Keywords

1 Introduction

1.1 Background of Study

With the development of computer hardware and big data, the AI technology has matured much faster than expectation and already entered into numerous fields such as engineering, medical treatment, machine control and so forth to cooperate or fully replace human beings (Li and Jiang 2017). Some scientists even have predicted that AI would lead to the next technological revolution, and Workplace Automation caused by AI robots would take place 47% jobs in the US in next two decades (Smith 2016).

With the dramatic development of AI robot technology, increasing applications has appeared and developed in the service sector as well. Since the 1980s, the AI technology has been used to help hotels to choose the most preferable room for the guests (Agnew 1987). Delivery robot, designed in a hominoid appearance, also has been successfully adopted by a number of hotels in Japan to assist staff to deliver amenities and room services (Kratochwill 2015). In 2015, the first fully robot-based hotel “Henn-na” launched in Japan, and it provides robot servers replacing all human labors and performing the job of bell service, checking in/out service and even as an in-room butler and so forth (“Welcome to the Robot Hotel: robots do most of the work at a new hotel in Japan,” 2015).

The reasons for hotel sector to bring in smart robots are mainly due to its high efficiency, low cost and novelty (Labs 2017). According to Weaver (2016), the founder of an AI operation system company, claims that AI technologies not only has the ability to learn from big data and derive the pattern to make prediction, but also are capable to provide more efficient as well as novel services that creates enormous added-value. Moreover, hospitality, as a labor intensive industry, robots powered by AI technology can greatly reduce the labor cost for the long run of the hotel operation, which has been considered as another one of the significant benefits for adopting service robots in hotels (Kuo et al. 2016). Hence, the great potential of industrial AI robots application have been recognized by numerous enterprises including service industry (Labs 2017).

1.2 Problem Statement and Research Objectives

Despite the advantages and perceived potential, for hospitality industry, the practice of attaching AI robots with hotel operation is still not so prevalent in the real practice compared with other industries (Weaver 2016). In this case, together with opportunities, both hoteliers and scholars are facing challenges and uncertainties in the meantime.

Therefore, this research aims to understand comprehensive user expectation to AI applications for full service hotels by gathering the detailed information from various dimensions. More specifically, it is based on service quality appraisal tool, and focuses on finding the pattern of interaction and expectation regarding different demographic factors and the interrelation between all survey variables.

2 Literature Review

2.1 Service Quality Management

In the hospitality industry, service quality is a significant factor for the success of a hotel, and the perceived high service quality directly increases guest satisfaction, which helps create profitable business, brand loyalty, positive word-of-mouth as well as customer relationship (Knutson et al. 2010). As one of the most widely used instrument to evaluate service quality in various industries, Parasuraman et al. (1985) devised SERVQUAL Model and Service Gap Model, defining that the service quality is the consequence between customer expectation and customer perception. After that, Zeithaml et al. (1990) refined the dimension of service quality respectively into Five Dimensions: tangibles, reliability, responsiveness, assurance and empathy respectively (Fig. 1).

Fig. 1.
figure 1

SEVQUAL 5 dimensions

Full size image

Tangibles.

For current application, Sacarino, a service robot, providing luggage delivery service to the room as a bell and offering the enquiry service for the guests about the hotel information (Zalama et al. 2014). For contact centres, AI system has been deployed in many service sectors (Kirkpatrick 2017), which highly increase the efficiency of solving routine questions and intellectually analysis the emotion of the speakers that is beneficial for improving the customer service. Bilgihan et al. (2010) stated that in-room entertainment technologies has the same weight as hotel operation to the guest satisfaction, and they are developing and changing in a very fast speed though many up-to-date High-Tec appliances are not in a high priority among guest choices as expected. In the robot-staffed hotel in Japan, all receptionists are played by three separate intelligent robot (“Welcome to the Robot Hotel: robots do most of the work at a new hotel in Japan,” 2015).

Reliability.

For robot service, as same as human service, it has the concern regarding reliability, especially for those task-specific robots such as delivery robot and lifting robots (Stoller et al. 2016).

Responsiveness.

Instead of measure the communication between human staff and guest, here the survey tends to measure the expected involvement between the machine and human beings.

Assurance.

Concerning service assurance dimension, the first point is the language. Although the IT industry has made a huge progress on language understanding, according to the research by MIT, the great obstacle affecting task accomplishment is the transformation of language into instruct (Knight 2016). Another concern for customer to use technological service is about the data security (Park et al. 2016). Besides, based on Technology Acceptance Model (TAM), the ease to use is an essential consideration of service assurance as well (Davis 1989).

Empathy.

Except for the existing question regarding customized service aspect, AI service has a special problem between SCIENCE and HUMANITY (Kakoudaki 2007). According to Kakoudaki, robot is a combined figure challenging not only technology but also traditional culture, value and justice, which are highly related to the ethical issues.

Despite the use of SERVQUAL in a wide range of industries, a number of refined models based on SERVQUAL have been developed to solely target on hospitality industries such as LODGSERV, HOLSERV, LODGQUAL and DINESERV. LODGSERV (Patton et al. 1994) and HOLSERV (Mei et al. 1999) are both for the measurement of accommodation industry. Knutson et al. redevised SERVQUAL five dimensions and then developed LODGSERV instrument, which is to focus on service quality in the lodging industry only (Getty and Thompson 1994). Although there are various schools of thought used to specially measure service quality in the hospitality industry, SERVQUAL is deemed as the most fundamental and reliable instrument having the most primary and comprehensive criteria (Stevens et al. 1995). For measurement of AI application in hotel industry, SERVQUAL is more appropriate to be selected as the basis template.

2.2 AI Robot Development

Even though AI application is still at the very beginning stage, a large number of application has been launched in the hospitality industry. In the near future, there will be increasingly advanced robot being introduced and tested as well (Fig. 2).

Fig. 2.
figure 2

AI development status in different countries

Full size image

3 Methodology

3.1 Research Design

A descriptive quantitative research based on a survey design was conducted, because it best served to answer the research questions and the purposes of the study, which is to collect market voice based on SERVQUAL dimensions toward staying expectation in robotic hotels. The design of the questionnaire was mainly a closed questionnaire firstly gathering the general willingness of staying in AI-robotic hotel, and then followed by modified survey questions based on SERVQUAL model to investigate the intrinsic motivation and how can they relate to the general willingness (Fig. 3). The questionnaire was structured into modified Likert fashion. To the expectations respondents should provide points from 1–5. 1 meant a very low level for expectations and 5 indicated the highest level of expectations. The specific content of questions was modified in accordance with the robotic hotel service blue print as well as the realizable application of AI technology context. There was one open question gathering the expectation of potential service robot. In the end, three parts of information were gathered for segmenting the market. Demographic information including age, gender, educational level and income level was the first part; traveling behavior and attitude toward AI robotic hotel were gathered in the following to also help formulate the market segmentation for AI-robotic hotels. Due to the high validity of SERVQUAL model, no pilot study was conducted. All data were primary data that would be used in data analysis process.

Fig. 3.
figure 3

The questions in the survey based on the SERVQUAL model

Full size image

3.2 Sampling Design

The target population defined for this study was the adult residents living in Beijing city and those who had lived in Beijing for more than one year. As Beijing is the capital city of China, as well as the center of culture, education and technology, the residents were expected to understand IT and High-Tec notion better than other tier 2 and 3 cities, and to have more chances to travel and stay in a hotel. As a result, the technological knowledge combing hotel experience can lead to a more effective conclusion that are more representative for the whole developed cities in China where there are more potential customers who have higher purchasing power. This is also more appropriate to be used to deduce future market potential for China.

Regarding the sample size and sampling method, since this research was limited by the time, budget and geographical convenience, it was impractical to use random sampling method to cover all population in Beijing. For this reason, the research was set a margin of error of 0.05. Thereafter, the survey was made into an online-based e-questionnaire, distributed by using a snowball sampling method to facilitate the dissemination, and finally reached a sample size of 108 in total. Eventually, to enhance the validity, those surveys that costs less than 3 min were excluded. Finally, 102 of them were chosen as valid samples, based on the time limit of answering the questionnaire.

3.3 Data Analysis Design

The data collected from the field were analyzed by four dimensions. First of all, a descriptive data analysis was conducted to understand the demography of respondents, travelling behavior and attitude, and then the average willingness and expectation level toward staying in the robotic hotels were calculated. Besides, the average expectation toward the service hardware was calculated as well to understand the specific needs from the market. For the second dimension, mean tests (t-test and one-way Anova) were used to detect the specific difference between various groups in terms of socio-demography, travelling behavior and attitude toward AI-robotic hotel, aiming at formulating the proper market segments for AI robotic hotel. For the third level, a correlation analysis was made to detect whether there is internal relation between SSERVQUAL five dimensions and living expectation. Lastly, a regression analysis based on the data from SERVQUAL and general willingness was made to understand the essential interrelationship among them.

To ensure the validity of questions, they were strictly based on standard questions used in original SERVQUAL model. Since SERVQUAL has been proved to be a valid tool and model to measure service expectation, the validity of questions should be very high. To ensure the validity of the results, all respondents are strictly selected by the requirement measured by time of answering questions; only the process lasting for more than three minutes could be chosen as valid sample.

4 Results

4.1 Descriptive Data Analysis

In terms of socio demography, the proportion of the gender distribution was relatively balanced, with 51.9% (male) and 48.0% (female). More than 70% of respondents are people below 40 years old. The majority of sample (77.4%) had bachelor degree background. For income level, the majority is below 12000 yuan per month (Table 1).

Table 1. Socio-demographic profile of respondents
Full size table

Based on Tables 2 and 3 (Fig. 4), around 59% of travelling activities were leisure ones and 41% were business and others. For the attitude toward robotic hotel, 63% of respondents had heard the notion of AI robotic hotel. Among them, 72% showed the interest in staying in the such hotels. Other than those who fell neutral toward this idea, only one respondent felt negative toward the idea of robotic hotel. Table 4 (Fig. 4) shows the overall willingness level toward staying in such hotels, and the final result was 3.73. Table 5 (Fig. 4) shows expectation level toward specific service hardware in future AI robotic hotel. In room butler and entertainment robots was the only one that had an expectation value over 4. Table 6 shows the average value of SERVQUAL five dimensions.

Fig. 4.
figure 4

Table 2. Traveling behaviors and attitude of respondents; Table 3. Respondents attitude toward AI robotic hotel; Table 4. General expectation toward staying in AI robotic hotel; Table 5. Respondents expectation toward specific AI robot’s application (n = 102)

Full size image
Table 6. Average value of SEVQUAL five dimension
Full size table

4.2 Reliability Test

The research questions had an acceptable reliability after the modification based on SERVQUAL five dimensions. As can be seen from Table 7, all of the Cronbach’s Alpha were over 0,7, which means the data were highly reliable and able to be used in further data analysis process.

Table 7. Cronbach’s Alpha value of the research questions
Full size table

4.3 Mean Test

Independent Sample T-test

Mean test was conducted to find the difference between various groups to set the market segmentation. By conducting independent samples t-test, Table 8 shows that there was no significant difference in staying expectation between genders, and whether they have heard about the notion of robotic hotel made no difference either (Significance values were both over 0.05). Meanwhile, there was no significant difference between leisure and business groups as well, with p value of 0.673. The results of T-test show that in-group average expectation was not affected, to the extent of statistical significance, by factors of gender, awareness or travelling type.

Table 8. Independent samples T-test results: sex v.s. staying expectation; heard v.s. not heard; leisure group v.s. business and other groups
Full size table

One-Way ANOVA

Table 9 (Fig. 5) shows that there is no statistically significant difference between means (p-0.097) in different age groups. However, there was statistically difference between group’s means by the criteria of educational level (p-0.006). Then, post-hoc test was conducted by using Scheffe method for these two groups. It could be seen in Table 10 (Fig. 5) that there was a significant difference between people having high school and below diploma and bachelor degree people toward the willingness to stay in an AI robotic hotel.

Fig. 5.
figure 5

Table 9. ANOVA: age group and educational level; Table 10. Multiple comparisons (educational level); Table 11. ANOVA (Attitude); Table 12. Multiple comparisons (attitude); Table 13. ANOVA (Income level); Table 14. Multiple comparisons (income level)

Full size image

In terms of attitude, it was found that among different types of attitude groups there was significant difference (p-0.01). By conducting post-hoc analysis, it could be clearly seen that people who are interested in the concept of IT and robotic related things have significantly different expectation to stay in robotic hotel to the group who hold neutral position toward the same topics and who have never heard this concept (Tables 11 and 12 in Fig. 5).

As determined by Table 13 (Fig. 5), the Income level had significant effect which was at .003 with an F statistic of 4.998. It could be interpreted that the income level had a significant influence on the overall expectation to stay in robotic hotel. Seeing from the post–hoc test, for the income groups, there was statistically significantly different expectation for low-income people (4000 and below) to people who have 8000–12000 monthly income. In conclusion, education, attitude and income level of consumers take significant effect on the expectation to stay in robotic hotel. Thus, in promotion process the hotel management should focus on these three dimensions to carry out market segmentation and market positioning.

4.4 Correlation and Regression Analysis

One main hypothesis of this study was SEVQUAL five dimensions affect the overall expectation level in staying robotic hotels. Pearson’s correlation analysis was employed to test the hypothesis and the results were presented in Table 15. It was hypothesised there is a positive correlation between the SERVQUAL 5 dimensions and the overall expectation level. The final result supported for this hypothesis (r = .657, 442, 602, 438, 501, and all p are less than .001). Based on these results, it supported the need to conduct further regression analysis as well. What’ s more, Pearson Correlation corfficients between SERVQUAL five dimensions significantly reach over 0.5, which shows that these 5 factors may affect each other.

Table 15. Pearson correlation coefficients results for overall expectation and SERVQUAL five dimension (n = 102)
Full size table

As all five dimensions are positively related with the general expectation level, a multiple regression analysis was conducted to find out whether SERVQUAL five dimensions can effectively predict the expectation level. Using the enter method it was found that SERVQUAL model did significantly predicted the expectation level (p < .05, R2 = .499, R2Adjusted = .472). From the Table 16, the simulation model of SERVQUAL passed F-test, showing that the equation established was efficacious.

Table 16. Model summary
Full size table

It can be found in the Coefficient table that only Tangibles (p < .05, beta = 0.566) and Responsiveness (p < .05, beta = 0.359) explained a significant amount of the variance in the value of expectation level. Each unit increase in Tangibles and Responsiveness expectation would lead to 0.138 and 0.353unit increase in the overall expectation level. In comparison, the responsiveness expectation played a more important role. However, other three factors did not significantly predict the value of staying expectation (Table 17).

Table 17. ANOVA (regression test)
Full size table

Combined with VIF values in Table 18, results in the collinearity diagnostics show that there was multicollinearity relation between factors of the formula. It can be found that Eigenvalues of dimension 2 to 6 approximate to zero and condition Index values are more than 10. As far as multicollinearity problem, on the one hand, insufficient samples may lead to weakening of statistical characteristics. On the other hand, the factors depict 5 dimensions of consumer expectation, which are closed related in the real sense (Table 19).

Table 18. Coefficients
Full size table
Table 19. Collinearity diagnostics
Full size table

Table 20 was conducted to examine residuals of the model. The absolute value of Standard residuals remained below 3, showing no unusual value. In the normal P-P plot, the observed residuals distribute around standard normal distribution curve. It can also be proved by the Histogram for the residuals that they are normally distributed in line with the condition of regression hypothesis. Figure 6. shows that most of the model residuals remain between −2 to 2, able to effectively explain predicted value of the dependent variable in the formula. In sum, residuals of the model apply to basic regression hypothesis.

Table 20. Residuals statistics
Full size table
Fig. 6.
figure 6

(A) Normal P-Plot of regression standardized residual; (B) Regression deleted residual scatterplot

Full size image

5 Discussion

5.1 Implications of the Study

First of all, based on the descriptive result analysis, it can be seen that at current stage, the general willingness toward staying at AI robotic hotel was between “neutral” and “expected” with average expectation value of 3.73. This means that for the general population as a whole, the market expectation toward living in an AI-based robotic hotel was not agreed to be very high. Thus, if currently an experimental business is established, market segmentation has to be well conducted to cater to those users who have better living wiliness.

In terms of market segmentation, based on the mean test, by socio-demographic factor, it can be seen that people who generally have higher educational level and income level are more likely to have a try in robotic hotel. Age and gender are not decisive factors affecting the willingness of staying in an AI Robotic hotel. In my opinion, the reasons include that high-income level people normally have higher educational background and are more confident on handling High-Tec and IT-related things. Therefore, education level might be the ultimate factor affecting the staying willingness. This hypothesis needs to be tested by further study.

In terms of travelling behaviors, there was no significant difference between leisure group and business group. It can be inferred that whether the aim of traveling is for enjoyment, business, and others, it did not manifestly affect the staying willingness. In other words, according to this study, the robotic hotel can be designed for people with any aim of travelling. Thus, for potential entrepreneurs, AI-based robotic hotels can be applied to business hotels, resorts and various types of hotel, because the study did not show a significant difference among people in different traveling groups.

For attitude toward AI robotic hotel, there was no significant difference between the people who had heard the idea and people who did not. Nevertheless, the people who has the interest of the AI and technology related notion had significantly higher staying expectation compared with those who has mild interest in such fields. In other words, people who are positive toward technology are more likely to be attracted by such hotels. Thereby, the promotion of the idea “AI-robotic hotel” should specially target on the group who is highly interested in and positive to IT topic (Fig. 7).

Fig. 7.
figure 7

General persona about target users of AI-based service hotels

Full size image

Among all current AI hardware, in-room entertainment robots were liked at most with an expectation average of over four, which is followed by the delivery robot. In this case, entertainment robots should be deemed as a realizable and welcomed hardware by the general market.

Another significant aim from this study is to understand how SERVQUAL five dimensions formulate and predict the overall willingness. From the study, we can conclude that Tangibles and Responsiveness are two significant factors positively affect the general expectation. Based on the survey, it is found that normally for people who is willing to stay in robotic hotels usually because that they have a higher expectation on tangibles and responsiveness. Based on this logic, it can be concluded that the intrinsic motivation of staying in robotic hotels is due to the expectation of trying the new service hardware and experience the communicational process with them. Thus, for industrial designer, it should focus on optimizing the hardware design and produce more fancy elements. Besides, communicational functions should be primarily enhanced to strengthen the interaction between the users and robots. For hoteliers, key factors increasing the potential of success are also the same points. When designing the service process and blue print, they should definitely focus on these two points, because they are highly related with the expectation level. the foremost is the service hardware that are attractive and special.

5.2 Limitations and Future Studies

The first and foremost limitation of this study is the sample size. Limited by the time, finance and access to relevant information as well as relevant resources, only 102 valid samples were collected. This, to some extent, affects the trustworthy and quality of the findings. After that, failure to use a probability sampling technique also limits the ability to make broader generalizations from the results and to make statistical inferences from the sample. In the future, a large-scale study conducted by random sampling method is expected to be conducted to verify the study result.

Another limitation of this study is the validity of the model modification. Although the survey questions are mainly based on SERVQAUL model that has high question reliability, some modification cannot be proved that has a similar validity compared with the original model. To enhance the model to evaluating robotic hotels still has a long way to go; it could be improved by more validity test methods such as professionals and literature verification. Pilot study for relevant professionals can also be conducted for more advices.

Last but least, for this survey, about some questions, it is quite abstract to describe the robotic service. There might be a gap existing and misunderstanding between the researcher and respondents. A more comprehensive description for those applications can be written for the future study to minimize the discretion.

In the future, a fixed-interval survey can be conducted regularly, such as 2 years per time, which is to monitor the change of expectation level of staying in robotic hotel as well as the preference for guests. With the development of service hardware and more practice in this field. It is expected to see a positive trend toward AI- based robotic hotel. Meanwhile, similar study could be conducted in different cities all around the world to detect whether similar pattern exist in difference cultural and socio-demographic background. The expectation might also vary from distinct economic scale. All these questions can only be answered by obtaining the voice from the market.

6 Conclusion

This study firstly set out to determine the current market expectation toward AI appliance in hotel context. It is reached that the current expectation level is between neutral and expected. Nevertheless, this study has found that some specific groups are more willing to try AI robotic facilities. They will be called as the targeted market segments for AI robotic hotel. As a conclusion for the market segmentation and target segment, the groups of people significantly expecting staying in AI robotic hotels are those who: (1) Highly interested in the notion of AI and technology-related things (attitude dimension) (2) Have higher educational background such as bachelor and above (sociodemographic dimension (3) With more than average monthly income (sociodemographic dimension)

However, through this study, business and leisure guests did not show a significant difference toward the willingness. Therefore, at this stage, hoteliers should specially target on the focused groups stated above and conduct relevant positioning strategy to cater to these people.

In terms of specific service hardware, in-room entertainment robot is the most welcomed robot, followed by delivery robots. This result provides the reference for designing the specific service robots in AI robotic hotel for both supply chain and hotelier.

The ultimate objective of this study is to understand how expectation toward robotic hotel formed from an academic side is. The results of the findings support the hypothesis of SERVQUAL model can explain the expectation level of staying in a hotel. Based on the regression test, it also proves that tangibles and responsiveness are two significant factors positively affecting the willingness of staying in such hotels. Thus, in order to improve the market attraction and meet the expectation of customer, these two focus points should be concentrated for promotion and further improvement. In other words, service hardware and the human-machine interaction are the areas where the market should pay the most attention to. For academic sides, this result helps predict the expectation value and build the basis for refining the new service evaluation model to specifically evaluate the expectation and perception in AI robotic hotel context.