A Proposal of the Emotion Hierarchy Diagram for Designing the Service Processes

Abstract

This paper proposes a method to design emotion-based service processes by decomposing highly abstract emotions into detailed emotions. First, we propose an Emotion Hierarchy Diagram (EHD) to decompose abstract emotions into detailed emotions and organize them for identifying and designing specific service processes. The identification steps for creating EHDs are listed. Further, we propose how to use the EHD for deriving service elements. Further, this study verifies whether the proposed EHD can derive service elements for high-quality emotions. We examine where individual differences appear in the EHD, and we find that these differences appear in the axes. Additionally, using statistical tests, we verify whether the service elements for high quality were listed when the EHDs are drawn in both cases, i.e., when the axes were fixed and when they were not.

1 Introduction

The tertiary industry currently accounts for 70 % of Japan’s gross domestic product (GDP). According to Reichheld [1], customers who receive highly valuable services will generate more profits through subsequent visits and positive word of mouth, while customers who receive poor service will do harm with negative word of mouth and by discouraging other potential customers. Therefore, it is necessary to study the process designs for highly valuable services.

Service delivers two types of quality: functional quality and emotional quality. A well-designed service can provide a cup of coffee in just a short time, which represents functional quality, while it can provide the emotions of “relaxation” or “excitement,” which represent emotional quality. The research focusing on emotional quality has not advanced much, while the research focusing on improving functional service is quite extensive [2]. For example, although there are some service frameworks such as the service profit chain [3] or the service marketing triangle [4], there is little discussion about how these frameworks can be used for improving emotional quality.

A few prior studies examined emotional quality in the context of services. Parasuraman et al. (1998) proposed a SERVQUAL method to evaluate the quality of service [5] using five criteria: reliability, assurance, tangibles, empathy, and responsiveness. Emotional quality is implicitly included in the survey questions. The SERVQUAL method is not always reliable since the evaluation results of a specific service will be relatively lower because the criteria for the questions are derived from the evaluation of a generic service. Customer journey [6] is a method for visualizing a service process by focusing on the emotion experienced by customers during the service. However, it is difficult to use the customer journey method to improve an emotion that is not yet experienced in the context of a particular service.

Next, we discuss the prior research on emotions in the psychology stream. Plutchik [7] proposed a three-dimensional circumplex model in which all emotions can be represented using eight basic emotions and their combinations. Higuchi [8] proposed a psychological model in which an emotion can be constructed with combinations of other emotions using factor analysis.

These models cannot decompose abstract emotions into detailed emotions because the emotions represented in the models are highly abstract and ambiguous. A highly abstract and ambiguous emotion could be associated with several detailed and inconsistent emotions. If a service is designed based on highly abstract emotions, it will deliver several detailed and inconsistent emotional services simultaneously; i.e., it will deliver service with low emotional quality.

For example, consider a café owner who has a service policy that the café will be based on a particular emotion, “excitement.” This policy will lead to a confused service process in the café because the emotion “excitement” is too abstract to allow for the design of a specific service process.

Decomposing an abstract emotion into detailed emotions and designing service processes based on the detailed emotions with a focus on emotional quality are significant because there is very little prior research on emotional quality. Moreover, abstract emotions cannot be used to design service processes even if the service provider focuses on emotional quality.

This study proposes an Emotion Hierarchy Diagram (EHD) to decompose abstract emotions into detailed emotions and to choose one or more emotions from among them to determine specific service elements for designing a service process with high emotional quality.

The scope of this study is as follows.

The service classification scheme [9] classifies service products into 2 × 2 categories based on (1) the direct recipient of the service and (2) the nature of the act. The direct recipient of a service is classified into persons and things. The nature of the act is classified into tangible actions and intangible actions. The service classification scheme does not specify whether the service provider is a person or a machine. The scope of this study is limited to the context where the direct recipient of the service is a person, the nature of the act is a tangible action, and the service provider is a person.

2 Proposal of an Emotion Hierarchy Diagram

Ueda and Hoshino [10] show that an increase in the number of loyal customers makes the service more profitable when a store constructs a system to remind customers of “hope” hidden in the customer’s unconscious depth psychology. In this study, “hope” is paraphrased as “excitement,” and this study proposes a tool called the Emotion Hierarchy Diagram (EHD) for deriving a service element using a sample of applications based on “excitement.”

There are two assumptions in the EHD:

• Assumption 1: An emotion has a certain level of abstraction.

• Assumption 2: The type of abstraction of emotions used is not a “has-a” type (composition or constituent) but an “is-a” type (inheritance or sub-typing), similar to the approach in object-oriented programming.

The components of the EHD are “top emotion,” “detailed emotion,” “branching line (for combinations),” and “axis.” The top emotion is the most abstract emotion. Detailed emotions are decomposed along the axis. Detailed emotions have hierarchical relations. The emotion in the upper layer is a parent emotion, and that in the lower layer is a child emotion. This relation is relative. The branching line decomposes an abstract emotion into detailed emotions. The axis specifies the condition for decomposing the emotions. Figure 1 shows the components and notations of the EHD.

Fig. 1

Components and notations of an emotion hierarchy diagram

While drawing an EHD, options for the axis are required. This study refers to the “global structure of emotion types” proposed by Ortony et al. [11] for these options. Table 1 presents the options for the axis.

Table 1 Options for the axis in an EHD

Figure 2 shows an EHD for “excitement.” To draw the EHD, this study chose 330 articles at random from the 1,876 articles obtained as search results when the keyword “excitement” was used in the Nikkei BP database [12] on 1 November 2014.

Fig. 2

The EHD for “excitement”

Figure 2 classifies the excitement expressed in the 330 articles along three axes (time, accomplishment, and cognition). First, the time axis is used for decomposing “excitement” in Fig. 2. This study refers to the dictionary definition of “excitement” for deciding the axis in the first layer. Sanseido’s Daily Concise Dictionary [13] defines “excitement” as “to make someone restless with pleasure or anticipation.” We apply this meaning when we examine the axes in Table 1. The axes that can decompose joy and expectation are time, accomplishment, and cognition. We consider time to be the best axis for decomposing “pleasure” and “anticipation” because we consider pleasure to be a current emotion (in the present) and anticipation to be a future emotion. We choose “present” and “future” as elements of the axis because excitements in the past are thought to be recollections of past excitements (in one’s mind). Since the real experience of excitement is important, not the recollections of past experiences, this study does not select “past” as an element of the time axis. Next, we choose a suitable axis for the child emotions. In this study, both the axes in the second layer are accomplishment. The emotion in the second layer of the present side decomposed by the time axis cannot be decomposed to further detailed emotions; the emotion in the future side decomposed by the time axis is decomposed to the second layer by accomplishment and to the third layer by accomplishment and cognition.

The EHD can be used to derive service elements via the following steps:

• 1. Decide an emotion that a service wants to build or stimulate.

• 2. Draw the EHD by examining this emotion using axes.

• 3. Choose the detailed emotions that a service wants to build or stimulate.

• 4. Derive the service elements from the detailed emotions.

3 Validation

This study conducts two types of verification.

Validation 1 determines where individual differences appear in the EHD because there may be individual differences in emotions. The verification is based on the assumption that individual differences appear in the axes if the variability of the detailed emotions when the axes are fixed is small and the variability when the axes are not fixed is large.

Procedure

• 1. The variability of the detailed emotions when the axes are fixed is calculated using cosine similarity in Sect. 3.1.

• 2. The variability of the detailed emotions when the axes are not fixed is calculated based on the choice of axes: different axes produce different detailed emotions, as discussed in Sect. 3.2.

• 3. Each characteristic of the verification is compared at the end of Sect. 3.2.

Validation 2 verifies whether the method proposed in this study can derive the service elements required to obtain high-quality emotion.

Procedure

• 1. The service elements derived from the abstract emotion, and the detailed emotions in the EHD when the axes are fixed are compared using the statistical result in Sect. 4.1.

• 2. The service elements derived from the abstract emotion and detailed emotions in the EHD when the axes are not fixed are compared using the statistical result in Sect. 4.2.

3.1 Validation of Individual Differences When the Axes Are Fixed

There were nine subjects (seven men, two women) in this validation exercise; they study in the second to third grade of university.

First, the aim of an EHD and the method for drawing an EHD were explained to the subjects. Subsequently, we gave the subjects the articles in the sample and the axes; the subjects drew the EHD using this information. We searched for articles in the Nikkei BP database using the keyword “excitement” on 1 November 2014 and found 1,876 articles. We chose 100 articles randomly from these search results. In this section, the axes and the order of use are fixed; however, the subjects were told that they did not have to decompose an emotion if they could not decompose it into detailed emotions. Table 2 shows the given axes ordered according to use in this study.

Table 2 Axes used in the EHD for “excitement” in Sect. 3.1

Table 3 presents the detailed emotions in the EHD created by one of the subjects. In fact, there were nine sets of the detailed emotions. Table 4 shows the correspondence between the number of detailed emotions and the elements of the axis, and Table 5 shows the various emotions for each child emotion. The ID numbers in Table 5 represent the subject’s number.

Table 3 Detailed emotions in the EHD of one subject

Table 4 Number of child emotions corresponding to the elements of the axis

Table 5 Each layer of excitement

This study measured the degree of similarity between two excitements based on the semantic distance, which was calculated according to a word’s frequency of appearance. We calculated cosine similarity based on the results of morphological analysis using MeCab [14]. Shiozu and Iwashita [15] used 0.25 as the standard for the degree of similarity. We follow this and assign green color to values with similarity greater than 0.25 and less than 0.5, blue color to values with similarity greater than 0.5 and less than 0.75, and red color to values with similarity greater than 0.75. Table 6 presents the cosine similarity of the first layer <1> and second layer <2>, for example, and Table 7 shows the percentages of each area of cosine similarity per detailed emotion.

Table 6 Degree of similarity of different excitements

Table 7 Percentage of each area of cosine similarity per detailed emotion

From Table 7, we understand that the detailed emotions that are high in the hierarchy are inclined to a high level of similarity, by examining the results into two parts separated by the first layer (future parts <1>–<7>; present parts <8>–<10>). The level of similarity decreases if we increase the axes because the emotions decomposed by each axis have a corresponding variability. Moreover, by self has a higher level of similarity compared to by others in the accomplishment axis. When decomposing an emotion into detailed emotions by self or by others, “accomplishment by self” is easy to understand and is determinative. Thus, the level of similarity is high in the “by self” part. Further, the elements of an axis that are “far from the meaning of the emotion,” “difficult to imagine,” and “not determinative” tend to have high variability.

The percentage where the level of similarity is greater than 0.25 is over 85 % for all the detailed emotions. When fixing the axes, the EHD has low variability.

3.2 Validation of Individual Differences When the Axes Are Not Fixed

There were 21 subjects (20 men, 1 woman) in this validation; they study in the second to third grade of university.

First, the aim of the EHD and how to draw the EHD were explained to the subjects. Subsequently, we gave the subjects the articles and the example of the axes; the subjects drew the EHD using these. The articles were the same as those described in Sect. 3.1. Table 8 presents the example of the axes provided to the subjects. The subjects chose three axes each from the examples in Table 8 on their own.

Table 8 Examples of the axes

Table 9 shows the detailed emotions in the EHD created by one of the subjects. Table 10 presents the axes used for Table 9.

Table 9 Detailed emotions in the EHD of one subject

Table 10 Axes used for Table 9

Table 11 presents the axes that were used by the subjects in the EHD. All the axes that were included in Table 1 are colored red in Table 11; the axis that was not included in Table 1 is colored black in Table 11. Table 11 (3) presents some of the axes that were examined in detail for the elements of the axes in Table 8 (shown in blue). Table 12 shows the number of each kind of axes, the total percentage, and the accumulation. The ID numbers in Tables 11 and 12 represent the subject’s number.

Table 11 The axes used

Table 12 The number of each axis used for classification and the percentage

The time axis is used for about 81 % of the classifications in the first layer in Table 12 (1). That is, the variability in the first layer is small. However, in Table 12 (2) the accomplishment axis has the highest percentage in the second layer (34 %). In the third layer, the cognition axis accounts for 25 % of all the axes. The percentage accounting for all the axes decreases down the hierarchy. That is, diverse axes are chosen in the lower layers. Further, no one chooses the same combination of axes, and no one draws the same combination of detailed emotions. The choice of axis changes according to the layer. Thus, when the axes are not fixed, the EHD has high variability.

To summarize the verification results, the variability of the detailed emotions when the axes are fixed is low; when the axes are not fixed, the variability of the detailed emotions is high. Thus, individual differences appear in the choice of axes. Further, for each axis, the variability tends to be high for the elements of the axis that are “far from the meaning of the emotion,” “difficult to imagine,” and “not determinative.”

4 Validation of the Derivation of Service Elements

Next, we verify whether the EHD can derive service elements with high emotional quality in both cases: when the axes are fixed and when they are not.

4.1 Validation of Derivation of Service Elements When the Axes Are Fixed

There were 12 subjects (11 men, 1 woman) in this validation exercise; they study in the second to third grade of university. The subjects were divided into four groups. Each group listed the elements of the excitement café in a brainstorming session before drawing the EHD. Table 13 presents the elements discussed by Group B before the EHDs were drawn.

Table 13 Elements of the excitement café discussed in the brainstorming session before the EHD was drawn (Group B)

Subsequently, the aim of the EHD and how to draw the EHD were explained to the subjects. Each subject created an EHD by decomposing excitement into detailed emotions with 100 articles. Further, the subjects identified the elements of the excitement café from the detailed emotions. Table 14 shows the selected detailed emotions and elements of the excitement café per subject in Group B. The ID represents the group-individual number.

Table 14 Detailed emotions and elements of excitement café selected by each subject in Group B

The subjects evaluated the elements of the café proposed by the group members in both cases (using EHD and without EHD) with five points (low is one; high is five). Table 15 shows the average score of the elements.

Table 15 Average score of elements of excitement café

After confirming that the average score is normally distributed by describing the normal plot, we conduct an F-test for the equality of the two variances. Here, we assume that the average score of the elements without EHD is distributed with N(μ ₀, σ ²₀ ), while the average score when using EHD is distributed with N(μ ₁, σ ²₁ ). The result of the F-test is $F=1.096<{F}_{0.05}\left(11,10\right)=2.943$. There is no significant difference in the two variances. Subsequently, we conduct the student’s t-test to test the difference in the averages of ${H}_0:{\mu}_0={\mu}_{1\ }\mathrm{v}\mathrm{s}\ {H}_1:{\mu}_0<{\mu}_1$. The result of the t-test is $\left|t\right|=3.874>{t}_{21}(0.05)=1.721$. Thus, the null hypothesis is rejected.

The population mean of the elements of the excitement café when using EHD is greater than that when not using EHD, as shown by the result of the t-test. Thus, when the axes are fixed, high-quality service elements are listed in the EHD.

4.2 Validation of the Derivation of Service Elements When the Axes Are Not Fixed

There were 21 subjects (20 men, 1 woman) in this validation exercise; they study in the second to third grade of university. The subjects were divided into four groups. Each group listed the elements of the excitement café in a brainstorming session before drawing the EHD. Table 16 presents the elements discussed by Group E before the EHDs were drawn.

Table 16 Elements of an excitement café discussed during brainstorming session by Group E before drawing the EHD

Table 17 shows the selected detailed emotions and elements of the excitement café per subject in Group E. The ID represents the group-individual number.

Table 17 Detailed emotions and elements of excitement café selected by each subject in Group E

The subjects evaluated the elements of the café proposed by the group members in both cases (using EHD and without EHD) with five points. Table 18 shows average scores of these elements.

Table 18 Average score of elements of excitement cafe

After confirming that the average score is normally distributed by describing the normal plot, we conduct an F-test for the equality of the two variances. Here, we assume that the average score of the elements without EHD is distributed with N(μ ₀, σ ²₀ ), while the average score while using EHD is distributed with N(μ ₁, σ ²₁ ). The result of the F-test is $F=12.777>{F}_{0.05}\left(20,20\right)=2.464$. There is a significant difference between the two variances. Subsequently, we conduct Welch’s test to test the difference in the averages of ${H}_0:{\mu}_0={\mu}_{1\ }\ \mathrm{v}\mathrm{s}\ {H}_1:{\mu}_0<{\mu}_1$. The result of Welch’s test is $\left|t\right|=2.842>{t}_{23}(0.05)=1.713$. Therefore, the null hypothesis is rejected.

The population mean of the elements of the excitement café when using EHD is greater than that when not using EHD, as shown by the result of Welch’s test. Thus, when the axes are not fixed, high-quality service elements are listed in the EHD.

5 Conclusion

This study proposes an Emotion Hierarchy Diagram (EHD) to decompose abstract emotions into detailed emotions and to select them for deriving specific service elements to design a service process with high emotional quality.

This study conducts two types of verification. Validation 1 examines where individual differences appear in the EHD. We find that individual differences appear in the choice of axes.

Validation 2 verifies whether the method proposed in this study can derive the service elements required to obtain high emotional quality. We find that high-quality service elements were listed when the EHD was drawn in both cases: when the axes were fixed and when they were not.

However, the validity of the findings needs further verification.

• Internal validity: The service elements were derived from the abstract emotion before the service elements were derived from the detailed emotions. Since the experience could be affected by deriving the service elements with high emotional quality, there were 3-week intervals between the processes. There were one or two women in each experiment. Thus, it is possible that gender difference affected the choice of axes or the service elements derived with high emotional quality. Whether gender has an impact needs to be verified in future research by increasing the number of women subjects.

• External validity: Questions could be raised about the external validity since the experiments’ subjects were only in the second to third grade of university. However, this method is not designed specifically for the second to third grade of university. The method could be useful for all people who have emotions. We would want to verify the utility of the proposed method for all ages and all levels of education by broadening the research objective.

To conclude, we present further directions for future research. Although the direct recipient of a service was defined as a person, the nature of the act as a tangible action, and the service provider as a person with regard to the scope of this study, the proposed tool and method could be applied in a context where the service provider is a machine. A machine could provide customers a service with high emotional quality if the machine could provide the service elements from the detailed emotions similar to what a person does.