez, 1991, Cialdini & Goldstein, 2004). For military operations, influencing typically means engaging with host-nation civilians, local religious leaders or foreign coalition forces, with goals of increasing civil-military cooperation, ingratiating into the local community or reducing insurgency through an increased information exchange (Elkhamri, 2007). Soldiers with good influence skills should be able to accurately assess a target’s receptivity and consequently, appropriately select and apply various influence strategies. Further, they should be able to implement these techniques with high self-regulation, and accurate interpretation and forecasting of their target’s
reactions (Van den Berg, van Hemert, van Beurden, & van der Kleij, 2015).
Perspective-taking is a psychological construct which has recently been promulgated as instrumental in aiding military influence interactions (Van den Berg, van Hemert, van Beurden, & Jetten, 2016). Described as the adoption of a mindset that differs from one’s own (Hodges, 2008), perspective-taking is seen as a malleable ability, whereby the level an individual may take perspective can be increased or decreased (Selman, 2003; Gehlbach, 2004). Empirical research has demonstrated that operating at higher levels of perspective-taking can enhance specific influence skills (Galinsky & Mussweiler, 2001). For example, target perception can be improved through a lowered susceptibility to confirmation bias (Galinsky & Moskowitz, 2000) or stereotyping tendencies (Galinsky, Ku, & Wang, 2005), and behavioural forecasting can be improved through an enhanced ‘theory of mind’ - the interpretation of another’s actions through information gained by taking their perspective (Wellman, Cross, & Watson, 2001). Indeed, specific influence scenarios such as negotiation (Galinksy, Maddux, Gilin, & Carpenter, 2008) or persuasion (Carpenter, Green, &
Vacharkulksemsuk, 2016) have been found to be more successful for those operating at higher levels of perspective-taking.
Perspective-taking’s propagated utility in military influence interactions, therefore, is logical. Soldiers operating at higher levels of perspective-taking should be better able to select an appropriate technique for their target using their enhanced perception skills. They should also better able to apply the technique using their heightened interpretation, regulation and forecasting skills. This is a valuable notion: psychological constructs that can enhance the effectiveness of these challenging influence engagements have great potential for increasing the success of international military and security operations. However, perspective-taking’s propagated usefulness has yet to be empirically tested in a military setting, and there are a number of features that are prominent in and specific to military influence interactions that may affect how perspective-taking functions in these engagements. The two studies presented in this paper, therefore, endeavour to empirically investigate these aspects, and their effect on perspective-taking in military influence interactions.
Perspective-taking in a military context
The context in which military influence interactions take place is one aspect that can be considered. Although perspective-taking’s benefit for influence has been heavily
empirically researched, the military context is vastly different to the empirical contexts in which this general influence research is carried out (Mackay & Tatham, 2009). One could argue that this general influence research does not capture many of the distinct and unique contextual elements found in military influence interactions. This invites the question of how perspective-taking functions when these elements are taken into consideration.
Typically, soldiers operate under more stressful conditions than participants
and often work under tight time constraints. Further, they are usually operating in a country whose language is unfamiliar, and whose cultural protocols are often extremely far removed from their own. Conversely, participants in general influence research are often in
environments familiar to them, such as university laboratories, which are clearly
unrepresentative of the aforementioned elements of military influence interactions. Coupled with this, the influence goals soldiers are tasked with are often much more specific and unusual than general influence literature. For example, while general influence paradigms typically use relatively commonplace scenarios such as real-estate negotiations (Mussweiler & Galinsky, 2001), deployed soldiers have suggested the intricate task of extracting
information about the locations of explosives from civilians as an influence goal they encounter daily (Woolfe & Arrow, 2005).
This specificity of military influence interactions, and the context they take place in, is something that has become apparent in research. A survey of military advisors working with foreign coalition forces revealed that, generally, only influence techniques that
established authority or provided foreign coalition officers with positive role-models, were viewed as effective by the advisors (Ramsden-Zbylut et al., 2009). With coalition officers being highly unmotivated and language barriers obstructing communication, proactive influence techniques such as using rational arguments or making emotional appeals were found to be largely ineffective and often resulted in noncompliance. The specificity of influence techniques that were deemed effective highlights the unique conditions that the military works under. The authors further highlighted how imperative the advisors viewed target receptivity assessments, and called for these processes to be explored further - arguably an indication of the need for research into perspective-taking from within the field. Although this study’s scope was limited to just interactions with foreign military forces, it acts as strong evidence for specificity in military influence interactions.
Further research which surveyed deployed soldiers who encountered a wider variety of influence targets has bolstered this notion (Wolfe & Arrow, 2013). By broadening the sample to include interactions with civilians, researchers were able to develop a basic taxonomy of military influence techniques, using factor analysis to divide techniques into clusters that specifically represent military influence. These clusters included
power-differential techniques such as punishment or threat, resource techniques such as offering a reward, positive-feeling techniques such as suggesting that a target’s compliance would be appreciated by others, positive techniques such as suggesting a target’s compliance would indicate positive traits, and negative techniques such as suggesting noncompliance would indicate negative traits. While forming this taxonomy, the researchers themselves highlighted that while in a broad sense, military and general influence techniques are similar, military influence techniques are far more specialised than general influence technique taxonomies, and should be seen as distinct (for examples of general influence taxonomies, see Marwell & Schmitt, 1967 or Yukl & Tracy, 1992). This assertion again demonstrates the disparity between general and military influence techniques, and by extension, contexts. It further highlights the necessity of verifying that perspective-taking increases influence skills within these distinct techniques and contexts.
Clearly, the unique challenges that deployed soldiers face, and influence processes soldiers use, are not represented in general influence literature. Therefore, it seems imperative that before the idea of perspective-taking being useful for the military is propagated, its benefit is validated using scenarios that are representative of this highly specific and unusual context. Only then can one be confident that perspective-taking functions in the same way in military influence contexts as general influence contexts. Therefore, a confirmation of whether high-levels of perspective-taking increases influence skill in military influence scenarios, forms the first aim of this investigation.
Perspective-taking in interpreter-mediated interactions
Related to this, the specific physical structures of military influence attempts are also not represented in general influence paradigms (Ramsden-Zbylut et al., 2014). The majority of general influence literature focuses on dyadic, unmediated interactions, yet civil-military interactions are almost always mediated by an interpreter (Vieria, 2013; Foottit & Kelly, 2012). With the endorsement of perspective-taking as an effective tool for military influence hinging on this general, dyadic literature, a further question needs to be answered to confirm its utility. That is: how exactly does perspective-taking function in interactions mediated by an interpreter, and is this different from its functioning in dyadic interactions?
Interpreters are incredibly valuable for the military (van Dijk, Soeters, & de Ridder, 2010), and particularly for influence interactions, which without interpreters would all but cease to be possible (Pouligny, 2006). Yet despite this apparently important role, surprisingly little research has looked at the effect of an interpreter presence (Rosendo & Persaud, 2016). Research that has, has indeed found that interpreters have a significant impact on the content and manner of the communications they are facilitating. Instead of acting as literal
‘translation machines’ - providing a direct translation of what each participant in the interaction says - interpreters instead often act as ‘cultural mediators’, who use their
discretion to omit, alter or advise whenever they deem necessary (Hoedemaekers & Soeters, 2009; Bos & Soeters, 2006). Much of this mitigation stems from motivations to keep ‘face’, for both the soldiers and in the case of this research, the local Afghan citizens. Through this active role in the interactions, interpreters alter conversations in order to placate both interaction parties.
Yet consequently, one can question how well perspective-taking can operate in such situations. If the expression of important opinions and feelings are indeed being altered, the
true intentions of both interaction parties can become obscured and lost. Compound this with the fact that soldiers will be dividing their attention between two individuals instead of one, and it seems likely that the interpreter mediation may jeopardise the efficacy of a soldier’s social skills, and by extension, hamper the success of an influence attempt. It therefore seems likely that when interactions are mediated by an interpreter, soldiers will find perspective-taking more difficult, which will in turn reduce its benefit for influence skill. However, investigation is needed to determine the veracity of this assumption; this forms the second question of the current research.
The present studies
In examining these two questions: whether perspective-taking increases influence skill in military scenarios and how interpreters affect perspective-taking and its effect on influence skill, the present studies endeavour to illuminate how perspective-taking operates in a context and physical set-up that is more relevant to the military. To manipulate the level of perspective-taking, a well-used military method of perspective-taking was taught to
participants. This was contrasted against a control manipulation. Two individual studies were designed to address these two questions, sequentially.
To validate perspective-taking’s benefit for influence skill in a military context, situational judgement tests were administered before and after the perspective-taking and control manipulations. Here, it was expected, in line with previous literature into perspective-taking and influence skill (Galinsky & Mussweiler, 2001; Van den Berg, van Hemert, van Beurden, & Jetten, 2015) that the perspective-taking manipulation would enhance influence skill. Specifically, an interaction effect between the manipulation type and the point at which influence skill was measured was predicted. In this, from pre-manipulation to
improvement when they received a perspective-taking manipulation, compared to when they received a control manipulation (Hypothesis 1). This was tested in Study 1.
Furthermore, to assess the effect of having an interpreter mediate interactions, and the subsequent effect this has on perspective-taking in military influence interactions, a
branching virtual interaction assessment was created. Here, it was expected that interpreter presence would reduce the ability to perspective-take, and subsequently, reduce influence skill. Therefore, main effects were expected for both interpreter presence and manipulation type. In this, participants who received a perspective-taking manipulation were expected to display higher influence skill than those who received a control manipulation, but this influence skill was expected to be lower when an interpreter mediated the interactions than when they did not (Hypothesis 2). This was tested in Study 2.
Study 1
Method
Participants. To determine sample size, a power analysis was conducted using G-Power. Using a mixed design, and given an effect size of f = 0.2, an alpha of .05 and a power of .8, a required sample size of 52 participants was indicated.
To ensure sufficient statistical power, 60 participants (Mage = 21.1; SDage = 4.14; 52% female; 92% Dutch) were recruited. 31 participants received the perspective-taking
manipulation and 29 received the control manipulation. Every participant was a psychology student from the University of Amsterdam, and was remunerated with a 0.5 research credit. The experimental protocol was approved by the University of Amsterdam ethics committee.
Cultural familiarity. To assess cultural familiarity, participants were asked to rate
how familiar they felt with the Middle Eastern culture on a 7-point Likert scale.
Baseline perspective-taking. Trait perspective-taking was measured using a six-item
scale. This scale was a subsection of the ‘Interpersonal reactivity index’ (IRI) (Davis, 1980), and items measured the extent to which individuals inherently seek to take perspective in others (e.g., “I sometimes find it difficult to see things from another's point of view”). The IRI has a reliability of α = .78 (Konrath, 2013). The subsection of items used in Study 1 can be found in Appendix A.
Perspective-taking manipulation. To manipulate perspective-taking level, an
educational video was created for the purposes of this investigation, which described the ERCA method. The ERCA method was designed as a four-step process of increasing one’s level of perspective-taking. It was developed based on trainings used in previous military research, which have used similar techniques to increase perspective-taking (Mueller-Hanson et al., 2007). The ERCA method was designed to be simple to implement, memorable and able to be done on the spot, making it ideal for soldiers in field operations, and therefore this research. The video consisted of four steps, where participants were taught to evaluate a target, to reflect on their role in the influence interaction, to communicate with the target using perspective-taking and how to adjust the interaction based on the perspective they have taken. Each step was thoroughly explained. Participants were unable to progress while the video was playing. Participants were asked to tick a box indicating they had watched the video, which appeared at the end. To view the manipulation, click here.
Control manipulation. For the control manipulation in Study 1, a video which
described the military history of the Netherlands was created. The video was a similar length and style to the perspective-taking manipulation video. It provided a brief overview of the
formation and international endeavours of the Dutch military, such as the Eighty Years’ War or World War II. The subject of Dutch military history was selected as it was relevant to the study thematically, but the material was deliberately kept very simple so that it would not afford the control group any advantage. Participants were again unable to progress while the video was playing, and as in the perspective-taking manipulation, participants were asked to indicate that they had watched the video when it finished. To view the manipulation, click here.
Situational judgement tests. To measure the dependent variable, influence skill,
Study 1 used situational judgement tests (SJTs). SJTs have been effectively employed by studies evaluating military programs training interpersonal skills, such as cultural knowledge for cross-cultural negotiations (Kim et al., 2009), or social awareness for military leader influence engagements (Mueller-Hanson et al., 2009). The first step of an SJT is for participants to be presented with a situation and several possible responses, actions or interpretations. The participants are asked to provide ratings on a Likert scale, of how likely they would be to use the specific response given the situation. This is done for each response, and by doing so, participants indicate a pattern of judgement or way of thinking about the situation. This way of assessing was favourable in this particular research, as influence assessment is therefore not an accumulation of binary, correct or incorrect answers, but instead, incorporates how there is often no single correct way to influence a target.
The second step is to correlate a participant’s pattern of ratings with the average of ratings provided by subject matter experts (SMEs). SMEs are individuals who already possess high levels of the construct being assessed. In Study 1, the SMEs were individuals with an in-depth knowledge of military influence. Taking the average of their ratings for each response therefore provides a rating pattern that would demonstrate high levels of the
construct. By computing the correlation between the participants and this average SME rating pattern, one can measure the extent to which participants demonstrate this measured construct in their own rating patterns. Therefore, in this case, a strong, positive correlation coefficient would indicate that a participant’s ratings were in accordance with the ratings of the military influence SMEs, and therefore, their ratings indicate high levels of influence skill. These computed correlation coefficients, therefore, are how the dependent variable - influence skill - was operationalised.
For Study 1, eight SJTs were developed, focussing on real-life situations that soldiers may encounter in the field. Previous studies indicated that illustration of the scenarios
provided a better engagement with the SJTs than written text (Moore, 2010), so an artist was employed to draw the SJTs in a graphic novel style1 (see Figure 1 for an example). Each SJT was accompanied by four possible response items, creating 32 in total. The possible
responses focussed on the application of different military influence techniques, and were developed using previous literature on influence techniques used in theatre (Ramsden-Zbylut, Wisecarver, Foldes, & Schneider, 2010; Wolfe & Arrow, 2011). Featured techniques
included inducing a negative self-esteem in the target by suggesting that others would view them poorly if they did not comply, or positive trait technique, such as suggesting that if the target complied, they would be displaying a variety of positive personality traits such as intelligence or generosity. The full array of SJTs, the images used and the associated possible responses can be found in Appendix B1.
Figure 1. Two examples of SJTs that the participants viewed.
The possible responses to the SJTs were carefully developed to be sensitive enough to capture the effect of differences in influence skill. Although each response was plausible, some were more appropriate and likely to be used by the military than others, given the influence target. The development of this argumentation was aided by previous research into military influence techniques (Ramsden-Zbylut, Wisecarver, Foldes, & Schneider, 2010). Notes on the development and argumentation for each response can be found in Appendix B2. To ensure that the intended influence technique was clear in each of the responses, a small pilot test was conducted. Ten students2 read the 32 possible responses, and matched them to the influence technique they thought the response was demonstrating. Any response
2
that was not successfully matched with the correct influence technique was altered to enhance its clarity.
Finally, before administering the SJTs to the SMEs, individuals with previous practical experience of influence in military theatre reviewed each SJT. They reviewed the plausibility of the situation and the argumentation of the responses. Upon satisfactory reviews, the SJTs were administered to five SMEs3 who anonymously gave their ratings of the likelihood of using this response given the situation, on a 7-point Likert scale. Descriptive analysis of these ratings revealed that the most common rating that SMEs gave was 5
(somewhat likely). This suggests that the SMEs saw a lot of the possible responses as at least somewhat likely to be used by the military. Furthermore, generally the pattern of ratings matched the argumentation that was used to develop the responses; the response rated as the most likely to be used by SMEs was typically the one intended as the favoured response during the creation of the SJTs. This would suggest that how the SMEs viewed the SJTs matched the thought process used to develop the SJTs. This bolsters the notion that they are valid tools of assessment when investigating military influence. A table comprising of the means and standard deviations of the SME ratings can be found in Appendix B3.
To ensure satisfactory interrater agreement between the SMEs, average deviation indices (ADI) (Burke, Finkelstein, & Dusig, 1999) were computed for the possible responses4. If interrater agreement for a particular response was low, then its mean rating may have been misleading and an inaccurate representation of the response’s likelihood. The ADI computed the average of the absolute deviations of individual SME ratings from the average SME rating of a particular response indicating how much the SMEs were in
3
Study 1 used five anonymous researchers from TNO Defense and Security, whose areas of expertise and day-to-day research includes military influence interactions.
4
Average deviation index was preferred over other methods of assessing interrater agreement as it is the most common method of assessing interrater agreement in situational judgement test literature (Bledlow & Frese, 2009; Guenole, Chernyshenko, & Weekly, 2017)
accordance in their ratings. Burke and Dunlap (2002) recommend dividing the number of possible Likert options for each response, in this case seven, by six to get a cut-off value that would correspond to roughly .70 reliability. This meant that any of the responses with an ADI of 1.167 or above were considered problematic – the interrater agreement about likelihood of this particular response was not satisfactory. Responses that had an ADI higher than this value were removed from the analysis. The results of these analyses can be found in Table 1. Removed response are highlighted.
Table 1
Average deviation indices for each possible response Situational judgement test ADI Situational judgement test ADI SJT 1 – 1 0.64 SJT 5 – 1 1.04 SJT 1 – 2 1.28 SJT 5 – 2 0.96 SJT 1 – 3 1.12 SJT 5 – 3 0.72 SJT 1 – 4 0.00 SJT 5 – 4 0.72 SJT 2 – 1 0.32 SJT 6 – 1 0.32 SJT 2 – 2 0.72 SJT 6 – 2 0.32 SJT 2 – 3 0.72 SJT 6 – 3 0.72 SJT 2 – 4 0.72 SJT 6 – 4 0.32 SJT 3 – 1 0.64 SJT 7 – 1 0.72 SJT 3 – 2 0.72 SJT 7 – 2 1.04 SJT 3 – 3 0.64 SJT 7 – 3 0.80 SJT 3 – 4 0.80 SJT 7 – 4 0.72 SJT 4 – 1 1.04 SJT 8 – 1 1.20 SJT 4 – 2 0.64 SJT 8 – 2 1.60 SJT 4 – 3 1.36 SJT 8 – 3 0.72 SJT 4 – 4 0.96 SJT 8 – 4 0.72
Procedure. Study 1 followed a 2 x 2 mixed design, with ‘perspective-taking’ as a between-subjects variable, and ‘measurement point’ as a within-subjects variable.
condition. The dependent variable, influence skill, was measured twice using the SJTs: before and after the manipulation stage.
Study 1 was implemented using the online survey software Qualtrics. Participants were welcomed, asked for their informed consent and then to fill out the cultural familiarity, baseline perspective-taking and demographic measures. They then began the
pre-manipulation SJTs, which provided the first measurement of the dependent variable, influence skill. They rated the likelihood of using 16 possible responses in provided scenarios. There were four possible responses for each of four situations displayed at this stage. The SJTs were counterbalanced within measurement points.
Participants then progressed to the manipulation stage. Here, half the participants watched the manipulation video about perspective-taking. The other half watched the control video. Both videos were similar in length and style.
Participants then progressed to the final stage, the four post-training SJTs. These assessed participants’ influence skill after the manipulation stage. This provided the second measurement of the dependent variable. This stage followed an identical format to the pre-manipulation SJTs, only with different content. Upon completion, participants were thanked and debriefed. On average, Study 1 took 32.7 minutes to complete.
Results
Overview.
General performance. As stated, to ascertain the dependent variable, influence skill,
each participant’s rating pattern was individually correlated5 with the SME average rating
5
There is debate about which method of correlation one should use to correlate likert-scale items. Pearson’s was selected here based on the methodologies of previous research that has sought to correlate participants and SME’s on likert-scale data (Harries & Gillhooly, 2011; Psotka, Legree, Belanich, Bludau & Gray, 2000).
pattern, both before and after the manipulations. Therefore, each participant had one
correlation coefficient for before the manipulations, and one correlation coefficient for after. Across the dataset, these correlation coefficients indicated a generally weak positive linear relationship between the participants’ ratings and the SME average ratings. Observed correlations ranged from r = -.36 to r = .69, with standard deviation of .24, and an
interquartile range of .28. This suggests that across all the participants, the displayed
influence skill was relatively low. The average difference between pre-manipulation to post-manipulation influence skill scores was -.07, suggesting generally no change in influence skill between measurements taken before and after the manipulations, when looking at the total sample.
Pre-measures. Results from the cultural familiarity pre-measure indicated that the
average rating of cultural familiarity across the sample was 3.13 (SD = 1.19), suggesting a generally low familiarity with the Middle Eastern culture in the participants. Results from the baseline perspective-taking pre-measure indicated that the average score across all
participants on the IRI was 4.37 (SD = 1.17). This suggests that, on average, participants had a medium inherent tendency to spontaneously adopt the psychological point of view of others.
Confirmatory analysis. To examine the specific effect of the perspective-taking manipulation, a comparison was made between the pre-manipulation and post-manipulation influence skill scores of both manipulation conditions. The descriptive statistics for this data can be found in Table 2. The average difference between pre-manipulation to
post-manipulation influence skill scores was -.09 in the perspective-taking condition and -.05 in the control condition, suggesting that the condition the participants were in made little
difference to the possible improvement in influence skill between the two measurement points.
Table 2
Descriptive statistics of the Pearson’s correlation (r) between participants and the SME average.
Perspective-taking Control
Mean SD Mean SD
Pre-manipulation .38 .24 .33 .24
Post-manipulation .28 .19 .28 .27
To statistically analyse the differences between the conditions, a mixed-design ANOVA, with manipulation (perspective-taking/control) as a between-subjects measure and measurement point (pre-manipulation/post-manipulation) as a within-subjects measure, was conducted. The interaction between the manipulation and measurement point was not significant, F(1, 58) = .35, p = .55. This indicated that the average change in influence skill between measurements taken before and after a perspective-taking manipulation was not significantly different to the average change in influence skill between measurements taken before and after a control manipulation. Therefore, hypothesis 1 was not supported.
Exploratory analysis.
Controlling for external variables. To control for cultural familiarity and baseline
perspective-taking, these variables were entered into the analysis as covariates. The additions of the covariates were non-significant (cultural familiarity: F(1,57) = 1.21, p = .28; baseline perspective-taking: F(1,57) = .18, p = .67). These analyses also did not result in any
significant modifications to the interaction between the manipulation and the measurement point, or its effect on influence skill (cultural familiarity: F(1,57) = .34, p = .56; baseline perspective-taking: F(1,57) = .40, p = .53). Results from independent sample t-tests indicated
that participants in the perspective-taking condition did not significantly differ from those in the control condition on cultural familiarity: t(58) = 1.33, p = .18, or baseline perspective-taking: t(58) = .08, p = .93.
Deviation analysis. A deviation analysis was performed to give more insight into the
results. This meant that the absolute deviation between participants’ likelihood ratings and the SME average were computed and extracted. These deviations were then averaged across each measurement point. In this, just as in the correlational analysis, each participant had a single ‘average deviation’ coefficient for both before and after the manipulations. This exploratory analysis continued with these derivatives acting as alternative markers of influence skill. A high deviation would indicate low agreement between a participant’s ratings and the SME average ratings, which would therefore indicate low influence skill.
These coefficients were entered into a second mixed-design ANOVA, which similarly to the correlational analysis, resulted in a non-significant interaction between the
manipulation and measurement point variable, F(1,58) = .02, p = .89. Again, this indicated that the average change in influence skill between measurements taken before and after a perspective-taking manipulation were not significantly different to the average change in influence skill between measurements taken before and after a control manipulation.
Deviation analysis was also performed on individual scenarios. Here, t-tests were used to assess the effect of perspective-taking on each individual SJT, in turn. There was one significant result: a significant difference in influence skill between participants in the
perspective-taking condition and participants in the control condition when isolating the sixth SJT: t(58) = 4.06, p = .04. However, inspection of the means showed that it was the control condition (M = 1.08) that had lower average deviation, and therefore showed higher influence
skill, than the perspective-taking condition (M = 1.39), a result which does not support the reasoning employed in hypothesis 1.
Influence techniques. The specific types of influence technique rated by participants,
and how the perspective-taking manipulation affected their ratings of the likelihood of use, was investigated. The responses to the 14 post-manipulation SJTs6 were divided using the five influence technique factors defined by Wolfe and Arrow (2005): positive trait
techniques, negative techniques, power-differential techniques, resource techniques and positive feeling techniques. The ratings that the participants indicated for these responses were then summed according to these factors. In this, each participant had a total rating score for each of the five influence technique factors. These scores were then compared between perspective-taking condition, using t-tests. The results of these t-tests can be found in table 3. A significant difference was found on the power-differential factor, with the average rating of the likelihood of using power-differential techniques being lower when rated by participants in the perspective-taking condition, than the control condition. This result suggests that watching the perspective-taking video led to participants rating techniques that highlighted a difference in power as less likely to be used, than participants who saw a control video. Table 3
Independent samples t-test results comparing factor scores between perspective taking conditions (perspective-taking vs. control).
6
Only the post-manipulation SJTs were selected as this is only where the effect of the manipulation on influence technique selection would be seen.
Influence technique factor Responses included Mean t p Perspective-taking Control Positive trait techniques SJT7-1, SJT8-3 11.55 10.96 1.21 .23
Discussion
Receiving a perspective-taking manipulation did not lead to participants showing a significant improvement in influence skill, when compared to participants who received a control manipulation. One explanation of this was thought to be that the perspective-taking manipulation was not long enough to teach the ERCA method, and therefore, perspective-taking levels were not properly manipulated. Indeed, military training courses can span days in order to adequately teach soldiers how to efficiently use perspective-taking in the field (Reivich, Seligman, & McBride, 2011). However, this is a relatively unconvincing
explanation given the purposes of this study. Other studies that have focussed on increasing perspective-taking levels have utilised less instructional, and arguably less potent or
memorable manipulations (e.g., a writing task was used in Galinsky & Moskowitz, 2000), the length of the manipulation was not thought to be the problem.
Another explanation may be the method of assessment that was employed. There is much debate about the correct and most valid way to assess situational judgement tests (De Leng et al., 2017; Bergman, Drasgow, Donovan, Henning, & Juraska, 2006), and with many different methods to assess performance, consensus has not yet been reached as to which is superior. Assessing performance on SJTs by looking at the general agreement - in this case using correlations - with SMEs is a noted and much-used method of assessing SJTs,
particularly those measuring human performance in security operations whereby there are no Negative techniques SJT5-3, SJT6-1, SJT6-2, SJT7-4 17.32 18.70 -1.57 .12 Power-differential techniques SJT5-1, SJT6-3, SJT6-4 9.42 11.62 -2.50 .01 Positive feeling techniques SJT5-4, SJT7-2 7.87 8.93 -1.66 .10 Resource techniques SJT5-2, SJT7-3, SJT8-4 13.77 13.41 .45 .65
clear, right or wrong answers (Psotka, Legree, Belanich, Bludau, & Gray, 2000; Pstoka, Streeter, Landauer, Lochbaum, & Robinson, 2004). This acted as the motivation for this assessment methods selection. Yet, to alleviate any doubts over the correlational analysis, exploratory analysis with deviations was also conducted. This analysis supported the results found using correlations: an insignificant interaction, yet a slight reduction in agreement with the SMEs. It therefore validated the results of the correlational analysis, and rendered an explanation of the lack of improvement due to the grading method as less convincing.
More compelling reasons for the lack of improvement were therefore believed to stem from issues brought about by the study’s repeated-measures design. Firstly, due to
complications with the assessment software, Qualtrics, counterbalancing could only be implemented within the measurement points. While great care was taken in ensuring clarity and systemisation of the information presented in the SJTs, the possibility that they varied in difficulty between the pre-manipulation and post-manipulation stage, and that this affected the chance to show improvement, although unlikely, is something that cannot be ruled out.
Secondly, observations made while Study 1 was running online, indicated that many participants who had only partially completed their responses to the study, had stopped at the manipulation stage. This means the online software Qualtrics had registered a long period of inactivity in participants, and the period began during the manipulation stage. This may suggest that participants actually became disengaged in the five minute period where the videos played. Indeed, watching the videos was the most passive part of the experiment, positioned between parts where participants were actively thinking about the SJTs and giving their responses. It is plausible that the ‘active-passive-active’ chronological order of the study, brought about by the fact that it had a repeated-measures design, led participants to lose focus in the second active part. Disengagement is a noted problem with online
methodologies (O’Neil, Penrod, & Bornstein, 2003; Dandurand, Schultz, & Onishi, 2008), and is especially likely given that often participants complete online studies anonymously, and without the physical pressures that come with a laboratory environment.
Study 2 sought to address these issues by changing to a between-subjects design. In doing so, the counterbalancing issue was removed, and in inverting the chronological order of the study so that the manipulation was first, the study provided a more natural flow of activity for the participants. The hope was that this created a better environment to measure influence skill in participants.
Study 2
Study 2 utilised a branching virtual interaction assessment to assess the effect of having an interpreter mediate interactions on perspective-taking ability, and the subsequent effect this has influence skill in military influence interactions. Study 2 assessed hypothesis 2, which predicted that interpreter presence would reduce the ability to perspective-take, and subsequently, reduce influence skill.
Method
Sample characteristics. A G-power analysis with an effect size of f = 0.2, an alpha of .05 and a power of .8, indicated that 180 participants were needed for a between-subjects design with four groups.
203 participants were sourced (Mage = 24.1, SDage = 5.67, 55% male). Two main sources were used for participant recruitment. The first source (n = 131) used was students from the University of Amsterdam. They were reimbursed with a 0.5 research credit for their participation. The second source (n = 72) was the online research recruitment platform ‘Prolific Academic’. These participants were screened through the platform so that only
students participated, to ensure that no unnecessary noise was added to the data. Participants sourced from Prolific Academic received a monetary reimbursement for their participation. 23 participants of the total sample were removed from the analysis as they failed a
manipulation check. Materials.
Knowledge of military deployment. To assess knowledge of military deployment,
participants were asked to rate how experienced they were with military deployment on a 7-point Likert scale.
Baseline perspective-taking. To measure baseline perspective-taking levels, the same
subsection of the IRI was used as in Study 1.
Manipulations. In Study 2, participants received the same manipulations as in Study
1, for both the perspective-taking and control conditions. As a flaw of Study 1 was thought to be disengagement - although mainly due to study order - the videos were condensed slightly to make it shorter and more concise. After each video, a manipulation check asked
participants to recount a particular element of each video (e.g., ‘Please name the second step of the perspective-taking method outlined in the previous video’).
Branching virtual interaction assessment. To assess the dependent variable,
influence skill, Study 2 used a branching virtual interaction assessment titled Influence in Ochen. The interaction depicted a military influence engagement, where participants assumed the role of a soldier who had been tasked with influencing a local civilian. Branching virtual interaction assessments are effective methods of assessing interpersonal skills (Nelson & Blenkin, 2007; Olson-Buchanan et al., 1998), as participants are better able to demonstrate ongoing skills in complex social interactions, while the standardisation of responses removes the inter-rater reliability problems of other interaction assessments such as traditional
role-plays. They also provide a content-rich, fluid and engaging environment for assessment (Olson-Buchanan, 2002). Branching virtual interactions are regularly used by military psychologists studying similar processes to influence (Zielke & Linehan, 2009; Mueller-Hanson et al., 2009; Prasolova-Førland, Fominykhand, Darisiro, & Mørsch, 2013).
The virtual interaction began with a briefing. Participants read a transcript of a
military briefing, where they learnt about the cultural profile of a fictional town, ‘Ochen’, and the country, ‘Olmova’, in which the interaction was set. They also received a brief history of fictional insurgency problems in Olmova, and a profile of their influence target, ‘Ovi Kabol’, a local resident in Ochen. Character and place names were generated randomly, and then selected for neutrality. Participants were told that Ovi was selected by the unit because of his influential position in the town, and that prior to this interaction, a good rapport with Ovi had been established over several meetings. Participants were then briefed that the objective of the interaction was to utilise this good rapport in requesting Ovi’s help in their counter-insurgency endeavours, for example, his help in identifying unknown individuals. This briefing intentionally provided certain information which could be utilised later in the
interaction. The idea was, as with Study 1, that the information would be used by those with a higher level of influence skill, i.e. those operating at higher levels of perspective-taking. Participants were then led to the beginning of the interaction.
The script for Influence in Ochen was developed using audio of real-life interactions between members of the Provincial Reconstruction Team (PRT) and Operational Mentoring and Liaison Team (OMLT), and locals in the Uruzgan province of Afghanistan7. Insights
7 Two-hour long audio recordings of interactions between the PRT and OMLT teams were studied. One recording was a shura – Arabic for consultation – concerning the building of a bridge in the local area by the PRT. Another was general progress meeting by the OMLT. Denser parts of the interactions were transcribed, and the distinctive conversation patterns, question phrasings and interpretation insights were extracted and implemented into Influence in Ochen’s script (see Appendix C4 for an example).
drawn from these were implemented in both the interaction construction and dialogue creation. Previous branching role-play tools, such as Connect with Haji Kamal – a military interaction trainer created for the US military, were also used as inspiration (Moore, 2010).
Thirteen scenes were devised, covering the entirety of an interaction and the different ways it could progress. As Study 2 sought to study the effect of an interpreter, there were two versions of each scene – therefore technically 26 film clips. The scenes were filmed using two actors. In the ‘translation’ condition, only the actor playing Ovi was present. He was directed with non-verbal cues, and spoke in an indecipherable language. The videos were edited so that a real-time translation was offered on screen, as the actor spoke. In the
‘interpreter’ condition, both the Ovi and interpreter actors were present. In these scenes, the actor playing Ovi performed in the same way to how he performed in the ‘translation’ condition, keeping vocal inflections as similar as possible to the ‘translation’ scenes. The script incorporated natural interjections by the interpreter, who would translate Ovi’s speech; a fluid flow between the interpreter and interaction subject was observed as common in real-life interactions. When these interjections occurred, the actor playing Ovi remained consistent in the displayed emotion, but performed no extra behaviour. This was in order to attain as much standardisation between conditions as possible. In both conditions, the corresponding choices were identical. A comparison of the same scene in both conditions can be found in Figure 2.
Figure 2. A comparison of the translation (A) and interpreter present (B) conditions in a scene in the virtual influence interaction, Influence in Ochen.
The assessment was developed so that participants were offered 2-3 response options at the end of each scene. These response options typically involved the application of a specific influence technique, as seen in previous literature (Ramsden-Zbylut, Wisecarver, Foldes, & Schneider, 2010; Wolfe & Arrow, 2013). These techniques were, as in Study 1, selected on the basis of the plausibility for the pathways. But further, they were selected based on what would be effective given the influence engagement. In this, each choice was carefully selected, and accompanying dialogue was carefully created (see Appendix C1 for the full script). It was important that each choice made sense both in how it follows and precedes other choices and scenes, both contextually and also in terms of the influence skill each individual pathway strove for.
As a response option was selected, the pathway branched, and participants were presented with a different scene and subsequent response options, depending their choices. The scenes were arranged in a complex branching structure, so that there were 40 possible routes (see Appendix C2 to see full structure of the interaction). The structure allowed for the possibility that one could reach the same outcome via different pathways; in real-life it is possible to make a social faux-pas and then recover, or vice versa. The structure was
carefully created to ensure that each of the forty routes flowed in a feasible and realistic way, with varying levels of influence skill. The script was reviewed by two experts8, whose
8
The first expert was an author of ‘Interaction rituals and language mediation during peace missions: experiences from Afghanistan’, the paper used to develop the second hypothesis. The second expert was a member of the Civil-Military interaction commando (CIMIC).
comments were incorporated and the script adjusted to enhance realism. The interaction structure was revised several times, to improve pathway plausibility and sensitivity.
In order to quantify the dependent variable for the experiment, a three-hour workshop was held to garner SME grades for each pathway. A group of five SMEs, comprising of professionals working in cultural psychology, influence and civil-military interaction, worked closely to grade each pathway individually9. Firstly, each pathway was described in detail, with the dialogue elaborated on, and SMEs gave their initial grading of influence skill. Following this, discussion of the pathway was invited10, before SMEs gave their final influence skill grade - if different from their initial grade. All forty possible pathways in the interaction were given an ‘influence skill grade’ out of 10. These individual grades were then averaged, and the average was used as the final ‘influence skill grade’. Therefore, depending on the pathway of choices, participants could arrive at a different, pre-established influence skill grade (for the full list of grades, see Appendix C3).
To view a sample version of Influence in Ochen, please click here.
Procedure. Study 2 followed a 2x2 between-subjects design, with interpreter
presence (interpreter present/translation) and manipulation (perspective-taking/control) as the between-subjects variables. Participants were randomly assigned to one of four groups. Therefore, there were two perspective-taking groups and two control groups. Within these two groups, one group had their interactions mediated by an interpreter, while one received a
9
For Study 2, we determined the grades we would measure our dependent variable, influence skill, by grading each pathway individually – as its own entity. This was favoured over attributing a score to each single branch/choice in a pathway, and accumulating these scores into a final grade, a method that has been employed by other studies (Mueller-Hanson, et al., 2009). This method can be faulted for introducing dependencies, and it can be difficult to properly score a branch that can be reached from multiple pathways which differ in the preceding influence skill. This approach can also be criticised for reducing assessments to individual assessments, and further, ignoring the continuous element of social interactions. Therefore, a more holistic grading approach, where each pathway is graded as its own entity, was chosen. Furthermore, it was less complicated and more practical given the size of the study.
10
real-time, on-screen translation of the influence target. The dependent variable was ‘influence skill’, measured using the pre-established influence skill grade.
The study was implemented using the online survey software Qualtrics. Participants were welcomed, asked for their informed consent and to fill out the pre-measures. They then progressed to the manipulation stage, where half watched a video of the perspective-taking manipulation and half watched the control manipulation video.
Participants then progressed to the virtual interaction. This involved a briefing about the task, including a page on the history of the setting, the culture of the setting, a background to the interaction and a profile of the influence target, Ovi. Having read this, they could then start the virtual interaction with Ovi.
Upon completion of the virtual interaction, participants were thanked and debriefed. Study 2 took 14 minutes to complete, on average.
Results
Overview.
General performance. To assess the influence skill, a coding system was developed.
Each response choice in the interaction structure was given a letter. Then, each pathway was plotted to represent a unique code of letters, and each unique code was matched with its grade: the average of the grades given by the five SMEs in the grading workshop. For example, a pathway where a participant selected to engage in more small talk, highlight the negative emotions of the village, praise Ovi as a moral person and then request collaboration (the highest graded in the interaction structure) would receive the pathway code AEHP, which would translate to a grade of 7.7.
The average score across the sample was 6.05. This suggests that the participants, in general, did well in the virtual interaction assessment, displaying good influence skill.
Pathways were analysed and simple descriptive statistics were extracted. In total, participants chose a total of 29 of the possible 40 pathways - 11 pathways were not picked by any of the participants. Within these 29 pathways, the most-selected pathway was ‘AEHP’ (18.9%), the highest graded pathway in the interaction by the SMEs. Collapsing across all conditions, four pathways dominated the choices (ADHP, ADHO, AEHP and AEHO), with 61% of
participants sticking to these four pathways. These pathways shared the common factors of engaging in more small talk and using a moral appeal to persuade Ovi to comply, and they were among the highest-graded pathways in the interaction structure, with grades ranging from 6.0 upwards. The remaining most highly-graded pathway was ‘CFKJTQ’ (grade of 6.6), in which participants would have chosen to directly address the issue of the meeting,
highlight how they and the target are on the ‘same team’, compliment the target, and then use the negative self-esteem and collaboration tactics influence techniques. This pathway was not selected by any participants. Conversely, 5 participants (2%) selected pathways involving choice ‘N’ (threat), which were considered the poorest pathways.
Pre-measures. Results from the military experience pre-measure indicated that the
average rating of military experience across the sample was 2.52 (SD = 1.32), suggesting a generally low experience with military deployment in the participants. Results from the baseline perspective-taking pre-measure indicated that the average IRI score across all participants was 5.02 (SD = 0.83). This suggests that, across the whole sample, participants had a medium inherent tendency to spontaneously adopt the psychological viewpoint of others.
Confirmatory analysis. A two-way ANOVA was run on these grades, with
interpreter presence as the first independent variable and perspective-taking as the second. No significant main effects were found. The influence skill grades attained after a perspective-taking manipulation were not significantly different to the influence skill grades attained after a control manipulation, F(1,177) = 0.34, p = 0.55. Further, whether the interpreter mediated the interaction did not significantly affect influence skill grades, when compared to viewing solely an on-screen translation, F(1, 177) = 3.06, p = 0.08. Therefore, hypothesis 2 was unsupported. The means and standard deviations of these groups’ grades are plotted in Figure 3.
Figure 3. Mean influence skill grade displayed by the four conditions.
Exploratory analyses. 3 3,5 4 4,5 5 5,5 6 6,5 7 7,5 Perspective-taking Control M ean i n fl u en ce sco re Manipulation Interpreter present Translation
Controlling for external variables. To control for military experience, this was
entered into the analysis as a covariate. The addition of the covariate was not significant, F(1,176) = .24, p = .62. This analysis also did not result in significant modifications to the effect of the manipulation, F(1,176) = .35, p = .55, or interpreter presence, F(1,176) = 3.06, p = .08, on influence grade. Similarly, to control for baseline perspective-taking, this was also entered into the analysis as a covariate. The addition of this covariate was not significant, F(1,176) = .26, p = .60. This analysis also did not result in significant modifications to the effect of the manipulation, F(1,176) = .35, p = .56, or interpreter presence, F(1,176) = 3.06, p = .08, on influence skill grade.
Grading method. Further exploratory analysis of the grading method was conducted.
Firstly, the same analyses were ran, using the grading predicted during the creation of the interaction structure instead of the SME average grades. This did not result in a different pattern of means, or any significant modifications to the effect of the manipulation, F(1,177) = .23, p = .64, or interpreter presence, F(1,177) = .44, p = .51, on influence skill grade. Secondly, the SMEs’ initial grades, given before the discussion, were used to form the average if they were different to the final grade. This analysis did not result in a different pattern of results, or any significant modifications to the effect of the manipulation, F(1,177) = .30, p = .58, or interpreter presence, F(1,177) = 2.99, p = .08, on influence skill grade.
Influence techniques. Similarly to the exploratory analysis of Study 1, the specific
types of influence technique used by participants, and how the perspective-taking manipulation affected their selection, was investigated. The 20 possible responses in
Influence in Ochen were divided according to the five influence technique factors defined by Wolfe and Arrow (2005): positive trait techniques, negative techniques, power-differential techniques, resource techniques and positive feeling techniques. The pathways selected by
the participants were scored on these factors so that each participant had a score for each of the five factors. For example, a participant who selected the pathway ‘CFKJTQ’ would have a score of ‘3’ on the power-differential techniques factor, but a ‘0’ on the positive trait techniques factor. The scores on these factors were then compared between perspective-taking condition, using independent samples t-tests. No significant differences between the conditions were found. A trend approaching significance was observed for the power-differential influence techniques factor. The results of these t-tests can be found in Table 4. Table 4
Independent samples t-test results comparing factor scores between perspective taking conditions (perspective-taking vs. control).
Discussion
Receiving a perspective-taking manipulation conferred no significant advantage to participants when engaging in a virtual, continuous, branching role-play, when compared to those who received a control manipulation. Despite adaptations to the manipulation after Study 1, including a change to a between-subjects design and a refinement of the content, it still failed to elicit a significant improvement to the participants’ influence skill. The results appear to show that the perspective-taking manipulation, as it stands, does not raise the level
Influence technique factor Responses included Mean t p Perspective-taking Control Positive trait techniques O, D, H, S 1.41 1.51 -.81 .42 Negative techniques E, G, I, M .52 .54 -.26 .79 Power-differential techniques J, N, L, C, Q, P .90 .72 1.86 .06 Positive feeling techniques K, T, R .30 .37 -.86 .39 Resource techniques A, B, F .98 .99 -.61 .54
of perspective-taking high enough to elicit a difference in influence skill, at least not in a way that can be captured by the present assessments. This is discussed further in the general discussion.
Because of the lack of effect in the perspective-taking variable, the effect of an interpreter mediation on perspective-taking and subsequent influence skill cannot be
understood. However, if anything, the results from Study 2 appear to point to a trend whereby interpreter presence is potentially increasing influence skill. This trend would partly
challenge the second hypothesis, which predicted that mediation by an interpreter would reduce influence skill across both the perspective-taking and control conditions. This is also discussed further in the general discussion.
Study 2 was the first use of Influence in Ochen outside of pilot testing. Therefore, exploratory analysis was conducted to address and disregard any initial usage concerns that may have affected how one can understand Study 2’s results. A first concern pertained to the validity of the grading method (see Materials). An invalid grading method would have rendered the results of Study 2 also invalid, as one could not trust the method of scoring as a valid assessment of influence skill. This concern was raised as one could suggest that
allowing time for discussion might have diluted the opinions of the experts, or meant that sway was introduced. Another problem may have been the sequential process of grading the pathways, which might have introduced a comparison element to the grading. Finally, it might have been possible that the communication of each pathway lacked clarity. Although numerous visual aids accompanied the workshop, it was possible that the scope of the assessment, with the nuances and interactions in each pathway, was too great to relay over the course of one, three-hour session. Perhaps giving the SMEs a longer period to become familiar with the assessment would have been beneficial.
Exploratory analysis, however, revealed the same distribution of grades and
subsequent pattern of results, after running the same analyses with the author’s grading. The author’s grading was done while the construction of Influence in Ochen was ongoing, and benefitted from the full insight into the reasoning behind each choice. This suggests that the experts’ opinions were more or less in line with the author, and their grades were aligned with the intended distribution of influence skill across the structure of the interaction. Further exploratory analysis with the SME grades prior to discussion also resulted in no differences to results, supporting the notion of a valid grading method.
Another concern explored was that within Influence in Ochen, there were no opportunities to properly utilise the skill enhancements that the perspective-taking
manipulation invoked. Previous literature has shown that a heightened level of perspective-taking should increase specific skills that enhance influence interactions (Galinsky & Moscowitz, 2000; Galinsky & Mussweiler, 2001; Van den Berg, van Hemert, van Beurden, & Jetten, 2015). Influence in Ochen was designed as an online, video-driven assessment, in order to better capture these heightened skills. However, it could be some skills did not adequately transfer to an online medium (Le & Beidel, 2017). For example, it might be that a certain technique can be delivered in a particular non-verbal manner, and this would be effective at building rapport. Indeed, vocal intonation and facial expressions have been found to alter interpretations, and be crucial in compliance-gaining (Peters, 2007). This may warrant the selection of a pathway that typically would not be advised, and may explain why the perspective-taking and control condition did not significantly differ.
This criticism of transferability is one that holds for almost every online assessment of social skills. Yet, in reviewing the wealth of significant findings in related research that used a similar branching virtual interaction assessment to measure influence skills (Ahamer, 2004;
Mueller-Hanson et al., 2009), the criticism becomes relatively unconvincing. Further, recent research has unequivocally supported the validity and use of branching virtual interaction assessments in measuring social skills (Le & Beidel, 2017). In light of this research, one can be reasonably confident that the online methodology of Influence in Ochen is adept enough to capture any differences in influence skill that may arise, and this is an unlikely explanation for why the perspective-taking and control condition did not significantly differ.
A final point pertained to the sensitivity of the assessment. If a certain pathway was clearly more favourable when trying to reach the influence objective, then the assessment would not be able to sensitively capture differences in the levels of influence skill.
Participants were given information before beginning the interaction that gave them an insight into the background of the interaction (see Materials). This information was provided intentionally; specific choices centred on this information (e.g., choices ‘A’, ‘D’, ‘K’) and were deliberately formulated to be effective if used during the assessment. This was done in order to give those participants functioning at a higher level of influence skill a chance to excel, ostensibly those in perspective-taking conditions. Yet, exploratory analysis revealed that the majority of participants favoured these choices. 72% of the sample began with choice ‘A’, a choice which would indicate higher influence skill than the other choices available when displayed. Further, 61% of the sample selected one of the top four highest rated pathways in the interaction. With large numbers of the participants selecting these high influence-skill choices and pathways, this could even point to ceiling effect, whereby the assessment can no longer discriminate between these operating at a higher level.
Ultimately, as online branching scenario assessments are a relatively new method of assessment, there is little documentation regarding how to adequately psychometrically assess them (Olson-Buchanan, 2002). Commonly used adaptive testing theories such as item
response theory, which could provide more psychometric certainty, cannot be implemented due to the provision that it adheres to a pre-established structure which adds new information after each response, unlike other adaptive tests which draw from a larger bank of items (Danver, 2016). Still, there appears to be room for improvement to the sensitivity. Future research may benefit from more discretely weaving the information into the briefing, or adding more response options that vary in the level of information that they use. This would increase sensitivity, and strengthen the assumption that only those who are operating at higher levels of perspective-taking are properly utilizing the information. A second option may be to completely remove this information, and rely purely on the enhancement of other skills that perspective-taking manipulations increase, such as response anticipation or body language interpretation (Galinsky, Ku, & Wang, 2005).
General Discussion
Across two experiments, using two distinct methodologies and two different operationalisations of influence skill, participants who received a perspective-taking
manipulation did not show a significant difference in influence skill, compared to those who received an unrelated, control manipulation. Furthermore, the presence of an interpreter who mediated the interactions was not found to reduce influence skill, when compared to
interactions that were facilitated solely by on-screen translations. Below, factors related to the findings are discussed in more detail. Firstly, the present perspective-taking manipulation and reasons why it failed to effectively heighten perspective-taking levels are elaborated.
Secondly, findings pertaining to the effect of an interpreter presence are briefly expanded. Thirdly, the disparity between general and military influence contexts, and how the two presented studies provide support for this notion is discussed. Recommendations for future research are provided where appropriate.
In both Studies 1 and 2, in comparison to a control manipulation, a perspective-taking manipulation, where participants watched an educational video about a 4-step method of perspective-taking, did not significantly increase levels of influence skill. Therefore,
hypothesis 1 was unsupported. One conclusion that could be drawn here would be theoretical: that perspective-taking is not useful in increasing influence skill. This conclusion would challenge the first hypothesis, and further, would fly in the face of a large body of research which has propagated higher levels of perspective-taking as instrumental in increasing influence skill (Galinsky & Mussweiler, 2001; Van den Berg, van Hemert, van Beurden, & Jetten, 2015). It therefore seems more sensible to suggest that if there is a fault with
perspective-taking, it could lie with the manipulation intended to invoke it, rather than the construct itself.
The ERCA method was developed specifically for the present studies, and was inspired by previous literature that had used a similar four-step, ‘in-the-moment’ method of increasing perspective-taking levels in deployed soldiers (Mueller-Hanson et al., 2007). The idea was that teaching participants the ERCA method would not only make perspective-taking more salient, but give participants a means to actively take perspective during the studies, replicating the context of military deployment. It was therefore created to be quick, direct and mobile: ideal for soldiers in the field. For the purposes of the two studies, a short, stylised video was used to teach participants about the method. A video methodology was employed to increase engagement with and subsequent recall by the participants (Choi & Johnson, 2005). In this, the entire manipulation consisted of participants solely watching an educational video.
Yet, if one reviews other perspective-taking manipulations, it becomes apparent that most share a common factor: they are active. Indeed, the most common manipulations are
