Public Response to “the MOOC Movement” in China: Examining the Time Series of Microblogging

In China, microblogging is an extremely popular activity and is proving to be an effective mechanism to gauge perceptions about social phenomena. Between 2010 and 2015 Sina Weibo, China’s largest microblogging website, generated 95,015 postings from 62,074 users referencing the term massive open online courses (MOOCs), a method of online course delivery popularized in North America that has spread globally. Time series analyses revealed distinct patterns in the volume of postings during a four-year period, and subsequently by month, by week, and by the time of day. The volume of postings during the week, for example, peaked on Monday and declined daily to a low point on Saturday. Relative to maximizing learner engagement, the findings may provide insight to parties who deliver MOOCs to employ or test strategies on timing (i.e., time of year to offer/not offer a MOOC, time of week to release/not release new material, time of day to schedule/not schedule chat sessions). The paper also serves to demonstrate a mechanism to retrieve big data from social media sources, otherwise underutilized in educational research.


Résumé de l'article
In China, microblogging is an extremely popular activity and is proving to be an effective mechanism to gauge perceptions about social phenomena. Between 2010 and 2015 Sina Weibo, China's largest microblogging website, generated 95,015 postings from 62,074 users referencing the term massive open online courses (MOOCs), a method of online course delivery popularized in North America that has spread globally. Time series analyses revealed distinct patterns in the volume of postings during a four-year period, and subsequently by month, by week, and by the time of day. The volume of postings during the week, for example, peaked on Monday and declined daily to a low point on Saturday. Relative to maximizing learner engagement, the findings may provide insight to parties who deliver MOOCs to employ or test strategies on timing (i.e., time of year to offer/not offer a MOOC, time of week to release/not release new material, time of day to schedule/not schedule chat sessions). The paper also serves to demonstrate a mechanism to retrieve big data from social media sources, otherwise underutilized in educational research.

Introduction
MOOCs are no longer a North American phenomenon ! 144 MOOCs are no longer a North American phenomenon. Since 2011, when Stanford University decided to offer several reputable courses online for free, MOOCs have expanded to engage learners around the world. The momentum is far greater than reported, with overseas learners enrolling in a MOOC at a preferred North American institution (Universities UK, 2013).
Institutions in at least 50 countries offer a MOOC (MOOC List, 2014), most commonly in partnership with a MOOC provider, such as Coursera or edX. The global appeal of institutions to design MOOCs domestically suggests there is a desire from learners for greater contextualization (e.g., curriculum, language, culture) that the more popular MOOCs in North America do not provide. The major challenge facing other providers of MOOCs is the persistent draw to the likes of Stanford or Harvard -unmatched in terms of prestige and perception of quality learning.
Newer MOOC providers that are situated outside of North America should continue to leverage their comparative advantage of contextualizing higher learning, but also engage more deeply into researching the behaviour of online learners. The big data that can be mined from MOOCs is invaluable, but underutilized. Considering the socio-economic returns on higher education, MOOCs present an alternative pathway for learning, particularly under conditions around the world where unmet demand persists.
This article aims to offer some empirical evidence relative to time series behaviour of prospective MOOC learners. The findings may be of value to prospective providers of MOOCs who are interested in utilizing time-oriented strategies to augment learner engagement online.

MOOCs in China: A unique context for delivery
Outside of North America, Chinese universities have been among the most active adopters of MOOCs. Like the immense interests in rankings and research output (i.e., publishing in high impact factor journals), the attention and resources devoted to MOOCs is, in part, a product of following western trends. Yet, like other imported technological trends, Chinese institutions have proven effective at altering phenomena with the infusion of Chinese characteristics. The product is altogether different, or even superior, as in the case of Alibaba (based on eBay) or Sina Weibo (based on Twitter). MOOCs have yet to follow a similar trend, but the social conditions in China seem highly favourable to utilize these learning tools in novel ways.
Among the many social metrics where China is atop, it is home to the world's largest internet population, at 640 million (Jacquot, 2015), and the world's largest higher education population, at 34 million (Ministry of Education, 2010, p. 22). Considering that holders of a university degree have shown to be the most active demographic enrolling in MOOCs elsewhere (Jordan, 2014) the appetite to enrol in MOOCs among China's educated is potentially big. A large proportion of MOOCs is offered by the country's top universities. The likes of Peking University and Tsinghua University have higher rejection rates for campus-based programs than Harvard or MIT (Wong, 2012). Their MOOCs present an opportunity that has been otherwise non-existent. Those who have completed some form of higher learning also have experience learning at an advanced level, increasing the chance of adaptation to an online learning environment. China's objective to create ! 145 a lifelong learning society over the next decade may also benefit from MOOCs as a means to reach the masses with some form of quality higher education.
Another segment of the population where MOOCs may prove beneficial is adult learners with little or no exposure to higher education. According to the National Outline for Medium and Long-term Education Reform andDevelopment (2010-2020), the government aims to reach enrolment of 350 million people into higher education or adult education by 2020 (Ministry of Education, 2010, p. 22). Campus-based solutions are clearly impractical opening up the possibilities for alternative pathways such as online learning that may include MOOCs.

The MOOC landscape in China
There are several prominent MOOC providers in China, namely, XuetangX, CNMOOC and iCourse163, which have links with edX and Coursera, respectively. Some institutions, such as Beijing Normal University, are offering MOOCs independently. The majority of these MOOCs are designed and delivered in Chinese; some are in imported from abroad and modified with Chinese subtitles or translations.
The research on MOOCs in China has generally focused on two aspects: 1) introductory studies on the MOOC curriculum model and definitions, characteristics, application modes, and case studies. For example, Li and Wang (2012)  Amidst the attention paid to MOOC providers and particular case studies from individual MOOCs in China, little focus has been directed towards aggregate data that may offer insights into the behaviour of learners enrolled in a MOOC. This is a reflection of both the infancy of research on MOOCs in China, and the challenge of acquiring adequate or big data to make informative generalizations.
Relative to this paper is the subject of time. Although MOOCs are characterized as being time independent, this is not always the case, nor is it always an appealing feature. Providing start and end dates to a MOOC remains the most common form of scheduling, and synchronous chat sessions between instructors and learners is extremely popular. The classroom culture, to which nearly all learners are accustomed, is often sought in some form when learning online. Although the flexibility of online learning is a draw for prospective learners, some consistency around timing may minimize isolation and attrition both of which characterize the darker side of MOOCs. Consider Chinese MOOCs. There is a high degree of uniformity with respect to learners who enrol in such MOOCs. They are predominantly located in China 3 , speak Chinese, and are immersed in its culture. Few MOOCs can be considered as contextualized to a learning population. Linking the geographical uniformity of enrolment to the topic of time may reveal some commonalities. After all, annual holiday periods, work weeks and time zones are identical across the country. Capturing such information in aggregate has become increasing possible through harnessing big data on social media, and in particular, social networking platforms.

Sina Weibo: China's most popular microblogging site
In China, social networking is extremely popular. The microblogging platform Sina Weibo (微博) in particular, enables users to engage on a range of issues with other users. The context is unfiltered, and subject only to periodic interference. No wonder there are over 600 million accounts, of which 30% are monthly users (Smith, 2015). Like Twitter, Sina Weibo limits a posting to 140 characters. Other common functionalities include initiating, re-posting, or responding to other users' postings.
Organizations, corporations and governments in China utilize Sina Weibo to interact with the wider usership. Newscasts often share with its viewers individuals' unsolicited postings reacting to varying social events. This adds a layer of authenticity and immediacy to reporting. Postings in aggregate also create a system of self-regulation. Inaccuracies, for example, are quickly weeded out from trending conversations. Repeat offenders are subject to scorn, and the permanence of the Internet leaves little room to erase an errant posting.
Because of the big data that is readily available, microblogging has presented opportunities to conduct varied research. One widely cited example was a study by Bollen, Mao and Zeng (2010).
They conducted a study that forecasted trends in the stock market by analyzing user behaviour on Twitter. They concluded that large volumes of tweets, as postings are called on Twitter, could be used to predict changes in the US stock market. Other studies have shown predictability in the transmission of infectious diseases (Sadikov, Medina, Leskoven & Garcia-Molina, 2011), a range of social events in China (Guan et. al, 2014). research was lagging behind in that the Twitter data has not received the considerable attention it merits. Overall, there is a lack of research using microblogging data to predict the trend of educational movements. The majority of research views Twitter as a learning tool promoting social interactions among students. Gao, Luo and Zhang (2012) conducted a critical analysis of research on microblogging in education, and illustrated that microblogging in education was mainly utilized as an educational tool to extend learning beyond the classrooms and blurs the line between formal and informal learning. As for MOOCs, microblogging sites (such as Twitter and Sina Weibo) act as an alternative to discussion forums and are used for a large number of MOOC courses. For example, the course "e-Learning and Digital Culture" offered by the University of Edinburgh used Twitter hashtags to create learning activities for their students. Van Treeck and Ebner (2013) also studied two consecutive MOOCs to explore the usefulness of Twitter for learning in massive communities.
In this paper, a time series analysis is provided over two periods: from 2010 to 2015 and 2013 to 2015. The purpose of the paper is to provide an analysis of data referencing MOOCs on Sina Weibo, a popular Chinese microblogging website. The aim is to provide insights that may be transferrable to learner behaviour in a MOOC. Another aim is to demonstrate the utility of data access through social media to inform theory or practice in the learning sciences.

Methods
In an attempt to study the public response to MOOCs in China, Sina Weibo data that made reference to MOOCs was analyzed. The microblogging site is the most popular of its kind in China. As noted in the previous section there are over 600 million registered accounts and approximately 30% of accounts are active monthly users who create a high volume of usergenerated content. As Sina Weibo has been used to gauge public perceptions about varying social phenomena it was deemed a valuable data source to conduct longitudinal data mining analyses about MOOCs.
To gather data, a posting was selected if it contained the keyword "MOOC" or "Muke" (慕课). For simplicity in this paper, the phrase MOOC postings includes either of these keywords. A web crawler was utilized to compile the postings. In total, 95,015 postings published by 62,074 users were retrieved on Sina Weibo from September, 2010 to January, 2015. The time series patterns of MOOC postings was analysed using R programming language. Four time frames were selected to analyse the volume of postings: by year, by month, by day of the week, and by the time of day.
It is important to note that the content of the postings do not necessarily reference an individual's experience of learning in a MOOC. Such content analysis is the basis for future research. Instead, the data reveals a set of time frames of MOOC postings. The results may provide insights for prospective designers of MOOCs on strategies to maximize engagement with learners. If, for example, there is a high volume (or low volume) of postings during a particular time of year, providers or designers of MOOCs may consider time-sensitive strategies in delivering a MOOC.
The assumption is that purposeful time-sensitive interventions may impact learner engagement, and ultimately enhance learning and minimize attrition.  Table 1).   In Table 2, the distribution of number of postings by number of users is displayed. The data demonstrates the volume of MOOC postings by groupings of users

! 152
Another notable finding is the lull in MOOC postings contributed on Saturdays. Observing patterns of behaviour in a MOOC during a given week may reveal similar findings. In any event, tracking user behaviours in a given week may reveal particular patterns that impact individuals' learning experience in a MOOC.
! Figure 3. The weekly distribution of postings that include the term MOOC on Sina Weibo.

Contributing postings after work
Like the patterns observed during the year and during the week, MOOC postings contributed during the day revealed that certain time intervals were more popular than others (See Figure 4).
Users contributed postings most frequently in the later evening followed by mid-afternoon and late morning. The comparatively high volume of MOOC postings contributed during after-work hours is not surprising as this is leisure time for many individuals. It also suggests that a significant number of users on Sina Weibo who contributed MOOC postings were working adults.
Another popular time to contribute MOOC postings, however, occurs during the workday. It may be inferred that there are a significant number of users who are not working adults, but those who work unconventional hours or have greater flexibility in the day such as students, retirees, or stayat-home caregivers. After all, the ease of submitting a posting to a microblogging site is highly convenient given the ubiquity of smart phones. One last notable finding was that the volume of

MOOCs for the Chinese Society
Worldwide interest in MOOCs was stimulated by Stanford's artificial intelligence course, which attracted over 160,000 students from more than 190 countries in the fall of 2011 (Barnes, 2013).
Since, several thousand MOOCs have been designed and delivered. Many remain active and new courses and new users are joining the MOOC space daily.
The appeal to MOOCs seems rooted in a cost-effective alternative to conventional higher education. MOOCs are effectively free, and many top universities are partaking in the design and delivery of these large online courses. The appeal is multifaceted. Optics suggest the institution designing and delivering the MOOC is trendy, pedagogically sound in regards to online learning, and engaged in some form of community service. It has also proven to be an excellent mechanism to promote a professor, a faculty, or, ideally, an entire institution.

! 154
In China, there are additional forces at play. There is an immense desire for quality higher education, particularly since the playing field is so uneven. A handful of institutions are recipients of enormous government funding, relegating a large majority of institutions to the margins. The growing draw to MOOCs seems to be linked to the desire to enrol in the country's top universities which are highly selective in their campus-based programs. Like the draw to Stanford or Harvard, an underlying assumption is that China's elite institutions provide quality instruction. If these institutions do not offer some viable credential that is recognized by government and by industry, incentives to enrol in MOOCs will wither. Other potential pathways are to utilize MOOCs to improve teaching domestically, much like the Chinese national top level course project (jingpinkecheng, "精品课程") aimed to achieve starting in 2003 (Haklev, 2010).
The other big potential for using MOOCs in China is to support a lifelong learning society, a primary objective identified nearly 20 years ago in the Action scheme for invigorating education towards the 21st century (CERNET, 1998 between platforms, the premise of connecting innumerable users online is common, and therefore transferrable to varying learning contexts that use MOOCs (Zhang, 2015). Assuming there is a link between the learning experience in MOOCs and MOOC postings on social media, how MOOC designers embed the use of sites like Sina Weibo as part of the learning experience will impact the volume of MOOC postings. Learners too may alternate between social media platforms to engage with other learners about their engagement in a MOOC(s). In the current setting in China, assessing user trends on MOOCs through social networking mechanisms will benefit from mining data from multiple sources.

MOOC Activity during the week
In a narrower sense, the activity patterns found in this study correlate with individuals' weekly routine. Yuan (2006) found that online students enrolled in distance education programs at East China Normal University tended to visit their course platform most frequently from Monday to Thursday (i.e., 600 student visits per day), with a noticeable drop on the weekend (i.e., 400 student visits per day).
The daily distribution of postings on MOOCs illustrates that, apart from the normal working hours during the daytime, a large number of users contribute MOOC postings in the later part of the evening. Findings also reveal that users are contributing MOOC postings at other intervals during the day. Two possibilities stem from this finding. One, more and more individuals are contributing postings using mobile phones, which are readily available and highly functional for microblogging. The other finding is that many users contributing MOOC postings are not working adults, but those who have greater flexibility in their work schedules, such as students. Clearly, further analysis is needed to discern the demographics of users in this study, as well as the rationale to rely on social networking to share information on MOOCs.
Most of all, content analysis is required to ascertain what users are talking about in regards to MOOCs, a topic for subsequent analysis with the data set used in this paper.

Conclusions
In this study, 95,015 MOOC tweets published by 62,074 users on the Sina Weibo platform between 2010 and 2015 were analysed to assess the public response to "the MOOC movement" in Monday, a finding that seems unusual considering this is the beginning of the work week and therefore a busy day for working adults. This trend gradually waned to Saturday before picking up again on Sunday. Finally, there appeared to be three distinct times during the day when the volume of postings was high. Evenings proved to be the most popular time frame when users were contributing postings on MOOCs.
Together, the analyses reveal potentially useful information for those who design and deliver

4.
Few tweets (n=83) were published between 2010 and 2011. To avoid noise in the data, these postings are excluded from analysis in this section.