There is a growing need for engineers to sustain our current life and maximize the best possible life for our future in our society (National Science Foundation, 2021). However, about half of the students who enter a post-secondary engineering program in the United States leave before graduation, and students who quit are disproportionately from minoritized groups (Besterfield‐Sacre et al., 1997; Huang et al., 2000; Beering et al., 2007). One reason for this discrepancy is that the structure of higher education engineering programs may lead minoritized studentsFootnote 1 to feel that they identify less with their program and with the identity of “engineer” than do non-minoritized students (Baumeister & Leary, 1995; Maslow, 1954). They are additionally more vulnerable to experiencing negative consequences (e.g., dropping out) when they do not feel a sense of belonging, because lack of belonging can be seen as evidence of confirming negative stereotypes from groups with which they identify (Steele, 1998; Steele et al., 2002).

Prior work to address this attrition gap has demonstrated that role models, figures that a student relates to and seeks to emulate over time (Lee & Eccles, 2021; Lockwood & Kunda, 1997), can promote students’ interest and achievement in Science, Technology, Engineering, and Mathematics (STEM) fields like engineering (Herrmann et al., 2016; Shin et al., 2016). Eccles and colleagues’ Situated Expectancy–Value Theory (SEVT; 2020) posits that socializers like role models can help increase students’ confidence to succeed in (“Can I do it?”) and their perceived value of (“Do I want to do it?) engineering—both of which are important predictors of academic choice, engagement, and performance. For minoritized students, exposure to successful non-stereotypical role models (i.e., those who are not White men in engineering) has been shown to positively influence their academic motivation (Cheryan et al., 2011). Unfortunately, given systematic historical inequalities in engineering, many students do not have access to such a role model in their immediate surroundings (Chubin & Babco, 2003).

In order to increase access and exposure to these role models, we leverage social media, particularly YouTube, as a platform where engineering students can connect with other students like them in a more geographically and temporally flexible way. Although research on social media has often focused on its detrimental effects on students, such as how excessive time spent on social media leads to decreased well-being (Brunborg & Andreas, 2019; Coyne et al., 2020; Kara et al., 2021) or how multitasking with social media during school work leads to negative academic outcomes (Demirbilek & Talan, 2018; Lau, 2017), less attention has focused on how social media can be used positively as an opportunity to connect with and learn from like-minded peers. Within the current study, we demonstrate the potential of social media for serving minoritized community college students in ways that can reduce attrition and foster positive STEM outcomes.

Theoretical Framework

We frame our work using Eccles and Wigfield’s (2020) Situated Expectancy–Value Theory (SEVT), formerly known as Expectancy–Value Theory (EVT; Eccles et al., 1983). In SEVT, Eccles and Wigfield (2020) describe central drivers of students’ achievement-related choices (e.g., subsequent course enrollment) and performance (e.g., course grades; see Fig. 1). Much of the prior work using SEVT for interventions has focused on directly promoting students’ belief that they can do well in an upcoming task (i.e., expectancies for success) and want to engage in a task (i.e., subjective task values)—the most proximal predictors of achievement-related choices and performance (Canning et al., 2018; Harackiewicz et al., 2012; Hulleman et al., 2010). Unfortunately, less discussed in intervention work is the middle portion of the SEVT model in which Eccles and Wigfield (2020) point out that there is a connection from socializers to individuals. Socializers have the ability to form, develop, and solidify expectancies for success and subjective task values, which in turn predict achievement-related choices and performance, such as subsequent course enrollments and course grades. Prior research has found that socializers, such as peer-related role models, can create an environment that stimulates positive motivational outcomes (Van Ryzin et al., 2009; Wentzel et al., 2010).

Fig. 1
figure 1

Situated expectancy–value theory model

Full size image

One way peers can bolster another student’s motivation, like expectancies for success, is through sending social belonging messages (Anderman, 2003; Goodenow, 1993). Social belonging in the school context is defined as “the extent to which students feel personally accepted, respected, included, and supported by others in the school social environment” (Goodenow & Grady, 1993, pp. 60–61). Students who see peers like them—in feature, structure, or purpose—express how they were able to accomplish goals despite difficulties are more likely to feel they belong, which can improve their expectancies for success (Anderman, 2003; Sánchez et al., 2005). Another way peers can bolster another student’s motivation is through talking about their subjective task values (Harackiewicz et al., 2012; Rozek et al., 2017). Harackiewicz et al. (2012) and Rozek et al. (2017) found that parents who watched testimonials from their children’s peers discussing the usefulness of mathematics and science courses promoted their children’s STEM motivation, and ultimately, course enrollment and career aspirations in STEM. If peer values can affect students’ motivation indirectly via their parents, then there is greater potential to promote students’ motivation when students can directly hear from their peers.

Community college engineering students, who are likely at a critical point in their lives determining their identity as engineers, may be positively impacted by watching and hearing from others discuss their expectancies for success and subjective task values in engineering. As we found in our prior work, engineering college students naturally bring up their expectancies for success and subjective task values in their peer-advice videos (Lee et al., 2023); they discuss how interesting, useful, relevant engineering is, as well as ways to overcome the struggles faced in engineering. The current study examines the extent to which peer role models in engineering can potentially influence subsequent course enrollments and course grades through delivering these messages in videos. Furthermore, as Eccles and Wigfield (2020) posit in SEVT that students’ prior experiences and stable characteristics of the individual predict subsequent course enrollments and course grades, the current study includes gender, ethnicity, socioeconomic status, and prior achievement as covariates in our analytical model predicting subsequent course enrollments and course grades.

The Community College Context

Community colleges play an important role in providing students broad access to a post-secondary degree and preparing students for the workforce due to their open admission policies, flexibility in course scheduling, short commutes to campus, and less expensive tuition (Juszkiewicz, 2015; Somers et al., 2006). The accessibility and affordability of community colleges means that they are home to a more diverse student population, including students who are first-generation, non-White, low-income, older, working full-time, and parents (Phillippe & Sullivan, 2005; Cohen & Brawer, 2013). For example, around 55% of community college students identify as non-White, 59% identify as women, and the average age of first-year students is 27, which is higher compared to four-year universities (American Association of Community Colleges, 2023). As certain students are more likely to attend community college, community colleges are especially useful for preparing students from non-traditional and underrepresented backgrounds in STEM subjects, such as engineering (Hoffman et al., 2010). Students from diverse backgrounds bring different perspectives to engineering, which are especially useful for promoting creativity, productivity, and innovation within the field (Smith-Doerr et al., 2017).

However, one challenge students face when entering community college is that they are often unprepared for the academic rigor of their courses. This challenge is in part because, for many underprepared students who may also come from historically underrepresented backgrounds, community colleges may represent their only access to higher education in the United States (Boggs, 2011). For example, first-generation students may not be prepared for their courses because they may overestimate their ability due to their lack of knowledge about the rigor of college-level courses (Cohen et al., 2013; Grimes, 1997). Further, this challenge is especially pronounced for minoritized students because of systematic inequalities (Bush & Bush, 2010). Another challenge STEM community college students—particularly those who are women—face is social belonging as a result of their identity. For example, Marco-Bujosa et al. (2021) found that women in a community college STEM program reported encountering gender stereotypes, social intimidation from students who were men, and academic isolation, all of which negatively impacted their persistence. Not only do women in STEM face a lack of social belonging within the classroom, but they may also encounter a “chilly climate,” characterized by an unsupportive environment within and outside of the classroom, along with negative comments from peers and instructors that can decrease their persistence within their major (Jorstad et al., 2017). This challenge is especially vital to address among students in their first few years in college, such as in community college, because oftentimes, students switch out of STEM during this time period (Bahr et al., 2022; Chen, 2013). Although much work has been done to address these challenges within community colleges (Fink et al., 2023; Freeman & Huggans, 2009; Gilken & Johnson, 2019), more work is needed to promote success for community college students from different socio-demographic backgrounds with multiple identities. Also, more work is needed in the engineering community college context, given the unique environment and demands of engineering programs and given existing inequities.

Prior literature has found two key predictors of success for community college students: demographics (Nakajima et al., 2012; Porchea et al., 2010) and socio-cognitive beliefs (Fong et al., 2017, 2018; Sorey & Duggan, 2008). Certain demographic factors have been negatively associated with student success. Low socio-economic status (SES) has been generally associated with lower success rates in community college students, in part because these students often have to work full-time, limiting the time they can devote to their academics (Cofer & Somers, 2001; Schmid & Abell, 2003). Research has also indicated that students’ identification with certain races/ethnicities may negatively predict success in community colleges. Black and Hispanic students in particular have lower academic outcomes and overall retention compared to White students, even when they actually report higher levels of academic engagement (Cofer & Somers, 2001; Greene et al., 2008). Even though racially/ethnically minoritized students may be more academically engaged, they may have less exposure to college-level academics prior to entering than their White peers, resulting in poorer academic outcomes (Greene et al., 2008).

In addition to demographic factors, certain socio-cognitive beliefs, including self-perceptions, motivation, and critical thinking skills, are positive predictors of student persistence and success within community college students (Fong et al., 2017, 2018). Motivational beliefs, in particular students’ intrinsic values, were shown to be strong predictors of student success (Seo & Lee, 2018). However, motivational beliefs differ with demographic characteristics, such as gender and race/ethnicity (Else-Quest et al., 2013; Starr et al., 2022; Wang et al., 2015). In one study, high school-aged boys reported higher expectancies and higher self-concept than high school-aged girls (Else-Quest et al., 2013). Similarly, research shows that racially/ethnically minoritized students who face structural barriers (e.g., racism) have decreased motivational beliefs in STEM (Starr et al., 2022). Thus, there is strong empirical grounding to suggest that minoritized students in STEM can benefit from motivation interventions that promote their socio-cognitive beliefs (e.g., expectancies, values).

Seeing the success of peers in engineering who come from similar demographic backgrounds and have high motivation can promote higher motivation in minoritized engineering students (Shin et al., 2016). Within community colleges, transfer students may be especially useful role models, given these students have successfully made the transition to four-year institutions (Lockwood & Kunda, 1997; Lockwood et al., 2002). For the current study, we focused specifically on community college engineering students who aimed to transfer to a four-year university to pursue a bachelor’s degree.

Transfer capital is another factor that may influence students’ success in transferring to a four-year university (Laanan, 2004, 2007). Community colleges provide students with an opportunity to accumulate different forms of capital, called transfer capital, that enable them to develop the skills necessary to successfully transfer to a four-year university (Laanan, 2001, 2007). These skills include students’ understanding of credit-transfer agreements between different institutions, the grades required for admission into the desired major, and the prerequisites required for their major (Laanan et al., 2010). By facilitating students’ development of transfer capital, community colleges play an important role in promoting students’ successful transfer. Institutions that provide clear transfer pathways facilitate access of transfer capital to transfer students, thereby increasing the likelihood that students have a more successful transferring experience (Laanan, 2007; Jain et al., 2011). In other words, within community colleges, STEM students’ motivation and success in transferring to a four-year university is influenced by their engagement with services aimed at promoting transferring, with more frequent engagement associated with greater transfer momentum (Wang et al., 2017). Prior work has found that STEM students who successfully transfer to a four-year university are more likely to have taken math and other STEM courses, while taking less credits in other, non-STEM courses (Wang, 2016). Particularly, successfully transferred STEM students took more courses concentrated within the physical sciences (Wang, 2016). Overall, findings indicated that STEM community college students may benefit from resources that target a diverse range of student needs, from advice on the transfer process, to course-specific information (Wang et al., 2017).

Relevant Intervention Research

Two bodies of intervention research support the promise of role model-based initiatives. First, motivational interventions (e.g., social belonging, utility value) have used aspects of role modeling to enhance students’ choice and performance (see Harackiewicz & Priniski, 2018 for review). In an example of social belonging interventions, first-year undergraduates read a letter or watched a video from upperclassmen (i.e., role models) discussing how everyone struggles through their transition to college, regardless of their racial/ethnic backgrounds, and how this struggle lessons over time (Stephens et al., 2014; Walton & Cohen, 2007, 2011). In typical utility value interventions, high school students (Gaspard et al., 2015; Hulleman & Harackiewicz, 2009), college students (Canning et al., 2018; Hulleman et al., 2010), or even high school students’ parents (Harackiewicz et al., 2012) write or watch a video articulating the usefulness of the material they are learning in class.

Although motivational interventions show positive effects on students’ academic outcomes like course grades, grade point averages (GPA; Hulleman & Harackiewicz, 2009; Walton & Cohen, 2011), and subsequent course enrollment (Canning et al., 2018; Lacosse et al., 2020), these types of studies have not approximated naturalistic impactful role modeling relationships. For instance, in social belonging intervention studies, students typically do not have agency in choosing a role model (i.e., who they wanted to see, hear, or watch). Prior literature on role models notes that people are more likely to view someone as a role model if they see a resemblance between themselves and the potential role model in features, structure, and purpose, as well as see the attainability of the role models’ success (Lockwood & Kunda, 1997). Leveraging choice in role model selection may increase perceived resemblance on the specific features most important to the student. In utility value interventions, students typically choose who they want to write the letter to; however, in these interventions they are in the position of being the expert (i.e., the role model). In other role model work, such as that conducted by Harackiewicz et al. (2012) investigating the effects of parents watching videos of successful college students who serve as role models for their high school children, participants also do not have a choice of role models.

Moreover, motivational interventions in the community college context have demonstrated mixed effectiveness. One expectancy-value based intervention delivered via text messages was shown to statistically significantly increase re-enrollment of pre-allied health students (O’Hara et al., 2022). But Canning et al. (2019) found that a utility-value intervention was not effective for increasing retention for STEM students in community colleges, and instead negatively affected students, especially low-performing students. One potential explanation for this may be that in the community college context the intervention was perceived as too challenging for students who were already struggling in the course, thereby causing disengagement for low-performing students (Canning et al., 2019).

In our study, we attempt to design and implement an effective intervention in a community college setting by having participants interact with someone who has overcome similar challenges to those they face (i.e., transferring as an engineering student) and we use writing prompts to encourage taking specific ideas from the role model. Furthermore, most motivational interventions cater to a specific message, such as challenges are normal in the transition to college and this feeling is temporary (Stephens et al., 2014; Walton & Cohen, 2007, 2011) and learning math and science is useful for my future career (Harackiewicz et al., 2012). These messages are generally crafted top-down from researcher to student, rather than asking students to naturally bring up topics that they believe are important to share to their peers. This bottom-up approach can be particularly useful for gaining greater insights into students’ beliefs without constraining their responses to predetermined categories developed by researchers using top-down approaches (Libarkin & Kurdziel, 2002).

A second promising body of intervention research on role modeling is work exposing students to non-stereotypical role models (Hernandez et al., 2017; Herrmann et al., 2016; Shin et al., 2016). Such role models support greater engagement and success in school among minoritized students (Herrmann et al., 2016; Shin et al., 2016). However, studies in this realm have typically exposed students to non-stereotypical role models via confederate interactions or narratives (e.g., Stout et al., 2011; Van Camp et al., 2019), which restrains ecological validity, because participants are generally in a lab setting where they interact with the confederate once rather than on a regular basis. People often consider someone who they interact with or watch frequently as a role model (Lee & Eccles, 2021). These interventions have had success in boosting participant confidence and values, and undermining negative group stereotypes (Hernandez et al., 2017; Herrmann et al., 2016; Shin et al., 2016; van Camp et al., 2019). However, more work is needed to bring together components from different interventions that have been shown to promote students’ choice and performance for long-term effects. Our approach allowed students to be repeatedly exposed to a non-stereotypical and similar background engineering student of their choosing rather than a one-time exposure to a non-stereotypical and similar background engineering student without choosing and using a writing component from motivational interventions with a role-model intervention. Writing provides the opportunity for students to make personal connections through lexical, syntactic, and rhetorical devices (Emig, 1977). In other words, students are able to see a role model like themselves and then reaffirm their personal experiences, expectancies, and values from what the role model said by writing.

Social Media and Education

With the advent of technology, students seek and share educational content through social media platforms (Mao, 2014; Moghavvemi et al., 2018). Social media is defined as the different technologies that facilitate collaboration and communication across the world, with notable examples being YouTube, Instagram, Twitter, and TikTok (Joosten, 2012; Tess, 2013). Although much research has examined the ways in which students use social media for entertainment purposes (Ezumah, 2013; Whiting & Williams, 2013), some research has explored the ways in which social media can improve students’ learning of content. In Moghavvemi’s (2018) study, undergraduate business students reported supplementing their traditional classroom learning by watching course-related videos that covered content their course textbooks and notes were not able to help them with. Similarly, Orus et al. (2016) found that students who actively participated in creating YouTube videos where they explained course topics had greater cross-curricular competencies. Outside of learning course-specific content, social media has also been utilized as a platform to develop social connections and share resources with peers and instructors in order to enrich learning outside of the classroom (Anthony & Jewell, 2017; Malik et al., 2019). For example, using Twitter, undergraduate and graduate students in social work reported that they were able to form connections with peers and other professionals within the field, as well as share resources and knowledge about real-world issues (Anthony & Jewell, 2017). Few studies, however, have leveraged social media outside the classroom to reduce challenges in the community college context.

The Current Study

In the present study, we bridge various interventions in an under-considered context: community college engineering. Community colleges are important because they play a crucial role in decreasing barriers to not only who can pursue higher education, but also who chooses STEM (Mooney & Foley, 2011). First, we contribute to the intervention literature by providing community college engineering students multiple interactions with a role model like them in a less artificial, lab-like environment through using YouTube videos. We also combine key mechanisms from other interventions in order to create longer-lasting effects. For instance, not only are engineering students in community college exposed to successful engineering students repeatedly throughout the academic term, but they are also asked to retain the information from the YouTube video as well as reflect on their own experiences after hearing from the role model through reflective writing, a mechanism that has garnered positive results in SEVT interventions. Third, engineering community college students are provided more agency to choose whose YouTube videos they want to watch—providing this kind of agency in choice of intervention engagement has been shown to increase motivational beliefs (Rosenzweig et al., 2019). More broadly, providing students more agency in the classroom has been associated with higher intrinsic motivation, greater competence beliefs, and improved performance on exams (Patall et al., 2008). Finally, we strive to close the gap in the literature that often focuses on using YouTube and other social media platforms for course-specific learning rather than non-course specific learning (e.g., experience and lifestyle videos as opposed to how to code using MATLAB videos). In other words, engineering students in the videos discussed their experiences regarding how they studied for their courses, transferred from a community college to a four-year university, or engaged in engineering opportunities outside of the classroom, rather than teaching content-specific lessons about engineering, such as how to use the MATLAB software. In the current quasi-experimental study, community college students watched a series of four advice-related YouTube videos from engineering students they selected who successfully transferred from a community college to a four-year program. These previous students are potentially valuable role models because they have relevant experiences as engineering students in community college and have successfully transferred to a feeder four-year institution. We examined the following research questions (RQs):

RQ1 To what extent do community college engineering students who participated in the role model YouTube intervention have higher achievement (i.e., course grades) compared to similar students who did not participate in the intervention?

RQ2 To what extent do community college engineering students who participated in the role model YouTube intervention have greater engineering course enrollment rates in the following term compared to similar students who did not participate in the intervention?

We hypothesized that students in the intervention group would have both higher course grades and subsequent course enrollments.

RQ3 For each of the above RQs, are there meaningful subgroup (i.e., gender, ethnicity, and socioeconomic status) differences in these associations?

We hypothesized that women, students with low income, and underrepresented racial minorities would benefit more from the intervention than would men, students with high socio-economic status, and students who did not identify as underrepresented racial minorities.

Method

Participants

The universities and faculty in this study are part of a larger National Science Foundation (NSF)-funded Scholarships in Science, Technology, Engineering, and Mathematics (S-STEM) program. The goal of the S-STEM program is to enable academically talented, low-income students to complete their STEM baccalaureate degree and contribute to the U.S. economy with their successful STEM careers. To achieve this aim, faculty, staff, and students from a four-year institution in Southern California established a collaboration or a S-STEM ecosystem with faculty, staff, and students from a local two-year institution. Our S-STEM program particularly focused on increasing the number of community college students who transfer to a declared engineering major at a four-year institution, the number of community college students who graduate with a baccalaureate degree in engineering, and the number of community college students who enter the engineering workforce or pursue a graduate degree in engineering. Collaborators on the NSF-funded S-STEM grant from the four-year institution worked with professors from the two-year institution on the grant to recruit students in their introductory-level engineering courses. The enrolled student population at the community college is 66% part-time, 31% White, 29% Asian, 23% Hispanic/Latino, 6% two or more races, 2% Black/African American, less than 1% Native Hawaiian/Other Pacific Islanders and American Indian/Alaska Native, and 41% receiving financial aid. Students in this study included those who took introductory-level engineering courses at the community college taught by the two engineering professors on the grant. There was a total of 11 different engineering courses: four core engineering courses (Programming and Problem Solving in MATLAB; Statics; Materials Science and Engineering; and Dynamics), six computer-aided design engineering courses (Electrical and Computer Engineering; Graphics and Geometry; Computer-aided Design Technologies; Architectural Drafting; Computer-aided Drafting; and Drawing and Design), and one hands-on lab engineering course (i.e., Methods). Out of the 11 engineering courses, seven are offered both fall and spring terms. Two engineering courses were offered only in the fall (Methods and Architectural Drafting) and two engineering courses were only offered in the spring (Computer-aided Drafting and Drawing and Design).

The intervention group consisted of students in any of the 11 engineering courses during fall 2020 and spring 2022 who decided to participate in the study for extra credit (see below for invitation procedures). The control group consisted of students who took the same 11 engineering courses during fall 2020 and spring 2022 but did not opt to participate in the study for extra credit. Table 1 presents the demographics across intervention (33%) and control (67%) groups for the 537 total students in the study.

Table 1 Characteristics of intervention and non-intervention group in full and matched samples
Full size table

Procedure

For each course in which the intervention was implemented, the professor uploaded the study sign-up information document to their course Canvas page in the first two weeks of the semester. This time frame allowed students to have enough time to opt into the study if they recently added the course. If students were interested in participating for extra credit, then they were directed to Qualtrics to consent. Participants who consented to participate in the intervention were informed that the study would involve the researchers collecting data on students’ experiences in engineering.

Figure 2 presents the steps of the intervention. Students had to complete pre- and post-surveys about their engineering attitudes and beliefs and four writing prompts via Qualtrics. The writing prompts included links to relevant YouTube videos. These pre- and post-surveys and writing prompts were due every two weeks of the semester. Students watched and responded to the video writing prompt in the same Qualtrics survey. Before each survey due date, two reminders, one a week before and another a day before the due date were sent for each survey. First, students were sent a Qualtrics survey link via email, which asked them to choose one out of six engineering students (or YouTubers) whose videos they wanted to watch. The YouTubers were selected for their appropriateness as role models in this specific context; they were recruited through the S-STEM program because seeing and hearing from engineering students who had successfully transferred to a four-year university can serve as role models for current engineering community college students (see Lee et al., 2023 for more information). Specifically, we recruited low-income engineering students who had successfully transferred to a four-year university via the S-STEM program. Among the YouTubers, four were men and two were women. Two of the YouTubers identified as White, three as Latinx, and one as Asian. All of the YouTubers were low-income. Each YouTuber had a brief biography, profile picture, and a description along with the title of their four videos to help participants choose (see Fig. 2 for example). Seventeen percent of students chose Hillary; 8% of students chose Eduardo; 10% of students chose Lukas; 20% of students chose Francis; 25% of students chose Adam; and 20% of students chose Emily. 59% of students chose a YouTuber of the same gender (62% of men chose a man YouTuber, 53% of women chose a woman YouTuber). 30% of students chose a YouTuber of the same ethnicity/race (31% of White students chose a White YouTuber, 41% of Latinx students chose a Latinx YouTuber, and 24% of Asian students chose an Asian YouTuber). Then, in the same survey on Qualtrics, participants were provided a link to their first video. The link directed them to YouTube, which included the title of the video, description of the video with timestamps of topics covered (e.g., “Here are the sections in the video to navigate easily: GPA, Exams, and Class Success at 1:26, Knowing Your Optimal Studying at 3:27, and My Own Take on Studying at 4:29), and channel description. The YouTube videos were uploaded as unlisted, meaning that the general public could not see the videos unless shared, in order to prevent participants from accidentally watching them before the intervention. After the students watched the video on YouTube, they went back to Qualtrics to respond to the writing prompt. Finally, students watched another video (from the same originally selected YouTuber) and responded to a writing prompt three more times during the semester, with two weeks between each activity session. The study was approved by the community college’s and four-year university’s Institutional Review Board.

Fig. 2
figure 2

Intervention steps

Full size image

Intervention

The intervention consisted of two main components: watching prepared videos and responding to the main topic discussed in the videos via writing. The 5-to-10-min-long videos were created in the 2019–2020 academic year by engineering students in a four-year university who transferred from a community college. The community college students in this study typically feed into the same four-year university. Given the importance of student choice and how it can lead to greater interest and engagement in a task (Patall et al., 2008; Rosenzweig et al., 2019), transfer engineering students were prompted to freely choose four topics that they thought were important to share with community college students. Video topics included: personal background (n = five students; six videos), study tips (n = four students; five videos), engineering opportunities (n = three students; three videos), COVID-19 (n = three students; three videos), transferring experience (n = two students; two videos), and adapting to the quarter system (n = two students; two videos).

Community college participants were prompted in Qualtrics to watch four of these videos. After watching each video, participants completed writing activities designed to help internalize the message from the YouTube videos. In particular, students were prompted to write a letter to a future engineering student wherein they connect the topics covered in the video to their personal experiences. The prompt was chosen because being in a position where individuals have to advocate for a message can increase the likelihood of internalizing that message (Higgins & Rholes, 1978). Students were asked to connect the video to their personal experience, because making connections to students’ lives has been an integral component of the effectiveness of utility-value interventions (Harackiewicz et al., 2016; Hulleman et al., 2017). All writing prompts started with a short recall description of the video, which was then followed by questions to incorporate into the letter detailing personal experiences and components from the video. An example of a writing prompt:

“In this video, Hillary shares her academic schedule. She talks about how she studied for different engineering courses and the different resources she used to excel academically. Write a letter to a future engineering student who is struggling academically. As you write this letter, imagine a student like you―in the same program and with the same ethnicity and gender. Include a specific time when you struggled in an engineering class. What did you do to overcome it? What tips would you recommend that proved to be helpful in your case? Was there anything you regretted doing or wished you changed? Incorporate detailed advice from your personal experience and components talked about in the video. Please be as specific as possible.”

Measures

Intervention Group

Participation was recorded by course instructors. The intervention group was dichotomized as 0 = not in the intervention group and 1 = in the intervention group.

Engineering Course Grade

Final grade in the engineering course was collected from institutional records. A continuous grade variable was created where 0 = fail, failing withdrawal, incomplete to an F, no pass, excused withdrawal, military withdrawal, and withdrawal, 1 = D, 2 = C, 3 = B, and 4 = A.

Subsequent Engineering Course Enrollment

Whether or not students enrolled in at least one engineering course the following semester when taking at least one course was obtained from institutional records. This variable was created only for students who enrolled in the 11 engineering courses during fall 2020, spring 2021, and fall 2021, because students had not yet enrolled for their next term courses during spring 2022 at the time institutional records were collected. Subsequent engineering course enrollment was dichotomized as 0 = the absence of a subsequent engineering course in the immediately following semester and 1 = the presence of a subsequent engineering course in the immediately following semester.

Engineering Course

From the 11 engineering courses, three categories of courses were created from the curriculum design: core, computer-aided design, and lab. Courses classified as core courses were Programming and Problem Solving in MATLAB, Statics, Materials Science and Engineering, and Dynamics. Those coded as computer-aided design were Electrical and Computer Engineering, Graphics and Geometry, Computer-aided Design Technologies, Architectural Drafting, Computer-aided Drafting, and Drawing and Design. Methods was coded as a lab course.

Term Description

Term description refers to the semester in which students were enrolled in the 11 engineering courses: fall 2020, spring 2021, fall 2021, and spring 2022. Indicator variables were created for each term, with fall 2020 as the reference category.

Instructor

Instructor refers to which instructor students had out of the two instructors where 0 = instructor one and 1 = instructor two.

Gender

Gender was obtained from institutional records and asked students to indicate whether they were men or women. Students who were not men were combined into one group where 0 = men and 1 = women.

Ethnicity/Race

Ethnicity/race was obtained from institutional records in which students had indicated whether they were Asian, Black or African American, Hispanic/Latino, Native Hawaiian or Pacific Islander, White, two or more races, or unreported/unknown. Students who did not identify as Asian and/or White were operationalized as racially minoritized students. (or URM; Asai, 2020).

Low Income

Low income was operationalized as whether or not a student received a Pell Grant, which is a U.S. federal grant for college students who have exceptional financial need. This measure was obtained from institutional records. A variable was created to indicate the student was low income (1) or not (0).

High School GPA

High school GPA was obtained from institutional records and on a continuous four-point scale.

Analysis

To examine engineering course grade outcomes of the role model YouTube intervention, we use the MatchThem package and procedure outlined by Pishgar et al. (2020) in R (R Core Team, 2018). Missing variables were first imputed using the mice package (van Buuren & Groothuis-Oudshoorn, 2011), which refers to the procedure of substituting missing values with a set of possible values that take into account the uncertainty in predicting the true unobserved values (Sterne et al., 2009). We imputed missing values for five datasets. Then using the multiply imputed data, we match to equate the distribution of covariates between exposure groups (Stuart, 2010). Covariates were term description, engineering course, instructor, women, URM, low-income, and high school GPA. Groups were intervention versus non-intervention groups. Matching was performed using a within approach (i.e., matching within each imputed dataset; Leyrat et al., 2019) as opposed to an across approach (i.e., propensity scores are averaged across the imputed datasets, and then matching is done; Mitra & Reiter, 2016), because the within approach typically displays better statistical performance (Leyrat et al., 2019). After matching occurred, we assessed the balance in the matched dataset—“the degree to which the balancing method was successful at achieving covariate balance in the exposure groups” (Pishgar et al., 2020, p. 294). Using the five matched datasets, we conducted a regression where the intervention group variable predicted engineering course grade controlling for term description, engineering course, instructor, women, URM, low-income, and high school GPA. As Ho et al. (2007) suggest, we included variables that were used in the matching when conducting the regressions because the matching does not completely eliminate imbalance between variables. Additionally, using the same five matched datasets, we conducted separate regressions interacting intervention by gender, ethnicity, and socioeconomic status (SES), controlling for the same covariates. Finally, we pooled the estimated models to get a single set of coefficient and standard error estimates from the imputed datasets.

To examine subsequent engineering course enrollment outcomes of the role model YouTube intervention, we followed the same procedure as above: (1) imputing the missing data in five datasets; (2) matching in each imputed datasets; (3) assessing balance on the matched data; (4) analyzing the matched datasets; and (5) pooling the causal effect estimates. The main difference in this analysis sample was that students who were in an implementing course in spring 2022 were not included, because the data for enrollment in an engineering course in fall 2022 did not exist at the time the institutional records were collected. This analysis was conducted using a logistic regression with subsequent engineering course enrollment (0/1) as the outcome. Moreover, we conducted separate regressions with interactions of the intervention group by gender, ethnicity, and SES with the same covariates.

Results

Descriptive Statistics and Missingness

All variables included in the regression analyses were normally distributed with skewness between − 2 and 2 as well as kurtosis between − 7 and 7 (Burdenski, 2000; West et al., 1995). Rank-order correlations showed statistically significant positive associations between the intervention group and engineering course grade (r = 0.23, p < 0.001) as well as subsequent engineering course enrollment (r = 0.17, p < 0.001). Chi-squared tests indicated that the intervention group, term description, engineering course, instructor, gender, low income, and URM did not predict missingness on the outcome variables. Kruskal–Wallis tests indicated that prior high school GPA did not predict missingness on the outcome variables. Therefore, missing data on variables of interest were missing completely at random (MCAR).

Success of Matching

Table 2 shows the maximum of the absolute standardized mean difference (SMD) for all covariates across the imputed datasets for course grade and enrollment analyses. The estimated maximum for the absolute SMDs for covariates are close to zero, indicating that the covariates are well-balanced in the imputed datasets.

Table 2 Balance from matching
Full size table

Intervention Group Predicting Achievement

Table 3 presents the pooled results of regressions of letter grades on the intervention group controlling for term description, engineering course, instructor, women, URM, low-income, and high school GPA. As expected, community college students who participated in the intervention had higher grades than those who did not participate in the intervention (B = 0.77, p < 0.001). The average student in the sample had a high C. The intervention would bring this grade to a middle B. A robustness check using the same matched sample without students in the spring 2022 showed similar results (see Table S1 in Online Supplement).

Table 3 Pooled causal effect regression analysis: intervention group predicting engineering course grade
Full size table

Table 4 presents pooled interaction estimates of the intervention group by gender, ethnicity, and SES. There were no statistically significant interactions between the intervention group and gender (B = 0.21, p = 0.58), the intervention group and ethnicity (B = − 0.22, p = 0.58), or the intervention group and SES (B = − 0.73, p = 0.08).

Table 4 Pooled causal effect moderator analysis where engineering course grade is the outcome: gender, ethnicity, and socioeconomic status
Full size table

Intervention Group Predicting Subsequent Engineering Course Enrollment

Table 5 presents the pooled results of subsequent engineering course enrollment on the intervention group controlling for term description, engineering course, instructor, women, URM, low-income, and high school GPA. As expected, community college students who participated in the intervention were substantially more likely to enroll in an engineering course in the next semester (OR = 2.18, 99% CI [1.22, 3.87], p < 0.01). In other words, the odds of taking another engineering course in the next semester were higher for students that participated in the intervention than for comparable students who did not.

Table 5 Pooled causal effect logistic regression analysis: intervention group predicting subsequent engineering course enrollment
Full size table

Table 6 presents pooled interaction estimates of the intervention group by gender, ethnicity, and SES. There was a statistically significant interaction between the intervention group and gender (OR = 4.27, 99% CI [1.07, 17.02], p = 0.04). But there was no statistically significant interactions between the intervention group and ethnicity (OR = 1.50, 99% CI [0.39, 5.74], p = 0.55) or the intervention group and SES (OR = 0.92, 99% CI [0.28, 3.03], p = 0.89).

Table 6 Pooled causal effect moderator analysis where subsequent engineering course enrollment is the outcome: gender, ethnicity, and socioeconomic status
Full size table

Discussion

Within this study, we examined the extent to which community college engineering students who participated in a YouTube role model intervention had higher performance and chose to enroll in more engineering courses compared to similar students who did not participate in the intervention. We also examined whether the intervention had differential associations with outcomes for student subgroups (i.e., gender, ethnicity, and SES). We found that community college students in the intervention had higher engineering course grades and greater subsequent engineering course enrollment rates than those who were not in the intervention. Also, we found that gender moderated the association between the intervention group and subsequent engineering course enrollment rates. However, there were no other statistically significant interaction effects between the intervention group, gender, ethnicity, and SES for engineering course grades and subsequent engineering course enrollment rates as outcomes. Although our study is not a randomized control trial, the results are results are consistent with the proposition that role models through social media videos can promote performance and choice in line with other motivation interventions delivered through different means (e.g., Herrmann et al., 2016; Shin et al., 2016; van Camp et al., 2019).

Together, these findings fill several important gaps in the educational literature. Specifically, our study expands research to date by incorporating social media, particularly, YouTube outside of learning course-related content to reduce challenges for community college students in engineering. Students were able to view the engineering students’ videos on YouTube on their channel page with a description. Each video also had a description of the video with timestamps of topics covered. Some engineering students who created the YouTube videos also left their contact information, such as their email to connect with them if viewers had follow-up questions. Despite motivation interventions typically using static materials in an online setting (e.g., McCarthy et al., 2017; Yeager et al., 2016, 2019), which is useful for scaling up to a larger population, our study began to allow for transformative interactions that are facilitated by social media platforms. The YouTube videos allowed students to be repeatedly exposed to non-stereotypical role models in a naturalistic context rather than a non-naturalistic context, such as a lab or reading.

Moreover, it is the only study to our knowledge that used transfer engineering students to promote engineering community college students’ choice and performance in a role model intervention. Community colleges play a crucial role in decreasing barriers for minoritized students to pursue higher education because of its accessibility and affordability (Juszkiewicz, 2015; Somers et al., 2006) as well as diversifying who chooses STEM (Mooney & Foley, 2011). This YouTube role model intervention placed in the context of community college can potentially diversify the representation of voices in engineering. In addition, prior research has shown that community college STEM students’ success in transferring is influenced by a variety of factors, including access to information about the transfer process, and preparing for university-level courses by taking the required courses (Laanan, 2007; Wang et al., 2017). This study provides a convenient and practical resource to support community college students in engineering, particularly those from minoritized backgrounds, in the transferring process.

Last, our study makes a vital theoretical contribution of integrating SEVT (Eccles & Wigfield, 2020) and role model theories (Lockwood & Kunda, 1997; Lockwood et al., 2002) in the intervention setting. Findings from our study showed that socializers, in this case, transfer engineering students, or role models who are competent, relevant, and attainable (according to role model theories; Lockwood & Kunda, 1997; Lockwood et al., 2002), have the potential to promote choice and performance in engineering (SEVT; Eccles & Wigfield, 2020).

Role Model YouTube Intervention Outcomes

Engineering Course Grade

Community college students in engineering courses who watched YouTube videos of former engineering students from similar backgrounds who successfully transferred to a four-year college and completed writing reflection exercises after each YouTube video had higher course grades in engineering compared to those who did not participate in the role model YouTube intervention. Prior research on motivation interventions has found a positive relation between participating in such interventions and course grades (Herrmann et al., 2016; Hulleman et al., 2017; Walton & Cohen, 2011). For example, Hulleman et al. (2017) found that students who participated in a utility-value intervention performed better on the final exam compared to students not in the intervention (β = 0.12). Similarly, a role-model intervention conducted by Herrmann et al. (2016) found that female students who read a letter from a female graduate student role model had higher course grades compared to students in the control condition (d = 0.24, 0.66, respectively). Likewise, social belonging interventions, such as that in Walton and Cohen (2011), found that students in a social belonging intervention had higher GPAs than students not in the intervention. These findings were particularly pronounced for Black students, who had increased grade GPAs across their time at university compared to Black students who were not in the intervention (B = 0.30; Walton & Cohen, 2011).

Consistent with these findings, results from this study showed that students in the intervention had higher course grades compared to students not in the intervention (B = 0.77). Although the effect size was larger than those seen in other motivation interventions (e.g., Hulleman et al., 2017; Walton & Cohen, 2011), this study was quasi-experimental in its design rather than a randomized controlled trial; meta-analyses of motivation interventions show that such design differences are associated with effect sizes (Lazowski & Hulleman, 2016). Our intervention has a number of features we theorized would be strengths. Unlike much prior work (e.g., Herrmann et al., 2016; Hulleman et al., 2010), this intervention was implemented more than once; in this case, it was implemented four times during the semester. Further, different from previous research (e.g., Herrmann et al., 2016; Shin et al., 2016), the present intervention asked students to watch role models talk about their experiences in engineering, retain that information, and reflect on their own experiences after hearing from the role model through reflective writing. Other work examining STEM course outcomes have often been in a lab context or through reading testimonials from role models and not in a community college context (e.g., Cheryan et al., 2011; van Camp et al., 2019) instead of a naturalistic context, such as actual transfer community college students in YouTube videos.

Subsequent Engineering Course Enrollment

Students in our intervention group also had greater subsequent engineering course enrollments compared to those who did not participate in the role model YouTube intervention. Previous work on motivation interventions have promoted students’ subsequent course enrollment (Canning et al., 2018; LaCosse et al., 2020; Porter & Serra, 2020). For example, Canning et al. (2018) found that students exposed to a utility-value intervention implemented in the first semester of a two-semester biology course sequence had greater course enrollments in the following semester compared to students not in the intervention (B = 0.30). Similarly, LaCosse et al. (2020) showed that first-year undergraduates interested in STEM who participated in a social-belonging intervention completed more STEM course credits at the end of the academic year compared to students in the control intervention, showing that they took more STEM courses after the intervention (B = 0.01). Likewise, in a role-model intervention, Porter and Serra (2020) found that female undergraduates exposed to female role models who had successfully taken the introductory economics course were statistically significantly more likely to enroll in subsequent economics courses.

Consistent with these findings, results from this study show promise for future research in promoting subsequent course enrollment. Aligned with SEVT (Eccles & Wigfield, 2020) and role model theories (Lockwood & Kunda, 1997; Lockwood et al., 2002), hearing from a minoritized student might raise another minoritized student’s choice like subsequent course enrollments. Drivers of choice in these theories include, socializers, interest (i.e., intrinsic value), personal and collective identities (i.e., attainment value), usefulness to daily life and career (i.e., utility value), and normalcy of struggles (i.e., cost). As students in the YouTube videos discussed these values, directly measuring changes in motivational beliefs to subsequent course enrollments in engineering can be used in future work to gain deeper insights into the intervention model of change.

Moderation by Gender, Ethnicity, and SES

The results of the present work indicate that gender moderated the association between the intervention group and subsequent engineering course enrollment. The association between the intervention group and subsequent engineering course enrollment is stronger for women compared to men. Aligned with previous research, Rozek et al. (2015) found that a utility-value intervention targeting parents was useful for increasing the amount of STEM courses that high-achieving female students (but not high-achieving male students) took after the intervention. Instructors should focus on providing women engineering students role models to promote choice in STEM as greater diverse voices can help solve problems the nation faces.

However, there was no other interaction between meaningful subgroups (i.e., gender, ethnicity, SES) and the intervention group that crossed the threshold of statistical significance for engineering course grades and subsequent engineering course enrollment outcomes. Community college students benefited from the role model YouTube intervention, regardless of their gender, ethnicity, or SES. One explanation may be due to selection bias, or those who choose to enroll in community college (Kenny et al., 1979). Students who enroll at four-year universities are oftentimes more academically prepared compared to students who enroll at community college (Bailey et al., 2010; Horn & Nevill, 2006), and motivation interventions have been shown to be most effective for students who are less academically prepared (Harackiewicz et al., 2016; Hulleman et al., 2010, 2017). Therefore, a motivation intervention in a community college may naturally benefit all students, regardless of their demographic background, as all likely have prior experiences that may not have supported their preparation for college work.

Prior motivation intervention research has revealed conflicting results as to whether similar subgroups are differentially impacted by motivation interventions. In regard to gender, although Rozek et al. (2015) found that intervention effects were moderated by gender, other research (e.g., Hong & Lin-Siegler, 2012; Wyss et al., 2012) did not find a statistically significant gender moderation. In regard to ethnicity, Sherman et al. (2013) found that a self-affirmation motivation intervention was useful for improving the grades of Latino students in their intervention, whereas White students remained unaffected from the intervention. On the other hand, Hulleman and Harackiewicz (2009) did not find statistically significant interactions with ethnicity in their utility-value intervention.

Finally, in regard to SES, although we expected low-income students to benefit more from the intervention (Bastedo & Jaquette, 2011; Engberg & Allen, 2011), SES was not a moderating variable, which aligned with previous work (Rozek et al., 2017). One explanation for this finding may be that SES operates on both individual and group levels and can therefore operate on the intervention differently at these levels (Rosenzweig & Wigfield, 2016). Low SES students may have benefited from the intervention because they had not been previously exposed to successful role models in their lives, whereas high SES students may also have benefited from the intervention because they had been exposed to role models in their everyday lives, thereby reinforcing the intervention (Rosenzweig & Wigfield, 2016). Future work should continue to examine for whom and under what conditions different motivational interventions work best.

Limitations and Future Directions

Within this study, we conducted a quasi-experiment using institutional data to investigate how participating in a YouTube role model intervention predicted course grade and subsequent course enrollment in engineering. Although this approach presented great opportunities to study community college students in engineering—a population not typically represented in motivation intervention research—relying on a non-randomized design also presented some limitations. We were not able to randomly assign students to an intervention or control condition. Students opted in to participate in the intervention. We attempted to deal with selection effects by matching students using all available administrative information: course (i.e., term description, engineering course, and instructor), demographic (i.e., women, URM, and low-income), and prior achievement (i.e., high school GPA) characteristics. There, however, still may be other unobserved factors, such as engineering interest and other demographic characteristics (e.g., age, first-generation status, full-time vs. part-time enrollment status, immigration status, whether a student has taken a remedial course or not), that could explain part of this association between the intervention group, course grade, and subsequent course enrollment. For instance, first-generation college students, those who are the first in their families to attend or earn a college degree, might benefit more from role models compared to continuing-generation college students because they lack inherited knowledge about navigating postsecondary institutions from their parents (Plaskett et al., 2018). Similarly, other demographic characteristics not accounted for due to data availability, such as age, enrollment status, immigration status, and remedial course enrollment, could potentially influence engineering students’ subsequent course enrollment and academic performance (Hagy & Staniec, 2002; Illich et al., 2004; Salamonson & Andrew, 2006). Future work should make an effort to measure and include these demographic characteristics as covariates. Further, given the promise of the relatively large effect sizes we found, replicating our work with other community colleges that do not benefit from an NSF-funded transfer partnership, as well as a community college with a different diverse student population, seems to be a fruitful avenue for pursuing continued work on increasing positive outcomes for engineering community college students.

Additionally, although we have given students the choice to select whose YouTube videos they wanted to watch, we were unable to inquire about their reasons for choice. Prior work has found that providing students autonomy in motivational interventions is more effective than those which include fewer choices (Rosenzweig et al., 2019). Researchers should aim to investigate the effectiveness of providing students with choices on their role models and reasons for choosing a particular role model. For instance, future work that varies the level of choice in comparing intervention outcomes or collects qualitative data to understand whether students’ reasoning for choice could have impacted intervention outcomes.

Third, even though our study is one of the first to leverage YouTube videos outside of learning course-specific content, future work is needed to incorporate more of the positive community aspect of social media. For example, increasing the accessibility of role models or shared personal experiences through comments or live streams might further increase community college students in engineering choice and persistence. More work on leveraging social media as a platform for interventions in a transformative way might open greater possibilities of effectively increasing persistence on a large scale for minoritized students.

Finally, this intervention was specifically conducted within a community college context with an established transfer pathway from a community college to a feeder four-year university. This articulation agreement between the two institutions often plays a key role in students successfully transferring to a four-year university and positive post-transfer outcomes, as transfer agreements can be easily assessed and kept up to date (Anderson et al., 2006; Stern, 2016). However, not all community college students in engineering will benefit from these readily accessible and up-to-date transfer agreements, as there is a lack of articulation agreement between some community colleges and four-year universities. Therefore, one limitation of our study is whether our findings would generalize to engineering community college students who attend institutions without clear articulation agreements with four-year universities. Future research can investigate to what extent watching role models through social media platforms, such as YouTube, helps community college students without a clear articulation plan.

Conclusion

We found that community college students in engineering who completed a reflection exercise after watching YouTube videos of recently transferred engineering students like them had higher performance and subsequent engineer course enrollment. Results contribute to the theoretical integration of SEVT (Eccles & Wigfield, 2020) and role model theories (Lockwood & Kunda, 1997; Lockwood et al., 2002) in the intervention setting. Seeing and hearing from a successful minoritized engineering community college graduate might increase the likelihood of another minoritized student’s belief that one can attain an engineering degree too (i.e., expectancies for success) and want to engage in engineering-related tasks (i.e., subjective task values), which ultimately predict subsequent course enrollments (i.e., choice) and course grades (i.e., performance). Our motivation intervention can potentially be used in engineering courses to promote community college students’ subsequent course enrollments and course grades.