Introduction

Massive Open Online Courses (MOOCs) are online courses aimed at broad participation and open access to everyone, regardless of their educational background or geographic location. As a result, MOOCs draw large numbers of students from diverse backgrounds, and then ask them to learn in the same online classroom. Adapting instruction to students’ cultural preferences has been found to improve learning gains, even in face-to-face classroom settings where students speak a common native language, live in the same geographic area, and learn in small groups (see e.g., Finkelstein et al. 2013). MOOCs are far more diverse than the classrooms in these studies. In the same online classroom, there are students who speak languages different from the language of instruction, students who have different cultural beliefs, students in different geographic locations, students with different educational backgrounds, and students at different levels of ability (Kasunic et al. 2016). Adapting instruction to student needs provides clear opportunities for improvement in enrollment, participation, and learning outcomes. However, existing adaptive systems, which have typically been intended for local use, may not yet be ready to support MOOCs. To understand how to integrate these systems effectively into MOOCs, and how to design new systems for adaptive instruction in MOOCs, insight into learners’ diverse needs and context is necessary. In this paper, we explore the issue of linguistic diversity in MOOCs, with a focus on English Language Learners (ELLs). We study the motivations for ELL students to take English-language MOOCs, and discuss the implications for adaptive design in MOOCs.

Although MOOCs have been widely adopted by the world’s most educated and affluent learners (Christensen and Alcorn 2014), the adoption of MOOCs remains limited for those who lack adequate infrastructure, digital literacy, and English language proficiency (Aborisade 2013; Liyanagunawardena et al. 2014; Nyoni 2013). While the requirements for infrastructure and digital literacy are fundamental to the MOOC format, English language skills are not, in theory, a requirement for participation. In practice, however, a review of all MOOCs listed on the aggregator site mooc-list.com as of May 2017 reveals that a disproportionate number of courses are offered exclusively in English.Footnote 1 Nearly 75% of MOOC content providers operate in English (Czerniewicz and Naidoo 2013), while only a minority of the global population speaks English as a first language (Noack & Gamio, 2015). This disparity between worldwide native speaking populations and the language of MOOC content is well known, and there are ongoing efforts to make MOOCs more accessible to audiences for whom English is not a native language. Khan Academy is currently working on translating much of its content to Spanish (Dolan 2013). Both Coursera and edX plan to extend MOOCs to Chinese students by providing Chinese language versions of course search portals, course synopses, platform orientations, discussion forums, and student testimonials (Coursera 2013).

These attempts can certainly increase MOOC access for students who have no knowledge of English. However, there is also a large global population of ELLs, or English Language Learners, who have some proficiency in English as an additional language – 603 million in one recent estimate (Hammarström 2015). In practice, ELLs are either treated as English native speakers (by the presentation of all content in English), or as non-English speakers (by translating MOOC content to their native language). We argue that they are in fact a third group that is large, that has unique needs, and that uses MOOCs in different ways than either English-native or non-English-speaking learners. Existing work has investigated the motivations of students from particular regions (Boga and McGreal 2014; Cutrell et al. 2013; Nyoni 2013), and general MOOC student motivations (Dillahunt et al. 2016; Kizilcec and Schneider 2015; Schneider and Kizilcec 2014; Zheng et al. 2015). We seek to understand the range of motivations for ELL students in MOOCs across many geographic regions, which may be different from the motivations of MOOC users in general, in order to design systems that support them as a distinct user group.

Because student motivations are difficult to extract from interaction logs, we interviewed 12 ELL online students about their needs and motivations for taking MOOCs and other online courses. Based on the findings from our interviews, we developed and deployed a large-scale survey in a MOOC targeted specifically to ELL students. We look at the complexities of these students’ motivations and experiences to give us insights into designing the kinds of interventions that may be effective at supporting ELLs who enroll in MOOCs, and generalize these insights with our survey data. Our research makes the following unique contributions to the MOOC research community:

  • We provide rich insights into the motivations of ELLs who are taking MOOCs and online courses, and how these motivations support long-term goals that are highly socialized.

  • We show that ELL students use MOOCs as a platform to prepare themselves for a future of economic, social, and geographic mobility, implying that MOOCs serve as more than a simple classroom replacement.

  • We present evidence that ELLs do want to interact with others in MOOCs, contrary to findings from recent studies, and discuss how AIED systems that support student interaction in MOOCs can be better designed to meet their needs.

  • We show that while translating MOOC content into students’ native languages does increase access for students with no English ability, it is equally as important to develop adaptive language support in English MOOCs for ELLs who believe that learning in English will help them achieve their goals.

  • We discuss the implications of these insights and propose design and implementation changes to large-scale AIED systems to support the social and economic goals of ELL students.

Related Work

ELL Participation in MOOCs

While there is no definitive estimate of non-native English speaking populations in MOOCs, existing studies suggest the number is substantial. 34% of the 13,887 students who enrolled in and completed a English-language Psychology MOOC were categorized as ELL students (Uchidiuno et al. 2016a, b). Preliminary analysis of data from a 2015 Statistical Thermodynamics course shows that approximately 32% of the 2971 students have a language other than English set as their preferred browser language. In 2014, Guo et al. (Guo and Reinecke 2014a) analyzed data from four edX MOOCs comprising 140,546 students from 196 countries. The four courses (6.00× – Intro to Computer programming, PH207x – Statistics for Public Health, CS188.1× – Artificial Intelligence, and 3.091× – Solid State Chemistry) show that the countries with the most certificate-earning students include Russia, Spain, and India. These three countries alone comprise 21–28% of the student population; this is similar to the combined representation of the U.S. and U.K., which is also 21–28% of the student population. In the Conversational English MOOC course discussed in this paper, over 150,000 students have participated in just over a year. These numbers show that, even with current designs, a significant number of MOOC enrollees are likely to be proficient in a language other than or in addition to English.

Even with the significant proportions of ELLs in MOOCs, research studies find that they are less likely than native English speakers to interact with other students online. Kulkarni et al. (Kulkarni et al. 2016) developed and deployed a video interaction MOOC platform, Talkabout, and showed that non-native English speakers were less likely to speak in discussion sessions than native English speakers were. Kizilcec and Schneider (2015) reported findings that “learners with the intention of improving their English skills were less likely to heavily engage on the discussion forum” (p.16). These authors clearly identify a problem that needs to be addressed, but do not provide explanations. Were students uncomfortable participating due to the language barrier? Were these students less motivated to interact? Or was the platform not effectively designed to meet their social interaction needs? In this paper, we demonstrate that ELLs are highly motivated to participate in MOOCs, but social interaction systems must address their unique concerns and motivations.

Existing Tools and Strategies to Increase ELL Access in MOOCs

A majority of the efforts to solve the issue of access specifically for ELLs has been dedicated to translating MOOC content into other languages (Dolan 2013). Some MOOC providers have developed courses entirely in other languages, such as Ewant (Chinese), Veduca (Portuguese), and France Universite Numerique (French), while many others such as OpenUpEd, Coursera, edX, and openHPI support multiple languages to accommodate a variety of students. Other efforts (in MOOCs such as CourseSites, Open Learning Global, and NovoEd) have built tools that allow people to build custom MOOC content (edX, 2015), which allows for the development of MOOCs from scratch to fit the needs of a local population (Cutrell et al. 2013). However, there is limited data available on the effectiveness of such tools to increase MOOC access for non-native English speakers. While translation can certainly make course content available to ELL students, there is no data that shows how this intervention affects either students’ personal long-term goals, or the other stakeholders that are involved in creating and deploying MOOCs. Finally, we note that some tools designed for MOOCs without considering the needs of ELL students, such as automated essay scoring, may disadvantage ELLs (Reilly et al., 2016). We argue that adapting MOOCs to meet the needs of linguistically diverse learners involves much more than providing the course in a language the students understand.

Students’ Motivation for Participating in MOOCs

There has been increasing interest in the last few years in understanding students’ motivation for participating in MOOCs, allowing researchers to better meet students’ actual needs (Dillahunt et al. 2016; Kizilcec and Schneider 2015; Uchidiuno et al. 2016a, b; Zheng et al. 2016). Kizilcec and Schneider (2015) analyzed free text survey responses from three Stanford University MOOCs, and developed categories for MOOC learner motivations. These categories were used to create the Online Learning Enrollment Intentions (OLEI) scale, which has been included in all Stanford University MOOC surveys since September 2013. So far, the researchers have included the OLEI scale in 14 MOOCs at Stanford, with over 71,000 respondents (Kizilcec and Schneider 2015). This instrument is useful for high-level self-report data about motivations. For example, their results suggest that a mean of 28% of all MOOC students across a variety of courses take MOOCs to improve their English skills.

Other research studies propose that it is possible to extract MOOC learner motivation by analyzing survey and interview data, or a combination of their click stream browsing data, and their demographic information such as their IP addresses to determine their physical location (Dillahunt et al. 2016; Guo and Reinecke 2014b; Zheng et al. 2015, 2016). Taken together, these studies report general motivations – complementing physical courses, gaining more knowledge for their profession, impressing an employer, improving college applications, satisfying their curiosity, and connecting with other people. Such studies provide broad categories of MOOC learning motivations, without providing an understanding of what those broad motivations mean in the context of non-native English speakers’ experiences, or their choice of MOOCs as the learning context. Finally, the metrics captured by existing surveys fail to capture the context and nuance of learner motivations, making them difficult to use for design. Perhaps as a consequence, the design recommendations proposed by current survey-based research do not include design proposals for adaptive systems.

Given that these surveys do not differentiate between native English speakers and ELLs, it is unknown to what extent these motivations represent ELL students specifically. Even when studies collect location data, it is not safe to assume that location predicts language proficiency; given that humans are increasingly globally mobile, a person’s physical location while logging into a MOOC may not be fully predictive of their language abilities (Uchidiuno et al. 2016a). We argue that such metrics are limited by their nature in providing actionable steps to design MOOCs to support non-native English speakers’ needs, without being informed by qualitative inquiry. We augment these research studies by providing rich qualitative insights to the motivations of non-native English speakers who are self-directed online learners, to reveal their explicitly stated motivations, as well as those implicit ones that can only be inferred from the context of their holistic learning experience. We substantiate our findings by deploying them in survey form in a MOOC targeted specifically at ELL students. Examining the motivations of non-native English speakers in MOOCs can inform research on adaptive designs for English-language MOOCs to meet ELL students’ long term goals in ways that account for the many different (and often invisible) stakeholders involved.

Challenges and Benefits of MOOCs for ELL Students

For a non-native English speaker, learning a complex subject area (e.g. astronomy) in an English-language MOOC can be especially challenging. Learning domain content in one’s non-native language has been shown to increase learners’ cognitive load – as if “taking a computer programming course where every exchange had to be written in computer code” (Sokolik 2014, p.18). The challenges of learning a complex subject area in English are not unique to MOOCs, but MOOCs provide unique benefits that could support ELL students. For example, MOOCs can achieve the same outcomes as a regular classroom at lower cost to students, (e.g. (Bowen et al. 2014; Collins 2013). For some underserved populations, MOOCs are the only way to afford a rigorous, college-level course (Schmid et al. 2015). The high quality and free cost of MOOCs are critical characteristics that allow non-traditional students to overcome geographic and economic barriers (Russell et al. 2013) and pursue their own learning objectives (Armstrong 2012). Students can learn content at their own pace; this reduces ELL students’ classroom stress which can arise from the instructor continuously challenging them to speak a language in which they are not proficient in a social setting (Kuksenok et al. 2013).

The use of digital video and multimedia-based materials to facilitate instruction in foreign language learning environments has been shown to increase student motivation, allow greater engagement with materials, provide feedback in the absence of an instructor, and enrich the teaching of cultural aspects (Chen and Oller 2005; Etienne and Vanbaelen 2006; Kuksenok et al. 2013; Tschirner 2001). Given these language challenges, one might expect that ELL students would simply utilize platforms that deliver content in their native language. However, this is not the case – as noted above, studies have shown significant participation in English-only MOOCs from countries where English is not an official language (Christensen et al. 2013; Guo and Reinecke 2014b; Snyder and Writer 2013). We aim to uncover the motivations of ELL students who willingly go through this additional hurdle by participating in English MOOCs, and to understand ways in which these learners are different from their course peers who are native English speakers.

Methods

We conducted semi-structured interviews with 12 ELL students (names changed for confidentiality). Recruitment targeted subjects with experience with self-directed online learning, specifically MOOCs, self-paced online courses, or related online courses. English-language advertisements were sent to Facebook groups related to a wide variety of online courses as well as community learning resources near the authors’ university.

Respondents were screened to ensure that they were at least 18 years of age and that they had previously participated in online learning courses. All respondents who were born and raised outside the United States, and whose first language was not English, were included in this study (Table 2). Respondents who did not meet these criteria were interviewed by the same research team for a separate study.

In order to study how ELLs engage with English-language MOOCs, we chose to recruit and interview in English. Our subjects had a range of English-language skills, but all were able to carry on a conversation and read simple text, as well as to give informed consent in English. Given that baseline English skills are required to discover and participate in an English-language MOOC, we used this as an inclusion criterion. A designated interviewer conducted semi-structured interviews in English, either in person or over Skype. Interviews lasted 45–60 min and included a brief demographic questionnaire. Interview topics included an open-ended description of the subject’s educational history; their motivations for participating in MOOCs or other online courses; stories about their experiences while participating in online learning; their beliefs about the value of online learning; and the quality of their social and emotional learning experiences. Sample questions included “How did you make the decision to sign up for [a particular course described by the learner]?” and “Tell me one thing you hoped to get out of the course when you first started.” Subjects were also asked about their life context as it related to learning, such as who in their family supported their learning experiences. Finally, interviewers followed up on learning-relevant topics introduced by the subject. All interviews were audio-recorded, transcribed by the research team, and entered into Atlas.ti for coding. Two researchers used Atlas.ti to identify and mark sections of the transcripts that related to any motivations for participating in online learning. At this stage, an inclusive approach was used in order to allow for the next stage of inductive, thematic analysis; therefore, any segment coded by either researcher was marked for further analysis by the broader research team.

Once all segments related to motivation were extracted, the team moved to an inductive approach (Braun & Clark, 2006). Three researchers reviewed the coded segments and identified emergent themes through an iterative process. First, the researchers sorted the segments into categories based on perceived similarity. Seventeen categories emerged. By examining these categories in a saturation grid, the team determined that data saturation had been reached (Brod et al., 2009). The researchers then identified the core concept of each category, and examined which participants had raised each point to look for prevalence and interrelatedness of themes (see Table 4 below). Through this process, the seventeen categories were reduced to six common themes, including “career advancement through English language skills” and “managing interactions in daily life.” These six themes, along with supporting data, were presented to colleagues outside the research team for feedback. The research team then incorporated these external insights to develop the three major motivations treated in this paper, as well as to shift from a preference-based (emphasizing what learners like or dislike about the course) to a goal-based (how learners use the course to satisfy their needs) approach.

After the interviews were completed and the data analyzed, we developed a survey instrument derived from this analysis, targeted at ELL students who take online courses in English. While the data gathered from our interview subjects was rich, our interviews could not tell us whether these insights would generalize. In order to discover how common the motivations of our interviewees were among the larger ELL population, we created survey questions based on specific elements of the open-ended responses provided by our interviewees. We selected elements that reflected each of the themes, which were mentioned by multiple interview subjects. Flesch-Kincaid readability scores, along with review by ELL individuals from a range of cultures, were used to ensure comprehensibility (Klare, 1974–1975). To tie our work to previous findings (Kizilcec and Schneider 2015), participants also completed the OLEI scale survey questions in Table 1. To most effectively compare our survey findings to those of the OLEI scale, all new survey questions (except those gathering demographic data) were presented the same way as the OLEI, using a 2 point Likert scale of “Applies to me” or “Does not apply to me.” Our aim was to highlight the similarities and differences between the motivations of general MOOC populations (which includes some ELLs), and ELL-only MOOC populations. The differences between these two groups provided our team with baseline differences in their motivation, and served as a lens for us to analyze both our interview and survey results (Table 2).

Table 1 OLEI Scale included in Stanford University MOOCs (Kizilcec and Schneider 2015)
Full size table
Table 2 ESL interview participants and their experience with online courses
Full size table
Table 3 Student demographic information
Full size table
Table 4 Co-occurrence of interview participants’ long-term goals for taking online courses
Full size table
Table 5 Survey questions and the percentage of students with “Applies to me” responses (N = 4651)
Full size table
Table 6 Emerging themes from Principal Component Analysis of survey questions (N = 3848 – See Table 5 for Question Text)
Full size table

To maximize the number of ELL respondents, we deployed the survey in a “Conversational English Skills” MOOC, an edX course targeted at ELL students. The course is aimed at developing conversational English skills, teaching students key words and expressions, and teaching students to listen to “dialogues and group discussions to better understand spoken English and cultural norms” (Conversational English Skills 2016). The course is comprised of eight units and a final exam, including listening comprehension videos, quizzes, and language use exercises for vocabulary and grammar practice. 146,304 students from 195 countries enrolled during a year of offering this self-paced course, with India (11.2% of total users), United States (6.8%) and Egypt (6.5%) representing the top three countries by enrollment.

The population of interest for this study was ELL learners who had previously taken a MOOC in English. To screen out any participants who were native English speakers, our survey used questions based on findings from previous research (“NHLRC: Heritage Language Survey Report,” 2014.; Reinecke & Bernstein, 2008) to ask students about their language abilities. Our survey drew 20,084 respondents from 166 countries; the top 10 countries represented include India (13.63%), Brazil (9.48%), Colombia (8.04%), Mexico (6.76%), Egypt (5.76%), China (3.5%), Spain (3.15%), Vietnam (2.21%), Pakistan (2.18%), and Russia (2.17%). Among the total survey respondents, 1.64% stated that English was their native language, and 1.20% indicated that English was the language they spoke the best; these participants were removed from the sample. 35% of the students indicated that they had spent some time in a region where English is spoken often. Tables 3 and 4 shows student demographic information for survey participants including age, gender, and education level; for other descriptive results and a list of survey questions, see Table 5 in the Results section. Our survey results below are reported using the 4651 (30% of the total) who had previously participated in a MOOC deployed in the English language, and who did not report native-level or English-best language abilities.

Results

Although the survey was deployed after the interviews, we begin our findings with the description of how Kizilcec and Schneider’s (2015) findings differ from ours in order to understand how ELL students might differ from what is previously known. We examined the differences between our ELL participants’ responses to the OLEI scale, and compare it to the general MOOC population as reported by Kizilcec and Schneider (2015). Figure 1 shows the results comparing both populations. While we cannot statistically compare the findings across the two studies due to the limitations of the data reported in Kizilcec and Schneider (2015), a visual inspection demonstrates that there are observable differences between the groups. This data likely underreports the actual differences, as 30–35% of students in the “general” MOOC population are typically ELLs, but their data is not reported separately in Kizilcec and Schneider (2015).

Fig. 1
figure 1

Results of OLEI Survey from ELL focused MOOC vs general population MOOCs as reported in K&S

Full size image

The biggest observed difference is in the stated need to improve their English. This difference is not unexpected, as a course on the English language likely selects for learners who are motivated to learn English. Other notable differences include the motivation to meet new people in MOOCs (38.6% ELL vs. 24% general); the desire to change careers (56.74% ELL vs. 36% general); and the use of MOOCs to serve their job (74.14% vs. 58%) and academic interests (51.2% vs. 37%). These findings provide evidence that ELL learners are different in their MOOC motivations from non-ELLs, and therefore deserve further study. These differences also suggest that there may be other, unreported motivational differences between ELLs in MOOCs and “general” MOOC populations. Surveys can only capture what they are designed to measure. To understand what might be omitted by the OLEI scale, we go back to the affected population and study them directly. Additionally, interview findings can give deeper context and meaning to the differences observed between ELL learners and “general” MOOC populations.

We now describe our findings from the interviews, interspersing survey findings as a check on the generalizability of the insights. As described earlier in this paper, the research team probed for motivations for learning using a range of prompts, including asking about motivation explicitly, having learners describe powerful experiences in online courses, and asking participants to connect their online learning experiences to their life context. Through the iterative thematic analysis process described above, three major motivations emerged: economic mobility, social mobility, and geographic mobility. These motivations were understood by learners as related to long-term personal and career aspirations rather than to the specific course they were taking. We therefore define them as follows:

  • Economic Mobility: This is the ability for individuals to move up or down the income distribution (U.S. Economic Mobility 2015). For the purposes of our research, any motivations aimed at gaining knowledge, meeting people, or moving locations specifically for the possibility of increasing their income are categorized under this goal.

  • Social Mobility: This is the movement of individuals or groups within a social hierarchy (Social mobility. 2017). For the purposes of our research, any motivations aimed at being accepted in a particular social circle are categorized under this goal.

  • Geographic Mobility: This is defined as a measure of how populations move over time (Geographic mobility 2016). For the purposes of our research, any motivations aimed at increasing the possibility of moving to a different geographic location are categorized under this goal.

Table 4 shows an overview of our interview participants, and the co-occurrence of their long-term motivations for enrolling in MOOCs and other online courses.

Economic Mobility

As shown in Table 4, all of our interview participants were highly motivated by the goal of increasing their income. All our participants took MOOCs and other online courses to increase their earning potential using one or more of three different strategies: mastering course content, obtaining certificates, and improving their English language skills. This finding is further strengthened by our survey results, as 90% of the survey respondents with English MOOC prior experience stated that they take MOOC courses to learn skills that help their career (Table 5: Q3). While the OLEI survey suggests that at least some native English speakers take MOOCs for economic reasons, we found that students’ ELL status affected the strategies they used to accomplish economic mobility through MOOCs.

ELL status was least directly relevant for students who were primarily focused on course content. For example, Andrej discussed how his high school calculus was “really easy”, and that he enrolled in MOOCs because he needed to know something “harder” in order to be admitted into college and earn a degree. On a more professional level, Ali discussed how taking MOOCs allowed him to be given priority in his job as a result of his newly acquired skills. He felt that the knowledge he learned from his university just taught him the theory, and expected him to learn the practical skills on his own – and MOOCs allowed him to obtain those practical skills. He never enrolled in courses with the paid certificate options, just the honor certificates, because the opportunity to practice his skills was sufficient to help him do his job. Pooja’s need to practice course content was even more specific. She was studying for her United States medical board exams; improving her knowledge of the exam materials, and practicing taking the exam in English, would help her do well on the test. In comparison to the doors opened by strong exam performance, a MOOC certificate held little value to Pooja.

In direct contrast to those who needed to master the course material, some participants focused on obtaining a course certificate. Among the survey participants who had taken an English MOOC before, 49.8% reported that it was important to them that their certificates stated that the material was completed in English (Table 5: Q12), and 44.3% stated that they already had the knowledge, but just wanted the additional certificate (Table 5: Q13). Given that some of our participants already had college degrees and worked as professionals in those fields, it was clear that they already had mastered the MOOC content. However, English-language certifications allowed them to concretely demonstrate their multilingual mastery of the domain, and furthermore functions as an implicit certification that the learner can collaborate with English-speaking co-workers about complex and specialized topics. These students felt that they could already communicate about work topics in English; they simply needed evidence that employers would find persuasive. Participants such as José, Carmen, and Filip stated that they were only willing to take MOOCs that gave them certificates at the end, even if they would have to pay extra for the certificates. Some jobs they applied for explicitly required that they show English-language certificates to prove that they have the skills needed to do the work. However, they also sensed that the credential would implicitly help them even if the potential employer did not require it. For example, Filip noted that he added his certifications to his LinkedIn profile to improve his marketability:

“I use certificates to show to my potential employers what I am doing. Obviously, you can’t compare a Coursera certificate to my degree, but I think it is good to show people what I am doing. That it’s not ‘Oh, I am just finishing my degree and doing nothing’. It is showing that I am still interested in the subject – it is showing my passion.” (Filip, M, 30’s, Polish)

Rather than learning course content, or obtaining certificates, some ELL students partake in MOOCs and online courses as a strategy to become better English speakers in ways they believed would be beneficial for their careers. The interview participants, especially those who lived in the US, stated that they took MOOCs to improve their work-related English language abilities, as they needed mastery of the language to communicate and obtain the jobs they desire. Participants took two different approaches to addressing this problem. Some participants chose to take English language courses, where the course content was focused on teaching English. Other participants improved their English language skills by taking MOOCs that were not English language courses. Instead, they chose to study other topics of interest to them, but to do so in English. They were able to use the courses to simultaneously improve their subject area knowledge as well as their English language skills. Even though learning content in another language increases cognitive load, these students chose to seek out this experience.

“I feel frustrated because I sometimes don't understand the English, but… I prefer it in English. Why? (laughing) I need to improve my English. I don't care if I don't understand, or if I don't answer correctly the questions. I have to listen and read in English” (Maria, F, 30s, Colombian)

Social Mobility

Much of the research on MOOCs treats social motivations as students seeking an opportunity to connect with others around similar interests. According to our ELL participants’ responses on the OLEI survey, ELLs are motivated to meet new people in MOOCs (38%), but not necessarily to take it with people that they have existing relationships with (14%). This result is surprising given recent research findings that ELLs rarely participate in discussion forums (Kizilcec and Schneider 2015; Kulkarni et al. 2016). Our findings suggest that social dimensions relevant to ELLs are being omitted from existing studies, which can explain this contradiction. Specifically, we found differences between our groups based on the type of social interaction available. Both interview and survey participants expressed interest in connecting with people around social and cultural participation, but interview participants were not interested in engaging with their peers around course content.

Among our survey respondents, 60% indicated they interacted with students to learn about a new culture (Table 5: Q15). This strategy was best exemplified by Alina. During her interview, she stated that she exercises every day to stay beautiful, loves to speak eloquently, and chooses to drink specialty tea from an expensive Russian brand. She does not use the word ‘older’ to refer to herself, but instead ‘better’. She decided when she came to the US that she would not achieve career success because she does not speak the same way and does not understand the context of conversations. While she was not directly motivated by economic mobility, she felt that she belonged to a certain social circle that her online courses would help her become a part of. She takes a GED class to improve her English language, but does not interact with the students as they mostly dropped out of school. However, she talks to the teacher a lot because “he is so smart”, and brings up topics that makes her “even feel high”. She stated that she preferred to interact with educated people, because it “prevents you from being shallow and narrow minded”. Alina didn’t want to meet just anyone; she had a specific class of people in mind to interact with while taking these courses:

“When in Rome, do as Romans do [laugh]…. It’s why I don’t communicate some people like from Russia. Because they still speaking like, you know [in exaggerated Russian accent] ‘Hi, my name, like, tomorrow, good morning!’ Oh my gosh, you know? I don’t even understand why just they didn’t learn, you know, good English?” (add participant citation in the same format as the previous quotes)

For other students like Diego, connecting with people is a means of networking, both to form deep friendships but also as a way to meet people that can potentially vouch for him in the future:

“In an online course, it’s very difficult to make close friends. You can talk with them, but they are not your friends. They only talk, maybe, once in a year. And then, they forgot you. But when you meet in a university, in a place, you meet something with the people— it’s different. They become your friends forever. They not forget you. They know you more. They know whether you are strange. What are your, in what you are good at. So if there is a job, they know, ah, you are good at this. So, why not, you can work with me? Something like that.” (add participant citation in the same format as the previous quotes)

Despite Diego’s focus on networking, interview subjects reported that they did not want to discuss course content with their peers, and that they largely did not participate in MOOC discussion forums. For some subjects, this is because they were primarily concerned with how their learning would affect specific pre-existing relationships. For Maria, for example, a job is just not a job. It is directly tied to her family’s financial power, but more importantly to her role within the family, and to her sense of self-worth. When asked the reason why she wanted to learn English, she responded:

“Because I want to do a master. I want to take a TOEFL test. I need to get a high score. My English is not enough for a high score in the TOEFL test…. I like to feel— you know, my husband right is studying his engineer. I am engineer too. And being mom is a wonderful job. But you have to combine mom and my career…. It’s very better to have two incomes, not just one…. And to feel important. It’s my feelings. It’s about me, about I. It’s like, to teach my kids that mom is not just a mom. Is a professional, too. Like their dad.”. (add participant citation in the same format as the previous quotes)

For other students, they felt that MOOC forums and course content discussions lacked a sense of intimacy, humanity, and trust. For example, Diego contrasted the ability to make close friendships in university with the lack of true connection in MOOCs. Pooja agreed:

“I have some inhibition. Like, you know, talking to a stranger? It’s not because I’m scared of them. It’s because they’re strangers, at the end of the day. But if it’s a real course, I would be able to more interact with the person.” (Pooja, F, 20s, Indian)

Interestingly, Pooja and Diego do not link their reluctance to engage about course content to their English language skills. Instead, they cite expectations for what online relationships are like – expectations that may be rooted in cultural factors such as a heightened sense of personal privacy, and hesitation to engage with complete strangers. Lily et al. (Al Lily et al. 2016) show that people’s acceptance of educational technology is highly dependent on cultural norms, such as their needs for online privacy, and the gender preferences. Such factors may not only affect people’s acceptance of technology, but also their ease in participating in forums that may violate some of their cultural preferences. Data on ELLs and Facebook suggests that ELLs may prefer platforms where they can understand their co-learners as full human beings (Kasunic et al. 2016). This finding aligns with Pooja and Diego’s descriptions of what online learning relationships are like, as well as with Maria’s focus on how the MOOC can transform some of her most intimate relationships for the better.

Finally, we note that our survey participants (66%) were far more interested in discussing course content than our interview participants (Table 5: Q16). We believe this may be an artifact of the English language course content, which may produce discussions similar to the “cultural” discussions that both groups value; however, this question requires further investigation.

Geographic Mobility

Finally, some ELL students participate in MOOCs as a strategy to learn the content and the culture of places to which they hoped to migrate someday, whether in search of economic betterment or otherwise. For example, 63% (Table 5: Q8) of our survey respondents stated that they take MOOCs in the English language to support their search for a job in an English-speaking country. The MOOC served as an opportunity to engage with course content, but not for a specific test or milestone; rather, students engaged with course content because they thought it was necessary in order to do the same in an English-speaking country. For example, José hoped the MOOCs he was taking would one day give him the opportunity to physically take courses in the US and in Europe. Pooja discussed how her medical license studying forum is filled with people from every part of the world, who hope to one day become a medical doctor in the US. Some of these participants, like Carmen, felt that there were better opportunities for her in English-speaking countries, and would want to one day study in America for the chance of growing her current career. Other participants, like Andrej, had career aspirations that they thought were difficult to obtain in their home countries. Andrej had taken the most MOOCs out of all our participants, and originally stated that he took them because he loved to learn. When probed further, he revealed that he wanted to become a physician (which he was formally studying), but also a physicist, and a researcher. He wanted to know and study everything possible about the human body, and he felt that such a multi-faceted career was impossible in his home country.

This goal also manifested in ELL students’ choice to learn additional languages (neither English nor their native language) to improve their geographic mobility. Pooja was not only studying for her medical board exams, but was also studying Spanish using an online course; she hoped to move to Texas one day, and serve both English and Spanish speaking patients. Minjun, on the other hand, was from an affluent family. He was taught by his businessman father that both China and the Arab world were going to be powerful in the future, so he felt it was necessary to learn both Chinese and Arabic. These students were motivated by both economic and geographic mobility; they thought that the ability to speak other languages made them more economically marketable in a particular geographic area.

Some ELL students participate in online courses as a strategy to help them prepare for and manage daily life in English. José, already living in the U.S., described incidents such as being asked questions by passers-by and being unable to respond, or needing to take his wife with him to get a haircut because he didn’t know how to communicate what he wanted to the barber. Alina also discussed how difficult it was to understand people – for example, not understanding what was funny about a joke someone told her, or feeling left out when everyone else around her got a joke and laughed. She also described her concerns with becoming a US citizen and being part of American society, when she cannot even navigate the language. For these participants, taking online courses mostly involved taking English classes that would not only teach them the formal language, but allow them to understand the cultural context of the language in their new homes, as well as interact with speakers who were learning the language and culture just like them.

Participants who described this motivation expressed a sense of fear, shame, or isolation related to their English language skills. For example, José noted:

“Yes, I feel comfortable when all the people in the room is in the same situation. So, it’s more relaxing. Because you don’t feel the pressure of being 100% correct.”

As noted in the previous section, participants who were reluctant to communicate about course content did not attribute this reluctance to their limited English skills, but rather to a lack of social connection with peers. Here, however, language skills themselves are experienced as a barrier. José describes the “pressure” to speak correctly, while Alina describes feeling left out when she cannot follow. The tension between the desire to learn English, and the discomfort of feeling excluded in English-language contexts, is uniquely important for these participants.

Co-Occurrence of Survey Participant Motivations

Finally, we ran a correlation matrix and a principal component analysis (PCA) on all the survey questions to explore the co-occurrence of concepts related to economic, social, and geographic mobility among these learners (Jolliffe, 2002). Table 5 shows the survey questions that ELL students responded to, as well as the percentage of students that responded with “Applies to me” for each question.

The question clusters from the PCA and our subjective interpretation of the factors are shown in Table 6. We ran a varimax-rotated Principal Component Analysis to determine the survey questions that mapped to the same factors. Of the ELL respondents to our survey, 4651 met the criteria of having previously taken an English-language MOOC and being less than native-level speakers of English. We employed a listwise method of handling missing variables, where all observations with any missing variables were excluded completely from the analysis. After excluding the observations with missing values, we ran the analysis with 3948 respondents. We observed that all 30 questions were correlated at least 0.40 with other variables, which suggests reasonable factorability. We used 5% variance as a threshold for factor inclusion; one factor was therefore excluded. The six remaining factors predicted a cumulative total of 58% variance. To test for sampling adequacy, a Keiser-Meyer-Olkin test measured 0.883 (recommended value is 0.6 and above), and Bartlett’s test of sphericity was significant (ϰ 2(435) = 53097.91, P < 0.001). Furthermore, the communalities were all above 0.40, confirming that each question had significant variance with other questions. A cutoff point of 0.40 was used to determine the questions that were correlated.

As seen in Table 6, these results show subtle but important differences from the patterns reported by our interview subjects. Factors 1 and 2 align with our interviewees’ focus on improving English in professional and everyday situations, respectively. Similarly, economic mobility is an independent factor for both our interviewees and the survey respondents (see Table 4, above). However, social and geographic mobility showed both independence and overlap among our interview subjects, and both factors always co-occurred with economic mobility. For our survey respondents, social mobility still co-occurs with economic mobility, but geographic mobility only appeared when intertwined with both economic and social factors.

Design Implications and Discussion

This work provides valuable insights on some reasons why ELLs participate in MOOCs, and why the current approaches of MOOCs may or may not be meeting their needs. To increase access and achieve learner goals, there are several ways that MOOCs currently make content more accessible worldwide, including translating current courses to local languages, developing new courses locally, and designing systems that improve access to existing MOOCs. All of these approaches have their place in supporting global learners: translation can make content created by subject matter experts accessible to those who prefer their native language, while creation of new, localized content can give a voice to frequently disenfranchised sources of expertise. However, bringing global learners together with a common language has been shown to have significant benefits – when MOOC students are given the opportunity to interact, the more geographically diverse the students are, the higher their performance scores (Kulkarni et al. 2016). While we do not claim that the factors of economic, social, and geographic mobility provide an exhaustive list of all ELL students’ motivations for participating in MOOCs, these three factors can guide the development of adaptive systems that help achieve these beneficial outcomes.

One of the most prominent – and currently unmet – needs expressed by our ELL participants is the need for language support in English MOOCs regardless of the subject matter, a direct result of the value learners place on the knowledge of English to improve their economic, social, and geographic mobility. This implies that adaptively supporting ELL students in English-language MOOCs is at least as important for this audience as translating MOOCs to their local languages would be. MOOCs can be designed to encourage and advertise English language interventions, such as content and context dictionaries, translators, or interactive video transcripts. These interventions can be deployed adaptively based on struggle behavior observable in students’ interaction logs in MOOC videos, as recent studies show that ELLs exhibit distinct behavioral patterns when engaging MOOC content (Uchidiuno et al. 2016b). Interaction systems such as Talkabout (Kulkarni et al. 2016) can be designed to encourage lurkers in discussion forums or even in live video chats to interact with other students, and conversational agents such as (Tomar et al. 2016) can be designed to not only facilitate conversation between peers, but actually interact with ELLs in a way that adjusts to their English language ability, and gives them the opportunity to practice their language skills without the pressure of speaking in public.

In addition to English language support, our findings suggest that the need for career advancement is especially prevalent for ELLs. Meeting this need implies that MOOCs should not just deliver content, but serve as a platform that helps students advance their career paths. Students’ prior qualifications could be pulled in from social networking sites such as LinkedIn and Facebook, which could form the basis of adaptive course recommendations for students. Adaptive recommendations could also be provided by allowing students to indicate the skills that they desire to strengthen and their career aspirations on the MOOC platform. Supporting strategic course choices can help MOOCs support informed student decision-making about how courses can advance their career paths, rather than leaving it entirely on students to make such choices. It also implies that more careful consideration has to be placed on how certificates are designed, conferred, and valued. For example, ELL students’ need for “proof of knowledge in English” means that certificates should not only include the name of the course, but the language it was completed in. Additionally, certificates could be adapted to demonstrate the degree of English mastery; even in domain specific MOOCs such as Psychology, students’ language abilities can be evaluated either using automatic grading mechanisms, or by their peers to minimize the burden on MOOC instructors. Further validation of language skills would provide valuable information for potential employers and a valuable asset for high-performing students. For students who already have prior knowledge of the course content, but need proof of that knowledge, better placement tests can be used to adaptively personalize and deliver the right content from within the course, rather than requiring all students to participate from start to finish. To increase the marketability of students’ skills, professional social networking sites like LinkedIn can be connected to student MOOC accounts so their certificates can automatically be included as part of their profile information.

Our findings also show that ELLs are highly motivated to participate in MOOCs, but given research findings that they are unlikely to participate in discussion forums, it may be an indication that MOOCs are not supporting them in a way that meets their needs. Our findings suggest this may be caused by language limitations, or cultural biases that influence their level of comfort interacting with strangers online, or both. Such learners can be engaged by humanizing interactions in MOOCs, and using existing platforms that these learners are already familiar with. For example, recent research studies show that ELLs are more likely to interact on social media platforms such as Facebook rather than the MOOC discussion forums (Kasunic et al. 2016). Pushing and pulling students’ social media account information to and from their MOOC accounts may allow students to learn more about their online classmates and increase the sense of familiarity, rather than just experience them as a username in a text-based discussion forum. Adaptive systems can also take these factors into account when matching students for course discussions, and encouraging them to participate in discussion forums. Students are currently being matched in discussion forums based on the time and order that they join (Tomar et al. 2016), but there is evidence that shows that such matching strategies are ineffective (Oura et al. 2015). Matching ELLs with other English learners may help to decrease their discomfort with speaking up even with their language limitations. Also, it may be desirable to develop systems to adaptively moderate, personalize, or prioritize discussion forum interactions to allow representation from all students in the forum. This suggestion is supported by research that shows discussion forums are much more effective when they are scaffolded by the instructor, set at scheduled times, and make up a percentage of the students’ final grade (Kotturi et al. 2015; Kulkarni et al. 2016). Students can also be adaptively prompted to discuss topics that are not directly related to course content before beginning class discussion, such as favorite foods or places they have lived. This may serve to engage ELL students who are more interested in social and cultural conversations than in discussing course content, while simultaneously humanizing all participants as suggested above. Finally, students could be provided with safe opportunities to continue interacting with students with whom they have taken courses with in the past, to foster a sense of long lasting community like the relationships formed in physical classrooms.

Meeting students’ needs for geographic mobility in MOOCs involves highlighting and celebrating the diversity of culture of MOOC participants. Tools such as Talkabout by Kulkarni et al. (2016) show tremendous potential to allow global students connect around course content and culture in a way that improves students’ performance scores, as well as ELL students’ social and geographic mobility needs. Possible interventions include allowing and encouraging students to create personalized discussion channels as they see fit including culture and language based forums. MOOCs can be designed to adaptively highlight, promote, and recommend discussion posts that align with students’ desired culture or language needs, rather than presenting them chronologically by time only to everyone. Student profile information can be augmented with information that shows their current location, the culture they identify with, and the languages they speak in order to encourage engagement from students who desire to participate in the same culture. This information should not only be provided for students who are currently enrolled, but should also include course alumni to increase the opportunities for students to find people who are part of their desired culture.

Finally, the relationships between these factors should also affect the design of adaptive systems. For example, students with the goals of geographic mobility may be motivated by economic mobility, social mobility, both, or neither. Among our interviewees, the goals of economic, social, and geographic mobility are not mutually exclusive, as shown in Table 4. However, the pattern among our interviewees is different from the pattern of co-occurrence that emerges from our survey respondents, based on the PCA analysis of our survey data. For interviewees, social and geographic mobility may appear together or independently of one another, but always in the presence of economic mobility. For survey respondents, economic mobility appeared as an independent factor and always co-occurred with social mobility, while geographic mobility only appeared in the presence of both economic and social mobility. We hypothesize that the difference between these groups relates to the current living situation of the learners. Only four of our 12 interview subjects were still living in their country of origin, and 50% had immigrated to the United States. In the course where we deployed our survey, only less than 1% of the students reported living in the United States. The prospect of geographic mobility requires planning for economic mobility (e.g. a job in the new country) as well as social support. However, when dealing with the consequences of a move, the need for social support is likely to vary with the individual’s specific situation. A purely geographic approach to supporting these students might be to help them learn about their target country or region, while a social approach might involve connecting them to learners who are also interested in connecting with their peers. Knowing that geographic mobility goals imply the need for social mobility, however, suggests a third strategy: reaching the right people in the right way. Systems to support these learners could connect them to peers in the geographic area they are targeting, and facilitate conversations related to the local economy and culture. However, for learners for whom economic and social mobility goals are not intertwined with geographic mobility, this strategy would not be appropriate.

Finally, how likely are MOOC providers to adopt any of these changes? Evidence suggests that popular MOOC content providers are interested in supporting ELL students. They advertise their platforms as an opportunity for providing “universal access to the world’s best education” (https://www.coursera.org/), and state goals such as building a “thriving worldwide community of educators and technologists who share innovative solutions to benefit students everywhere” (edX, 2015). However, their commitment to supporting ELL students goes beyond advertising. These content providers are continuously making changes to their platform to increase access for ELL students. For example, we reviewed archived courses deployed on Coursera in 2013, and the only language support features available were English language captions accessed by the ‘CC’ icon. As of 2016, the platform has been changed to allow students to access captions using a more universal subtitles icon, subtitles are now available in 8 different world languages, video transcripts are automatically opened below the video pane, and students now have the ability to search the transcripts while the video is playing. These features are available even in a Machine Learning course (Machine Learning - Stanford University 2015), and is evidence that MOOC platform developers care deeply about supporting the needs of ELL students in MOOCs. Our work contributes to this research area that is gaining increased attention, and will benefit not only ELL students, but the general MOOC student body at large.

Limitations

There are several limitations to our research study. All of our interviewees were ELLs and had at least some basic fluency in English; therefore, we cannot draw conclusions about those who have little to no English language fluency. Our interviewees all had some form of education, and a few were already professionals in their field; our findings may not be extensible to students who have no higher or formal education. Also, our survey was deployed to students who had some interest in learning English, so different motivations may emerge if it is deployed in non-ELL focused MOOCs (e.g, ELLs who take MOOCs in their local languages rather than in English). We did not independently verify their participation in MOOCs with content providers. Finally, our participants are not representative of all cultures from which MOOC students may come, and there may be other motivations influenced by specific cultures. However, regardless of their English fluency, cultural needs or prior education, our data show that a singular way of redesigning MOOCS is unlikely to meet the needs of all ELLs, and – more broadly – non-native English speakers.

Conclusion

Non-native English speakers differ in their learning and cultural needs, and they enroll in MOOCs for a diverse set of reasons – all factors that can potentially impact the ways to scaffold their long-term goals. Our findings show that designing for individual motivations without the context of their long-term goals, and without considering the role of other stakeholders (such as employers), may not actually meet their needs. We show that although ELL students express similar goals of economic, social, and geographic mobility, they aim to achieve the same goals using a range of strategies that call for different design interventions. Rather than defaulting all ELL students to one-size-fits-all approaches such as translation, researchers must develop adaptive ways to understand students’ needs and motivations, and present them the necessary personalized interventions, if they are to satisfy their needs.

Education is not only a means for people to learn educational content. Higher education is a means for people to develop life-long friendships and bonds, become independent, develop new interests, form beneficial networks, and improve their economic situation (McGuire 2011). The difference between MOOCs and traditional classroom-based learning is in the means of instruction, and so we cannot assume that the only reason that students participate in MOOCs is to learn course content. In the same way that schools and instructors adjust to the needs of groups and individuals, MOOCs have to be designed so that they can both recognize and adjust to needs of diverse students to support their needs and ultimately, to increase access for students that MOOCs are not currently serving.