1      Introduction

Social media provide considerable benefits for social and economic developments. Users optimizes their performance on online platforms with convenient and omnipresent access. However, increasing prevalence of social media application has challenges, such as cyber security and privacy risk. The protection of customer data is one of top priority of users’ concern. Moreover, there is a growing recognition that many negative effects on users are also present. This can involve illegal use, cyber- bullying, undesirable content, distorted news information, and rising levels of despair and worry.

This objective of research project is to investigate:

• How often Facebook shows happy moments or sad moments for users
• Whether the level of education are associated with users’ actions on what showing up their Facebook News Feed
• Whether users age groups affect the degree of their acceptance when trading off data to receive better online experiences

2      Research questions

RQ#1: Do users think Facebook more frequently shows them posts that remind them of a happy (Happy) time in their life, or a sad (Sad) time?

RQ#2: Is there a relationship between the level of education (Education) and users’ actions have taken to optimize better their personal Facebook News Feed? These actions are: (a) managing friending or unfriending certain people (FriendList), (b) liking, sharing or commenting on certain types of content (Comment), (c) indicating they want to see less of certain people or types of content (Request), (d) changing their privacy settings or ad preferences (Privacy), and (e) following or unfollowing certain groups or organisations (Following)?

RQ#3: Does user age (Age) affect how open they are to social media sites using their data and online behaviours to: (a) suggest events in their area (Recommend), (b ) show them product or service ads (Products), (c) suggest someone they may like to meet as a friend (Friends), and (d) show them political campaign messages (Politics)?

3      Research methodology

In order to clarify the research questions, the study performs a survey with Facebook users to investigate users’ behaviour. The data provided will allow you to understand if age influences how receptive users are to social media sites using their data and online activities, whether education level affects the actions users have taken to try and influence what shows on their Facebook News Feed, and whether respondents think Facebook shows them posts that remind them more of a happy time in their life, or a sad time.

Data was collected by a major US Research Centre in 2018, from a probability-based online panel of adults aged 18 or older living in households in the United States. The sample participants were asked to respond to a number of questions on their social media behaviour and related attitudes. Demographic information was also collected. The data set provided for this project is a subset of the total dataset and includes 781 randomly chosen respondents indicating they had an active social media profile and were Facebook users. The project data includes usage of social media sites, feelings about social media content, and Facebook activity; as well as various demographics including their location, age, family income, race, religion, marital status, education level and gender.

4      Analytical results and Findings

Along with research questions, we apply appropriate quantitative methods to figure out the findings of the report. We explain the findings in the Analytical results sections, and detailed results are shown in Technical Appendices sections.

4.1     Research question 1

4.1.1        Analytical results

In detailed, the survey carries out questioning of how often Facebook reminds users of happy or sad memories with the Likert scale 1-4: 1 = Frequency, 2= Mostly often, 3= Rarely, 4 = Never, 99 = Refused/ Don’t know. We take a look at descriptive analyses of the frequently shows them posts that remind them of a happy (Happy) time in their life, or a sad (Sad) time in Table 4‑1. The table shows that the mean of the frequency of Happy time is much lower than that of Sad time (1.96 < 2.75), hence, users are reminded more often of happy moments rather than sad ones in Facebook. The frequency of reminding of happy time is asymmetric distribution with right-tailed (Skewness = 0.765) while that of sad time is nearly symmetric data due to a skewness near zero.

Figure 4‑1: The frequency that Facebook reminds users of Happy time and Sad time

Then the study conducts the frequency distribution of Happy time and Sad time in Facebook in Table 4‑2 and Table 4‑3. The Figure 4‑1 above shows the frequency that Facebook reminds of Happy time and Sad time of users. The majority of users contend that Facebook remind them of happy moments frequently and sometimes often while about 20% admit that they are rarely and even never recalled happy memories by Facebook. Moreover, 38% users state that sad times are often recalled by Facebook while the rest of users rarely or never reminisce about sad memories due to Facebook.

The Paired-sample T-test is showed in Table 4‑4 to investigate the difference between two variables related to Happy time and Sad time reminded by Facebook. With the 95% confidence interval, the p-value is less than 0.05, thus, there is a statistical difference between happy time and sad time that Facebook reminds users in life. The mean difference of two variables is 0.79. There is a difference in the degree to which Facebook recall happy or sad memories, or in other words, Facebook tends to remind users of positive emotions over negative.

4.1.2        Technical appendices

Table 4‑1: Descriptive statistics

Table 4‑2: Frequency table of Happy time in Facebook

Table 4‑3: Frequency Table of Sad time in Facebook

Table 4‑4: Paired Sample T-test of Happy time and Sad time

4.2     Research question 2

4.2.1        Analytical results

In this section, we gain insight about users’ behaviours related to their personal Facebook New Feed with the question: “What actions have you taken to try and influence what shows up in your Facebook News Feed?”. There are five action groups including (i) Monitoring friends in Facebook, (ii) Controlling likes, shares, and comments, (iii) Optimizing the type of contents showing up, (iv) Changing privacy setting, and (v) Following groups. Five actions are coded by Friend List, Comment, Request, Privacy, and Following respectively. Any blank answers or selections means that users have not tried to affect and modified their personal Facebook New Feed. The Figure 4‑2 shows that the percentage of users who try to optimize their New Feed and otherwise. Approximately 30% users, on average, undertake to control settings and optimize their personal New Feed. The “Comment” action is gained the least attention and efforts of users to influence what shows up in their Facebook even though the difference between actions are not significant.

Figure 4‑2: The frequency distribution of users who try and influence on Facebook New Feed

In order to investigate the association between Education level and users’ actions to control their Facebook, we conduct the Pearson correlation to identify the direction and strength of relationship between these variables based on the method of covariance. The Table 4‑5 indicates that the education level of users has statistical relationship with Friend list, Request, and Following actions (Significant value p < 0.05). The correlation coefficients between Education level and three actions are -0.188, -0.236, -0.163 in respectively. There are weak negative associations between users’ education level and their actions on Facebook. Users who are higher education tend to be more concern about what shows up on Facebook New Feed, and be willing to modify settings in associated with their preference.

Then, we apply ANOVA test to identify whether users’ actions have difference by the level of education. The significant value of ANOVA tests (p < 0.05) indicate the difference in users’ actions to control Friend list, Request, Following among education groups, . We use LSD Post Hoc to generate multiple comparison of mean of these users’ actions by education category and check which groups differ in Table 4‑6, Table 4‑7, and Table 4‑8. Regarding Friend list actions, Table 4‑6 shows that only H.S Graduate group and College Graduate group have different actions about how users friend with certain people on Facebook ( p = 0.001 < 0.05). Table 4‑7 indicates that there is a significant difference in mean of Request actions between H.S graduate and College Graduate (p = 0.00 < 0.05), between H.S graduate and Some College (p = 0.002 < 0.05). Table 4‑8 illustrates that H.S Graduate group and College Graduate group have significant difference in managing to follow certain groups on Facebook ( p = 0.004 < 0.05).

Figure 4‑3: Mean plots of Friendlist, Request, and Following elements among Education level category

Figure 4‑3 shows that the mean plots of users’ actions among education level category. Users who achieve college and high education are highly aware of managing what appears on their personal Facebook ().

4.2.2        Technical appendices

Table 4‑5: Pearson Correlation between Education and Users’ actions

Table 4‑6: ANOVA Post-Hoc test of Friendlist among Education level category

Table 4‑7: ANOVA Post-Hoc test of Request among Education level category

Table 4‑8: ANOVA Post-Hoc test of Following among Education level category

4.3     Research question 3

4.3.1        Analytical results

Some of people agrees to trade off a particular amount of data in order to perceive usefulness of Facebook, as a desirable compensation. In the section, we investigate the degree of acceptance on data sharing for specific purposes including content recommends, products, friends, and politics. We use Likert scale 1-4: 1 = Very acceptable, 2 = Acceptable, 3 = Neutral, 4 = Not acceptable at all, 99 = Refused. Figure 4‑4 shows that the degree of acceptance of users on data sharing for specific purposes. Facebook wins acceptance of 80% users to trade off data for recommending better contents while about 20% users are not willing to sharing data to gain recommend benefits. By contrast, 60% users oppose Facebook to collect and use their data to serve political purposes. Regarding showing suitable advertisements or friends as Facebook benefits, about half of users accepts and another half refuses.

Figure 4‑4: The frequency distribution of the degree of acceptance on sharing data for specific purposes

In order to investigate the association between Age groups and users’ benefit when experiencing Facebook, we conduct the Pearson correlation to identify the direction and strength of relationship between these variables based on the method of covariance. The Table 4‑9 indicates that the age group of users has statistical relationship with Content recommends (Recommend), Products ads (Products), Friends recommends (Friends) (Significant value p < 0.05). The correlation coefficients between Age groups and three benefits are 0.167, 0.109, 0.293 in respectively. There are weak positive associations between users’ age group and their purposes gained from Facebook. Users who are older tend to be more concern about privacy issues, thus their acceptance on Facebook is lower than.

Then, we apply ANOVA test to identify whether users’ acceptance on data sharing have difference by the group of age. The significant value of ANOVA tests (p < 0.05) indicate the difference in users’ acceptance on data trade off among age groups, . We use LSD Post Hoc to generate multiple comparison of mean of purposes by age category and check inter- groups difference in Table 4‑10, Table 4‑11, and Table 4‑12. Regarding Recommend purpose, Table 4‑10 shows that younger generation (18-29 age groups) and older one (50+ age groups) have different willingness about sharing their personal data to receive better contents ( p = 0.00 < 0.05). Table 4‑11 indicates that there is a significant difference in mean of Products benefit between 18-29 age group and 65+ age group (p = 0.004 < 0.05), between 50-64 age groups and 65+ age group (p = 0.004 < 0.05). Table 4‑12 illustrates that 65+ age groups have significant difference in managing data to recommend friends in comparison to other age groups.

Figure 4‑5: Mean plots of Recommend, Products, and Friends purposes among Age category

Figure 4‑5 illustrates that users who are 65+ age group maintain cautious attitude against Facebook actions, thus they tend to show unacceptance on data trade off. By contrast, younger ones are less aware of data issues. In other words, they assess Facebook benefit over data issues, and show more willingness to sharing data to enhance online performance.

4.3.2        Technical appendices

Table 4‑9: Pearson Correlation between Age and The purposes of sharing data

Table 4‑10: ANOVA Post-Hoc test of Recommend

Table 4‑11: ANOVA Post-Hoc test of Products

Table 4‑12: ANOVA Post-Hoc test of Friends

5      Recommendations

Privacy awareness is essential for users whenever they encounter the situation that required them to seriously consider the intention and the amount of data, they should disclose to Facebook. Whereas the data disclosure refers to the personal decision and the threats of possible interruption caused by other parties, therefore, should encompassed within personal scale. There is a paradoxical issue. Thus, Facebook must account for the sensitivity of community in designing tools for privacy-awareness. Those supports should help to increase available privacy-relevant information in order to balance the scale. We propose some of possible approaches in the following:

Regularly measuring users’ privacy concerns: The primary responsibility of Facebook must not stop at protection, but also, include educating and reminding users about their privacy attitude no matter what education levels or age groups. Nonetheless, Facebook seems to be struggle in finding suitable ways to assess citizen privacy attitude. Normally, in capturing the privacy preferences of people, their choices are limited in two options only: (i) ask them directly or (ii) gather preferences from observation of actual behaviour. However, monitoring of the behaviour might be privacy-invasive itself and, second, asking citizen directly might not satisfy objective criteria, thus, deprive effectiveness of the survey.

Transparent privacy policy: A privacy policy set the way in binding manners to secure users privacy, and establish legal right between service providers and users. Facebook should put effort to produce a privacy policy as legal document that is both specific and easy to understand, and it is not appropriate to obscure clauses between sentences. Regarding the risk perception of users, Facebook should provide clear information, and even ask users about what kind of information that need to be collected, and stored, and what purposes of providers. A privacy policy is necessary on any site that collects any data about its users, even for providers simply collect information related to users servers such as browser cookies and IP addresses in attempt of targeting suitable contents based on users’ habits. As these programs include the sharing of information, service providers are expected to have transparent privacy policy published before being eligible to participate in any program. The specific data information collected varies based on the applicable policies and purposes. The definition of what purposes providers take advantage of users information should be presented evidently. In order to mitigate users’ concern to use Internet service, Facebook should require users to ensure that they have read the privacy policy before they accept to provide their information to use service. It may be obvious that users have a right to know personal details collected. If there are data breach incidents, Facebook should commit a right and obligation related to security measures to protect data for each party.

A flexible approach to diverse groups: The choice of methods and approaches need to be friendly and understandable for different groups of users, not only for those who have experiences in this field. In fact, the threats of losing data is equal for everyone, but the ways they perceive their losses and the value of each personal data is different. A majority of users is not experts and their level of technical knowledge differs as well. Further, the concept of “bounded rationality”, which is well known in cognitive science, signifies the limited ability of individuals to acquire, process, and remember information. That is, even if users would theoretically have all privacy-relevant information available, they will not be able to use all the information for making a rational decision, however they apply a simplified mental model handle it cognitively. By that means, it is not sufficient to rely on rigorous policies determine what may be useful to display and what may not, and how to inform people. It is important to identify and consider perception, awareness and needs of targeted groups to design a set of practical policies.