Introduction

Feedback is especially critical for personalized learning. To provide students with timely and meaningful feedback, it is important to understand not only what students did or did not learn but also why (Sadler 1989). There are numerous factors that are associated with why students learn or fail to learn. Affective and motivational factors are considered especially important for the traditionally underrepresented and disadvantaged students (National Mathematics Advisory Panel 2008). Students in remedial mathematics courses have already experienced repeated failures during prior mathematics classes. Affective and motivational factors influence their success or lack thereof in subsequent courses (Kim and Hodges 2012). However, these factors are generally overlooked in providing feedback to students and in efforts to promote student success (Schiefele and Csikszentmihalyi 1995). This oversight of affective and motivational factors in providing feedback results primarily from practical limitations (e.g., time constraints) in an instructor’s ability to provide appropriate support, especially in online courses.

The purpose of this paper is to report the development of design guidelines that address students’ affective and motivational factors in order to promote personalized learning and provide meaningful feedback in online remedial mathematics courses. The use of virtual change agents (VCAs; see Fig. 1) in an automated, dynamic, and personalized support system does not place additional time burdens on instructors. The design guidelines provide directions to the design, development, and implementation of VCAs in a personalized, online learning context. The design guidelines also provide support for practitioners and researchers striving to integrate both affective and cognitive learner factors in their efforts to promote personalized learning. This work contributes to a theoretical and practical understanding that recognizes diverse paths to support and facilitate students’ personalized learning.

Fig. 1
figure 1

An example of a virtual change agent

Full size image

College remedial mathematics

Deficits in mathematics knowledge still remain among students entering colleges. One out of every three college students in the United States are required to take remedial mathematics courses (Bettinger and Long 2005). To respond to this pervasive need for mathematics remediation, approximately 70 % of 4- and 2-year colleges offer remedial (developmental) mathematics courses (National Center for Education Statistics 2003). However, few studies have described efforts to improve teaching and learning in college remedial mathematic courses (Dees 1991; Ironsmith et al. 2003; Stern et al. 1996) even though remedial mathematics education is central to the mission of preparing students for the twenty-first century workforce in the United States (NCES 2003; Perin 2004).

Moreover, efforts to improve remedial mathematics education have rarely incorporated or accounted for affective factors related to mathematics learning. Since remedial mathematics courses are often taken by first-year college students (Stern et al. 1996), most enrolled students have already experienced repeated failures during prior mathematics classes in high school. Emotional and motivational problems (e.g., math anxiety and task avoidance) potentially confound and complicate efforts to examine student performance in remedial courses (Ironsmith et al. 2003).

Challenges in online courses

The rapid growth of online courses is a major phenomenon in the higher education. The percentage of students who took at least one online course was approximately 10 % in 2003, 25 % in 2008, and 30 % in 2009 of the overall student population in higher education (Allen and Seaman 2009). Rapid increases in the number of students taking online classes is a phenomenon not only in higher education but also in elementary and secondary education; 55 % of K-12 public school districts report that their students have taken online courses in 2009–2010 (National Center for Education Statistics 2011). However, online courses do not guarantee every student’s success; up to 50 % of enrolled students drop out of their courses (Angelino et al. 2007; Clark 2003). Challenges in online learning include feelings of isolation and a lack of social support (Erichsen and Bolliger 2010; Muilenburg and Berge 2005; Song et al. 2004).

Challenges in online learning can worsen in online remedial mathematics courses. Students experience negative emotions (e.g., shame, anger, frustration) due to their past failure on a college placement test, which required them to take the remedial mathematics courses. Academic emotions are defined as emotional experiences that stem from achievement-related ongoing activities (e.g., anxiety during an exam) or the outcomes of the activities (e.g., shame linked to an exam failure) (Pekrun et al. 2007). Research has shown that emotions impact the process and outcome of learning and performance (Gläser-Zikuda et al. 2005; Goetz et al. 2006; Pekrun 2006; Pekrun et al. 2002). Cultivating positive emotions and reducing negative emotions can benefit learning processes and outcomes (Astleitner 2000; Fredrickson 2001; Goetz et al. 2006; Pekrun 2006). The need to include individual and personalized support for positive emotional experiences in designing instruction is the central concern in the design guidelines developed herein for personalized learning.

Optimizing emotional experience: Emotion regulation

Emotional experiences for learning can be optimized through emotional scaffolding (Rosiek 2003; Rosiek and Beghetto 2009). Emotional scaffolding refers to pedagogical activities that are tailored to specific aspects of the content of teaching as well as to emotional experience of students in the classroom (Rosiek 2003). Emotional scaffolding requires teachers’ knowledge of interactions among “curricular content, cultural discourses, community histories, students’ personal histories, and general attitudes about schooling that precipitated students’ emotional response to their lessons” (Rosiek 2003, p. 406). For example, to provide effective emotional scaffolding, a teacher should understand that uncertainty embedded in an assignment (even when purposefully planned for discovery learning) could trigger frustration on the part of some students. These students may interpret uncertainty as their inability to complete the assignment due to their personal academic histories, cultural background, and so forth.

When face-to-face interactions are not available or convenient, for example, in online teaching and learning environments, agent-based emotional scaffolding has been used to promote positive emotions and reduce negative emotions using affect-aware tutors, relational agents, and embodied conversational agents (Bickmore and Cassell 2004; Campbell and Green 2009; Woolf et al. 2009). For example, an affect-support agent that communicated a sense of empathy to frustrated users helped them persist in using the computer program that had caused their frustration (Klein et al. 2002). However, the frustration was pre-planned by designers who embedded glitches in the computer program. This means that when and why the users would be frustrated was already known when the agent’s emotional scaffolding for users was programmed. In contrast, in personalized, remedial math learning environments, students’ emotional experiences are not totally predictable. It is difficult for agents to be as flexible as human teachers in providing just-in-time, individualized support corresponding to students’ emotional experience in such environments. This is in part due to the limited capacity to “fully elicit, communicate, measure and respond to students’ affect” (Woolf et al. 2009, p. 158).

Thus, students need to be supported in the process of regulating their own emotions. Emotion regulation is the process by which individuals change themselves and their environments in order to influence their emotional experience (Gross 2008; Schutz and Davis 2010). For example, Schutz and Davis (2010) examined the emotion regulation processes of students during test-taking. One example of emotion regulation (i.e., regaining a task-focusing process) would be a process in which students regulate themselves to focus on the test by avoiding wishful thinking or self-blaming (Schutz and Davis 2010). Not all students regulate their emotions while struggling with difficult tasks. Some dwell on self-blame, for instance, along with distracting emotions such as hopelessness, anger, and shame.

Strategies for the promotion of emotion regulation in online environments have rarely been studied in the field of instructional design. Many researchers have stated that theory-driven interventions should be designed and implemented to cultivate positive emotions (e.g., Astleitner 2000; Gläser-Zikuda et al. 2005; Goetz et al. 2006; Kim and Hodges 2012; Pekrun 2006). Several theory-based models and approaches describe how emotions occur and are regulated (Kim and Pekrun 2013). However, there has been little research on theoretically-guided instructional design that aims to promote learners’ emotion regulation (Kim and Hodges 2012).

Reciprocal impact of emotions and motivation

An effort to promote students’ emotion regulation alone may not help optimize emotional experiences (Kim and Hodges 2012). As illustrated in Fig. 2, emotions and motivation impact each other to act in a certain way (Op ‘t Eynde et al. 2006; Pekrun 1992, 2006). Student motivation can be defined as the desire to pursue learning activities (James 1890; Pintrich and Schunk 2002). Actions result from different emotions that form motives (Frijda et al. 1989; Plutchik 1980; Roseman et al. 1994). For example, a student’s absence in midterm exam (action) induced by the feeling of hopelessness (emotion) is to give up the course completion (motivation). Likewise, actions result from different motives that create emotions. Studying to prepare for the midterm exam (action) induced by the desire to complete the course (motivation) creates anxiety (emotion).

Fig. 2
figure 2

Reciprocal impact of emotions and motivation

Full size image

Emotions and motivation can bring about the same action but the two constructs are not the same. For instance, experiencing either negative emotions or a lack of motivation to study can lead to course dropout. This does not mean that feeling negative emotions and lacking motivation are the same but does mean that the two are the source of the same problem. It would not resolve the problem if only part of the source (e.g., negative emotions) is ameliorated. This is why both emotions and motivation need to be understood and optimized to promote student learning and performance. Buck’s (1985) analogy of matter and energy confirms the need for an integrative view of emotions and motivation: “Just as energy is a potential that manifests itself in matter, motivation is a potential that manifests itself in emotion. Thus motivation and emotion are seen to be two sides of the same coin, two aspects of the same process” (p. 396).

Recent research has shown close relationships between emotions and motivation in online learning environments. Negative emotions such as anxiety, hopelessness and boredom were interrelated with a lack of or discontinued motivation (Alexander and Onwuegbuzie 2007; Artino 2009; Haycock et al. 1998) and positive emotions such as pride and enjoyment were observed with high level of motivation (Kim and Hodges 2012). The reciprocal impacts of emotions and motivation need to be considered for online learning in the context of promoting either emotional experience or motivation.

In fact, both student motivation and emotions are critical to success in college mathematics courses (Schiefele and Csikszentmihalyi 1995). Limited motivation can make mathematics especially challenging for students who do not plan on majoring in science, technology, engineering, and mathematics (STEM) fields (Kim and Keller 2010). Such students reluctantly enroll in remedial mathematics courses due to college requirements, but few embrace mathematical concepts and processes (Hagedorn et al. 1999; Stage and Kloosterman 1995). Such students often demonstrate a lack of engagement and negative emotions (Ironsmith et al. 2003).

There is a need to support students’ emotion regulation while improving motivation in remedial mathematics education. This need is particularly acute for online remedial courses where the course growth is most apparent and student needs are greatest. Besides, the lack of personal contact between students and instructors as well as among students may limit the individual attention required to help students overcome specific motivation and emotion problems.

Virtual change agents (VCAs)

Time and curriculum constraints often limit an instructor’s ability to address student needs beyond mathematical concepts and processes. VCAs, which can be easily integrated into online mathematics courses, represent a potential solution. VCAs (see Fig. 1) are three-dimensional, human-like, animated characters designed to facilitate positive changes in learners in terms of academic emotions, motivation, beliefs, and so forth (Kim 2012; Kim and Baylor 2008; Kim et al. 2007). The concept of VCAs is based on research on pedagogical agents (Azevedo and Hadwin 2005; Baylor 2002; Biswas et al. 2005; Chase et al. 2009) and on change agents (Fullan and Stiegelbauer 1991; Rogers 2003). A change agent is an individual who facilitates his or her group’s acceptance and implementation of a new thing. In this context, strategies for motivation and emotion regulation would be the new thing for students to accept and implement. VCAs have characteristics of pedagogical agents that respond to a learner through human-like interactions (Kim and Baylor 2006b) along with the characteristics of a persuader who change others to adopt new things (Rogers 2003) such as emotion regulation strategies. The term, change agents, used in the diffusion of innovation literature refers to human agents. The uniqueness of VCAs is that the persuasion and social influence of change agents are enabled in online environments without the presence of a real human by using the technology of pedagogical agents that are shown to have social presence (Kim and Baylor 2006a).

Thus, VCAs can be advantageous not only to overcome the problem of limited instructor time or inclination, but also to help students to adopt strategies for emotion regulation and improve motivation. In this paper, VCAs are said to promote students’ success by improving their emotion regulation and motivation in online remedial math online courses without increasing demands on instructors (see Fig. 3).

Fig. 3
figure 3

Support for emotions and motivation through virtual change agents

Full size image

Design guidelines for virtual change agents

VCAs support personalized learning by responding to student needs with regard to academic emotions and motivation and encouraging them to implement emotion regulation and motivational strategies. The accessibility of VCAs can be greater than that of human instructors because of their computer-based nature (Kim and Baylor 2008) that enables responsiveness to accumulated records of an individual student’s past activities as well as current activities, which makes the learning environment personalized. The fact that agents can be customized to support particular needs of students is one critical aspect of the utilization of agents (Baylor 2011). Such customizability enables VCAs to provide personalized support to individual students (Lee and Park 2007; Shute and Zapata-Rivera 2007).

However, VCAs do not provide course content that is adapted for students’ needs, as in typical adaptive systems (e.g., Papanikolaou et al. 2003; Park and Lee 1996). Rather, VCAs aim to help individual students acquire the skills (e.g., emotion regulation skills) that are necessary to overcome difficulties in courses. As Kinshuk et al. (2009) argued, it is important that students have the opportunity to be trained to cope with courses that are not aligned with their preferences. The availability and customizability of VCAs are utilized to maximize the benefits of personalized learning environments for this purpose.

In the design guidelines for VCAs to enhance personalized learning environments, how the agents look is not as important as what they say and when they say. Many studies on agents have focused on visual aspects of the agent such as gender and general appearance, rather than how the agent interacts in response to specific students (e.g., Baylor 2009). For example, there is a study reporting that the effects of an agent were the greatest when the look of the agent was perceived as the most similar to the study participants (Rosenberg-Kima et al. 2008). Some researchers highlight that the visual and verbal communications of an agent should be carefully blended in order for the role of agents to be performed as planned (Lester et al. 1999). In brief, attractiveness, gender, race, and age that determine the agent appearance as well as message delivery methods that include voice and gesture have been often the focus in the context of examining the effects of agents (Baylor 2011). However, despite the importance of aesthetic, animation, and audio components for the agent design, still, what students do with agents should be of primary concern. For instance, the learning-by-teaching, protégé, effect occurred when students taught their agents (i.e., teachable agents in Chase et al. 2009). What determines the functionality and effects of agents is what agents offer and how students interact with agents rather than how attractive agents look. Therefore, two design components form the focus of concern in this paper as described in the following sections: (1) the strategies that VCAs convey; and (2) interactions between VCAs and students. Guidelines are stated for each of the design components.

Design Guideline 1: Design VCAs to convey strategies that facilitate students’ reappraisal of the situation

  • Corollary 1.1: Design VCAs to convey strategies that promote students’ perception of task value.

  • Corollary 1.2: Design VCAs to convey strategies that promote students’ perception of controllability.

Theoretical foundations

These guidelines are based on a conceptual framework that is grounded in integrative knowledge from previous research as well as theories on motivation and emotions such as appraisal theories (Ellsworth and Scherer 2003; Scherer 1999; Schutz and Davis 2000), attribution theory (Weiner 1985), control and value theory (Pekrun 2006), and an emotion regulation model (Gross 2008). The strategies were constructed to facilitate students’ positive reappraisal of a course activity or its outcome (e.g., assignments in an online math course) by maximizing their controllability of the four emotion process components described by Gross (2008): (a) situation, (b) attention, (c) appraisal, and (d) responses. The strategies also include ones increasing students’ perceived task value (see Fig. 4). This conceptual framework and specific strategies are elaborated in the following sections.

Fig. 4
figure 4

A conceptual framework for creating strategies that VCAs convey

Full size image

Emotions arise when a person appraises a given situation. The situation is evaluated in terms of its meaning and causal structures and the person’s controllability over the situation (Gross 2008; Pekrun 2006; Scherer 1999; Schutz and Davis 2000; Weiner 1985). This cognitive appraisal process is not always conscious or deliberate; it can be non-conscious and automatic (Gross 2008; Johnson-Laird and Mancini 2006; Op ‘t Eynde et al. 2006; Pekrun 2006; Schutz and Davis 2000, 2010). The aim of using VCAs for personalized learning is to empower students’ conscious and deliberate appraisal processes which can ultimately become non-conscious and automatic for effective, engaging learning.

Meaning structure analysis

A cognitive appraisal process is initiated by analyzing the meaning structure of a given situation (e.g., an assignment or an exam score). Extrinsic value and intrinsic value of the situation are evaluated in the meaning structure analysis (Carver and Scheier 1990; Pekrun 2006; Schutz and Davis 2000). For instance, the extrinsic value of a remedial mathematics course is appreciated if students see that the completion of the course leads to a step closer toward graduation (instrumental usefulness). The intrinsic value is appreciated if math is perceived as a fun subject.

Causal structure analysis

In addition to the analysis of the meaning structure of a given situation, analyzing the causal structure of the situation provokes emotional responses (Pekrun 2006; Weiner 1985). The causal structure analysis forms students’ perceptions of the locus of control (internal vs. external) and the stability of control (stable vs. unstable) (Weiner 1985) and such perceptions bring about emotional reactions. For example, students would not experience the emotion of pride if they think they passed an exam because, due to luck, they got easy questions. At the same time, their motivation to study for the next exam would not be apparent due to external and unstable control, luck.

Controllability

Not only students’ analyses of the meaning and causal structures of a given situation but also their controllability over actions is also critical in the appraisal process (Pekrun 2006; Weiner 1985). Even if students think that their ability, not luck, is the key to success in a remedial mathematics course, unless they perceive controllability of actions (e.g., exerting effort) to sustain or improve their ability, unconstructive emotions (e.g., hopelessness) can be experienced.

In summary, the cognitive appraisal process that forms emotional responses and motivation should be considered for emotion regulation and enhanced motivation in a way to help students perceive the value of a situation as well as their controllability.

Practical applications

To fulfil Design Guideline 1, two corollaries are proposed. For Corollary 1.1 (Design VCAs to convey strategies that promote students’ perception of task value), both intrinsic value and extrinsic value should be addressed in strategies as described in the following:

  • intrinsic value: capturing the interest of students; stimulating curiosity to learn; and

  • extrinsic value: meeting students’ needs/goals to make learning experience be instrumentally useful for them.

Specifically, to help students perceive intrinsic value of tasks, the strategies are tailored to students’ interest. As described in Keller’s (2009) motivational design model, making tasks relevant to students’ needs such as interest is critical in motivating students. Also, interest influences learning processes and outcomes (Hidi and Harackiewicz 2000; Wade 2001) by facilitating engagement (Hidi and Baird 1988; Krapp 1999) and constructive emotions (e.g., less boredom and more enjoyment) (Ainley et al. 2002; Harp and Mayer 1997). As illustrated in Fig. 5 and Table 1, a sample strategy to promote intrinsic value would be to provide students with choices to see how algebra can be used. The choices (i.e., motorcycle, weight loss, smart phones) can connect to individual students’ interests, provoke their curiosity, and encourage them to think that mathematics does not have to be always boring. To facilitate students’ perception of extrinsic value of tasks, the strategies focus on instrumental usefulness. For example, as shown in Table 1, the strategies include a reminder of school requirements for the completion of the remedial mathematics course and students’ goals requiring college graduation.

Fig. 5
figure 5

A screenshot of a VCA aiming to promote students’ perception of task value

Full size image
Table 1 Sample strategies
Full size table

For Corollary 1.2 (Design VCAs to convey strategies that promote students’ perception of controllability), appraisal-oriented regulation and emotion-oriented regulation (Pekrun 2006) should be considered in the development of the strategies for controllability. Appraisal-oriented regulation strategies are designed to facilitate students’ deliberate efforts (a) to become aware of the situation and their emotions as well as (b) to conduct conscious appraisals and reappraisals (e.g., recognize the feeling of boredom when studying online materials). Emotion-oriented regulation strategies are designed to help students provoke or suppress emotional responses (e.g., provoke enjoyment, suppress anger). Both strategy types are meant to increase students’ perception of their controllability of the situation (i.e., learning processes and outcomes). Specifically, four strategies used before the activation of certain emotions and one response-focused strategy used during the activation of certain emotions (Gross 2008) were applied to the construction of the strategies that VCAs convey, as described in the following:

  • Situation-selection—choosing to be in environments that are likely to diminish negative emotions;

  • Situation modification—changing a certain environment to reduce negative emotions;

  • Attentional deployment—shifting attention to something else;

  • Cognitive change—cognitively re-evaluating the situation; and

  • Response modulation—suppressing certain emotions activated.

As illustrated in Table 1, when students feel discouraged and hopeless due to overwhelming work to do for the course, VCAs emphasize that such feelings are natural. VCAs recommend the adoption of a situation selection strategy, in which students go to a math tutoring lab or attend an online tutoring session only to see that many people are there because they experience difficulties as well. Accepting negative emotions as normal facilitates their awareness of their current emotions, which is effective in regulating regulation (Seo and Barrett 2007). Another example strategy is a situation modification strategy that helps students modify debilitating anxiety to facilitative anxiety when they prepare for an exam. An attention deployment strategy is to help students shift their attention to something else that evokes more pleasant or less unpleasant emotions; for example, VCAs recommend that students be attentive to “tasks to do” rather than “the exam to take” and “questions were on an exam” rather than “scores from the exam.” A cognitive change strategy is to facilitate students’ cognitive re-evaluation of the situation (without selecting a different situation or modifying the situation). For example, when the course materials are too boring, students try to figure out what of the materials do not bore others. A response modulation strategy is to help students to suppress their negative emotions that are already occurring by cognitively blocking them from their current thoughts (e.g., locking worries in a mental box).

Design Guideline 2: Design VCAs to promote students’ interactions with VCAs

  • Corollary 2.1: Design VCAs to assess and respond to students’ needs.

  • Corollary 2.2: Design VCAs to illustrate scenarios where strategies are embedded.

Theoretical foundations

According to the theory of “Computers Are Social Actors” (CASA), human-to-computer interactions are comprised of social presence and responses of both computers and users (Reeves and Nass 1996). Research has shown that users treat and respond to computers in the manner that they would do to people. For example, users perceived computers as social actors when they read error messages on websites (Tzeng 2006), computer cartoon character and synthetic speech (Lee 2008) and computer icons (Hall and Henningsen 2008) and when they work with online tutorial and evaluation feedback (Karr-Wisniewski and Prietula 2010) and interactive mathematics learning programs (Tung and Deng 2006).

It is noteworthy that interactivity of computers enables users’ perception of CASA. Interactivity is critical in differentiating computers from other objects (although phones have interactivity nowadays—they are now smartphones, which seems to imply that users perceive phones as social actors who are smart). For instance, when users click on an online tutorial menu, how the tutorial website responds to the action of clicking the particular menu determines how useful, friendly, polite, reasonable, informative, and credible users perceive the online tutorial is. And this perception influences users’ willingness to continue their interactions with the tutorial and learn with it. A computer’s constant responsiveness to users’ needs allows the computer’s interactivity and forms and maintains interactions between the computer and users, which ultimately influences how effectively the information (or something else) conveyed by the computer is received by users.

Along these lines, the level of interactivity of VCAs (i.e., VCAs’ responsiveness to students’ needs) should be carefully considered during the design of VCAs. Interactions between VCAs and students make students think what VCAs offer them to enhance their emotion regulation and motivation is credible and important. The interactivity can be achieved by VCAs’ understanding of students’ needs and presenting strategies that are entailed to students’ needs for emotion regulation and enhanced motivation. Students’ needs can be assessed by asking them about their motivation and emotions and the strategies can be presented in the scenarios that students can relate to.

Practical applications

To fulfil Design Guideline 2, two corollaries are proposed. Corollary 2.1 (Design VCAs to assess and respond to students’ needs) holds that VCAs should assess what students think and how they feel about specific contexts. For example, a VCA can ask, “How relevant do you think this math class is to you?” to assess students’ perception of the task value. The VCA can then ask, “Are you getting anxious before the upcoming exam?” to assess students’ emotional statuses. Questions to determine how to proceed with conversations can facilitate the interactions between VCAs and students as well. For example, a VCA can say to a student, “Last semester, I found interesting uses of math in everyday life. For example, some people use algebra for dieting, exercising, cooking, and even for shopping. Would you like to hear how?” If a student clicks “Yes, I would like to hear how” and selects “Algebra for cooking,” then the VCA talks about the relevant example to help the student perceive the task value in learning algebra. Based on students’ responses, personalized strategies are selected dynamically and automatically from a database.

As an extended example, Alex, a student of an online remedial mathematic course, logs into the course website to study the week’s course materials. He works on the practice midterm. A VCA appears when 50 % of his responses to the practice test are still inaccurate after his second practice. The VCA asks Alex questions about his readiness for the exam and any difficulties such as emotional experiences. Alex clicks on a choice associated with each question and a pre-programmed voice that corresponds to Alex’s needs is played with additional information. A simplified illustration of such interactions between a VCA and a student is presented in Fig. 6.

Fig. 6
figure 6

A sample flowchart for interactions between a VCA and a student

Full size image

For Corollary 2.2 (Design VCAs to illustrate scenarios where strategies are embedded), VCAs incorporate scenarios to convey strategies. The VCAs themselves are students who have had to take the remedial mathematics course. They share their stories with learners about overcoming emotional and motivational problems in the course. For example, when a VCA responds to a student’s negative perception of values in learning math (described above), the VCA presents a scenario like the following, “When I was taking this course last semester, math did not seem important or relevant to me either. I couldn’t understand why I had to take it. I had never planned to major in math and I still don’t. I had never thought I would need to learn math when I began the course. I could not see how it could possibly be relevant to my future studies or life.” Then, the VCA shows what changed over the semester.

Using scenarios of VCAs as peers is to facilitate social interactions between VCAs and students grounded in the concept of agents as social models (Kim and Baylor 2006a). When students can relate to VCAs as peers, they are more likely to follow the model of VCAs for their own actions (e.g., implement the same emotion regulation strategy that the VCAs say that they used in studying for the class). VCAs are designed to persuade students that the strategies are effective by enabling their vicarious experience and suggesting methods for implementing the strategies (Kim and Baylor 2008).

How to validate the effectiveness of design guidelines for VCAs

The design guidelines have not been fully validated. Only some have been tested in empirical studies. This does not validate the effectiveness of the design guidelines as a whole. The design guidelines need to be implemented in an integrative manner. Only the strategies that promote students’ perception of controllability (i.e., part of Design Guideline 1) were implemented in previous studies (Kim and Bennekin 2010, 2011; Kim and Hodges 2012). The limited effects of the interventions in those studies led to the inclusion of strategies promoting students’ perception of task value (i.e., Design Guideline Corollary 1.1), also based on theoretical foundations that explain reciprocal impacts of emotions and motivation and both should be considered for students’ reappraisal of the situation. In contrast, strategies with regard to emotion regulation implemented in some previous studies (Kim and Keller 2008, 2010, 2011) did not aim to increase students’ perception of controllability. Also, the interactions between the interventions and students were limited in these studies because personalized emails were used as delivery methods although strategies were personalized according to students’ needs throughout the use of diagnostic questions. Prior studies are listed in Table 2. In brief, as learned from the previous studies as well as the synthesis of theoretical foundations, the integrated use of all the strategies and interactions proposed in the design guidelines should be tested for the purpose of improving students’ emotional and motivational experiences. An example of a comparison group in validation studies would be a control group without scenarios having VCAs as peers, which examines if the whole set of the design guidelines works as proposed or if some components should be revised.

It would be ideal if the validation studies are conducted in online remedial mathematics courses for two reasons. First, the design guidelines have been developed for personalized learning of students with a failure experience (e.g., not passing the mathematics placement test) who do not intend to major mathematics but are required to take mathematics courses for remediation. Second, VCAs are intended to help students in online environments where there lacks social support for affective and motivational problems. However, these guidelines may be applied to contexts with similar student needs to those in online remedial mathematics learning. Prior studies have taken place in college large lecture general education courses with similar affective and motivational challenges. Nonetheless, if validation studies are conducted in online remedial mathematics course, further math-specific guidelines could be added to the design guidelines (e.g., for numerical anxiety, math test anxiety, and abstraction anxiety; Ferguson 1986). The guidelines could also be made more prescriptive in nature.

A multi-method approach (Meyer and Turner 2002) should be used to examine the effectiveness of the design guidelines. Observations and analyses of students’ online interactions and conversations on discussion boards, via email, and synchronous chats can provide information related to the effects of VCAs. Also, surveys and interviews on students’ perceptions of VCAs (including their thought about the strategies and interactions) should be examined as well. In addition, longitudinal studies should be done to validate the design guidelines. Even if positive changes in students’ motivation and emotion regulation are found, it should be investigated if such changes persisted over time and how they impact learning so that improvements can be done in the design guidelines accordingly.

Table 2 A summary of previous studies on emotion regulation and motivation
Full size table

Discussion

Summary

In this paper, design guidelines for VCAs were proposed to address students’ affective and motivational factors to promote personalized learning in online remedial mathematics courses. Automated, dynamic, and personalized support was promoted by maximizing interactions between VCAs and individual students. The strategies that VCAs conveyed were constructed to support emotion regulation and motivation based on theories and prior research on emotions and motivation. The availability and customizability of VCAs enabled the strategies to be provided in real-time and customized for individual students.

An integrative approach that has been used to construct the design guidelines for VCAs includes the following emphases. First, addressing only student motivation would not be sufficient in learning contexts where there are many challenges like remedial math online courses. Emotions should be considered as well as motivation. Second, understanding the common factors that influence both emotions and motivation (e.g., task value, controllability, social interactions) are important in designing interventions that promote student success in such challenging courses. Third, VCAs can be designed in a way to address such factors but especially to promote students’ emotion regulation, which would be more practical than to program to address every single possible problem that students can encounter.

The design guidelines are meaningful in that they guide the creation of interventions to promote both motivation and emotion regulation unlike other existing models and theories. For example, Keller’s (2010) ARCS (the acronym of attention, relevance, confidence, and satisfaction) motivational design model focuses on motivation; Astleitner’s (2000) FEASP (the acronym of fear, envy, anger, sympathy, and pleasure) approach focuses on emotions but not on motivation or emotion regulation. Deci and Ryan’s (2000) self-determination theory focuses on motivation and does not provide explicit guidelines for how to design interventions to promote autonomy, competence, and relatedness. There are some models for emotion regulation (e.g., Gross 2008) and research on emotion regulation (e.g., Schutz and Davis 2010). However, these do not provide guidelines for how to design interventions for emotion regulation in the contexts of online learning. Pekrun’s (2006) control-value theory of achievement emotions provides a comprehensive picture of both emotions and motivation but its main purpose is not to provide guidelines for the design of interventions for online remedial mathematics learning.

Limitations and suggestions for future research

Further research should continue to validate the design guidelines. Also, as the interaction between VCAs and individual students is what makes the functionality of VCAs effective, a rich database is required to permit meaningful interactions with more customized strategies for individual needs. Within the current design of VCAs, interaction proceeds with VCAs’ questioning and students’ answering, which solely depends on students’ self-assessment (e.g., VCAs reacts based on students’ responses to VCAs’ questions). The assessment occurs in real-time, which responds to some researchers’ criticism of the lack of real-time assessment in emotion research (Ainley 2006; Kim and Hodges 2012; Pekrun 2006; Schutz and Davis 2010). However, the assessment still relies on self-report data. A model predicting students’ affective and motivational experiences based on prior-knowledge, prior failure/success experience, and/or some other variables may need to be developed. The predictive model can be utilized to automatically require students with certain dispositions to go through particular strategies that have been identified as necessary for them to cope with predicted difficulties in class. In addition, advanced technologies such as psycho-physiological sensing systems to detect affective states by analysing data of students’ mouse-click behaviors, facial movement, skin responses, etc. (e.g., D’Mello and Graesser 2010; El Kaliouby et al. 2006; Berleson 2011; Picard 2003; Scheirer et al. 2002) can be used along with self-assessment questions in the future studies.

The design guidelines presented in this paper maximize the possibility of studying and supporting students’ affective and motivational factors within the learning and performance contexts. Still, issues pertaining to the situatedness of emotions, according to which emotions are dynamically linked in a specific, social-cultural context, are not fully resolved. For example, it is not clear how VCAs could be used in test-taking contexts, although it is clear that negative emotions can adversely affect students taking tests. Personal, interpersonal, community-level, interactive, socially-situated appraisal processes (Op ‘t Eynde and Turner 2006) need to be examined when considering the situatedness of emotions.

Implications

The design guidelines provide guidance for the design, development, and implementation of VCAs in a personalized, online learning context. Further, they can serve as a foundation for creating theory-based interventions and evaluation programs using a variety of technologies for emotion regulation and motivation enhancement. The guidelines also provide support for practitioners and researchers striving to integrate both affective and cognitive learner factors in their efforts to promote personalized learning. Moreover, since all the data from students’ interactions with VCAs are retrievable, VCAs can be used as “a tool that identifies when students have difficulties in learning” (Kinshuk et al. 2009, p. 751). This work presented here contributes to a theoretical and practical understanding that does not solely rely on knowledge tests but, rather, recognizes diverse paths to support and facilitate students’ personalized learning.