Study design
We used a social survey to collect data concerning elderly people in China using structured questionnaires and interviews.
Structured questionnaires
Structured questionnaires are also called standardized questionnaires. This method of data collection was pioneered by Francis Galton [8]. It has many advantages in terms of collecting data, such as its high efficiency, objectivity, consistency, and wide range [9]. All questions in the structured questionnaire were designed within a standard framework. The format, question content, and order of the questionnaire were all fixed and used to collect information from respondents.
The structural questionnaire adopted in this paper includes two parts: basic information and robot acceptance. Basic information includes age, gender, marital status, education level, health status, and housing arrangements. The robot acceptance questions addressed personal acceptance of robot services, and what types of services they were able to receive from the robot.
Structured interviews
Structured interviews, also known as standardized interviews, are interviews conducted according to unified design requirements. In a structured interview, the entire research project is highly standardized during the design, implementation, and data analysis stages. Specifically, structured interviews have uniform requirements with respect to the criteria and methods of selecting interviewees, the designed questions asked during the interview, the manner and sequence of questioning, the manner in which the interviewees answer, and the manner in which the interview is recorded. Sometimes even the time, place, surroundings, and other external conditions of interviews are kept consistent across all interviews. Control of irrelevant variables is emphasized [10].
Structured interviews have specific advantages. First, they can reduce the number of “I don’t know” responses and refusals to answer, and they are suitable for research into sensitive questions. Second, through structured interviews, researchers can observe the words and nonverbal responses of interviewees, which is impossible in questionnaires. Third, the recovery rate and efficiency of survey interviews are much higher than that of questionnaires filled out by the respondents [10].
During the social surveys, certain elderly people expressed their views on the degree of robot acceptance and provided their specific reasons. Investigators conducted structured interviews with these elderly individuals who were willing to undergo the interview and recorded these participants as typical cases.
Stratified random sampling
We used stratified random sampling to sample the elderly in different areas. Stratified random sampling stratifies the survey population according to a certain standard. We then used simple random sampling to collect samples from each stratum for investigation. The samples in this paper are divided into four layers. We selected urban population size as the basis of the first layer of stratification, districts as the second layer, streets as the third, and communities and institutions as the forth layer.
During the first stage of stratified random sampling, random sampling was performed in mega-cities, XL cities, and I cities. During the second stage, random sampling was performed in urban districts (counties). During the third stage, streets were randomly sampled, and in the final stage, sampling was performed in communities and nursing homes. The samples included people 60 years of age or older.
This paper used Feng’s sample size formula [12]:
[Due to technical limitations, please see the formula in the supplementary files.]
In this formula, is the sample size, is the critical value corresponding to confidence, and is the allowable sampling error. The confidence interval in social research typically is 95%; thus, t = 1.96. When the population exceeds 500,000, the sample size should be greater than 800, . To keep the sample error to a minimum, we determined that a sample size of 3000 would be needed (e = 1.83%). Therefore, we surveyed 3060 individuals to assure that attrition, incomplete data, and invalid responses would not lower the total number of surveys to less than 3000 (Fig. 1).
Survey implementation
Due to the large amount of information to be collected and the large geographic distances between the investigated cities, the survey was conducted in August 2015 and July 2016. The investigation team contacted each local government in advance, and the local government helped coordinate with the communities and the relevant personnel at the pension institutions. They allowed the investigation team to conduct the social survey but they did not inform the participants in advance so they would not prepare answers to the questionnaires. During the investigation, the participants were informed that this investigation had been approved by the government.
From August 15 to 23, 2015, the research team conducted surveys in Suzhou, Xi’an and Baoji. A total of 1546 elderly people received the structured questionnaire; 20 structured interviews were completed, and 23 elderly people declined to take the survey.
From July 20–28, 2016, the research team carried out 1534 structured questionnaires in Beijing, Nanjing and Yulin; 30 structured interviews were compeled, and 45 elderly people refused to take the survey. The reasons that the elderly declined to accept the structured questionnaire generally included the following: they did not want to reveal any private personal information or they did not know about service robots.
In the social survey, we recruited 40 investigators, including undergraduates, postgraduates, and doctoral students who had social survey experience as interviewers and whose own research field was also elderly service. Each investigator was paid. They were trained before the survey to unify their knowledge of each question. During the social survey, each investigation team had two members, one of whom asked the questions using the questionnaire and filled in the respondents’ answers. The other investigator recorded the question-and-answer process, including which questions the participants were confused about and which questions turned out to be ambiguous. Structural questionnaires were usually conducted outdoors, and some participants were willing to participate in household surveys and underwent structural interviews in their homes. Investigators gave older people small gifts, such as soap or towels, to encourage them to answer questions.
At the end of each day, the team organized the data according to the recorders’ records. Invalid answers were removed and the valid answers were retained. The specific survey times, places, and number of surveys collected are shown in Table 1.
Table 1 Survey implementation
|
Time
|
Place
|
Quantity of sample
|
|
August, 2015
|
Suzhou Xi’an and Baoji
|
1546 participants
|
|
July, 2016
|
Beijing, Nanjing, and Yulin
|
1534 participants
|
Data analysis
The statistical software SPSS (Statistical Product and Service Solutions) was used for data entry and data analysis. Descriptive statistical analysis data are used to identify the proportion of age, gender, marital status, level of education, health status and housing arrangements of the elderly, as well as the acceptance rate of the elderly to the robot.