Cite this article: Yulong BIAN, Chao ZHOU. Motivation effect of animated pedagogical agent's personality and feedback strategy types on learning in virtual training environment. Virtual Reality & Intelligent Hardware, 2022, 4(2): 153-172 DOI:10.1016/j.vrih.2021.11.001

Motivation effect of animated pedagogical agent's personality and feedback strategy types on learning in virtual training environment

Yulong BIAN ^1,² , Chao ZHOU ³

1. School of Mechanical, Electrical and Information Engineering, Shandong University, Weihai 264209, China

2. Engineering Research Center of Digital Media Technology, MOE, Shandong University, Jinan 250100, China

3. Tsinghua University, Beijing 100084, China

Abstract

Background

The personality and feedback of an animated pedagogical agent (APA) are vital social-emotional features that render the agent perceptually believable. Their effects on learning during virtual training need to be examined.

Methods

In this paper, an explanation model is proposed to clarify the underlying mechanism of how these two features affect learners. Two studies were conducted to investigate the model. In Study 1, the effect of the APA's personality type and feedback strategy on flow experience and performance was reexamined, revealing significant effects of the feedback strategy on flow and performance and a marginally significant effect of the personality type on performance. To explore the mechanism behind these effects, a theoretical model is proposed by distinguishing between intrinsic and extrinsic motivation effects. In Study 2, the model was evaluated, and the APA's personality type was found to significantly influence the factors in the path of the extrinsic motivation effect rather than those in the path of the intrinsic motivation effect.

Results

In contrast, the feedback strategy affected factors in the path of the intrinsic motivation effect.

Conclusions

These results validated the proposed model. Further distinguishing the two motivation effects is necessary to understand the respective effects of an APA's personality and feedback features on learning experiences and outcomes.

Keywords： Human-centered computing ; Visualization ; Visualization design and evaluation methods

Content

1 Introduction

Recently, an increased interest in the potential use of animated pedagogical agents (APAs) in virtual learning environments has been shown in several studies. APAs are animated, lifelike characters designed to enhance learning and motivation by simulating social interactions with a learner^[1]. The presence of an APA in a multimedia environment may lead to positive learning effects, and this is known as the "persona effect"^[2,3]. APAs can provide learners with instructional information^[4] and evoke social and emotional responses using verbal and nonverbal forms through human-computer interactions^[1]. However, different APA properties may have distinct effects on students' learning experiences and performance. The personality and emotional feedback of an APA are vital social-emotional features that render the agent perceptually believable. In several previous studies, the effects of the personality and emotional feedback of an APA on learners have been separately investigated^[2,5,6]. However, in only a few studies, the effects of both features on learning have been simultaneously examined and the mechanism behind how different APA properties lead to certain consequences has been evaluated.

In a recent study^[7], the potential effects of APAs' personality and emotional feedback on students' learning experience and performance during virtual tai chi training were systematically examined. Both factors were found to significantly affect learning outcomes, but the effects were vastly different. The emotional feedback of an APA influenced both learning experience and performance, whereas the personality of an APA influenced learning performance. Compared with a phlegmatic APA, a choleric APA led to better performance. However, phlegmatic APAs were helpful in promoting the effect of emotional feedback on learning satisfaction. Therefore, these two factors may influence learning in different ways; however, this hypothesis has not been tested, and theoretical exploration is still scarce. According to previous studies on self-determination theory (SDT), motivation may play a crucial role in understanding these effects. The theoretical explanation of the motivation effect is a framework for explaining the persona effect. This has been explored in only a few studies; thus, additional research should be conducted to clarify how APA properties can lead to different effects on learners.

This study makes the following contributions:

(1) A theoretical model is proposed to clarify how APAs' personality and feedback types affect users' flow experience and performance through paths of different motivation effects.

(2) Two empirical studies are conducted, and the APA's feedback strategy is found to affect learners through the intrinsic motivation effect, whereas the APA's personality type affects learners through the extrinsic motivation effect.

Further, this study provides theoretical and empirical references to optimize the design of APAs for virtual reality (VR) training/learning environments.

2 Related work

2.1　APA's personality and feedback

Personality and feedback are key social-emotional features that make an APA perceptually believable. Agent personality has been concerned in many studies in the human-agent interaction field^[8‒14]. In numerous studies, merely the ways for assigning personality to agents have been described, instead of investigating how users were affected by the APA's personality^[8‒11]. In these studies^[8‒11], the design of APAs' personalities was based on personality theories and psychological knowledge about the links between physical appearance, behavior, and personality judgment. For example, the personality theories of the five-factor model (the Big Five) and Eysenck's trait model have often been used as the theoretical basis for designing agents' personalities^[9,12]. However, the effect of an agent's personality on user experience has been investigated in only a few studies^[13,14]. Dekker evaluated the effect of an agent's personality on user experience instead of personality creation and found that an agent's personality enhanced some emotional and motivational aspects of learning, but these effects were not deeply investigated^[13]; thus, it was difficult to draw strong conclusions.

More research has been conducted on APAs' feedback than on personality. In several previous studies^[5,2,6,15], the implementation of APAs' various feedback was investigated and their effect on learners' learning experience, behavioral intention, outcomes, etc. was evaluated. Various types of feedback can be designed for the agents. According to a previous study^[6], feedback classification is connected to some of the following characteristics: (1) "when" feedback is triggered (in advance, immediate, and delayed); (2) the context of the feedback is targeted at one of the three different mind dimensions (cognitive, emotional, and conative); (3) the effect and outcome that a feedback will produce (reward/positive, neutral, or punishment/negative); (4) the multimedia type of the feedback (text, pictures, audio, video, etc.) could serve as a distinguishing feature; (5) other feedback characteristics include frequency, duration, interactivity, personalization, and educational context.

It is clear that APAs' feedback influence learners' emotional, motivational, and behavioral aspects of learning. Praise for correct answers can enhance the sense of competence, self-control, self-efficacy, and curiosity, thus enhancing intrinsic motivation^[16]. However, whether it affects learners in the same way as the APA's personality is unclear.

In a recent study^[7], the effect of an APA's personality type and feedback strategy on students' learning experiences and learning performance in tai chi training was evaluated. To the best of the authors' knowledge, that was the only study in which the effects of these two features on learning were simultaneously examined. In that study, four APAs with two distinct personality types (choleric and phlegmatic) and two distinct feedback strategies (positive and negative) were created. Subsequently, the potential effect of these two factors in a virtual tai chi training studio was investigated; the APAs' feedback strategies were found to significantly influence students' learning experience and performance. A positive feedback strategy led to more favorable learning experiences and performance compared with the negative feedback strategy. Moreover, although the APA's personality type significantly affected learning performance, it had no considerable influence on learning experience. Therefore, the APA's personality and feedback strategy may influence learners in different ways.

2.2　Studies on motivation effect

To clarify how an agent's personality and feedback affect training experience and performance, the previous results must be further discussed and explained theoretically. The theoretical explanation of the motivation effect is a possible perspective for explaining the persona effect^[15,17,18]. Some studies^[19-21] have revealed that the use of APAs can enhance students' motivation. Chen and Chou noted that motivation improved in computer-assisted learning systems in which an APA was used, and the reactions of the APA contributed to this trend^[19]. Both feedback and personality influenced learning through motivation-related factors. Specifically, learning was significantly enhanced when the APA displayed positive emotions compared with negative emotions^[1]. Students' attitudes and motivation were found to improve when an APA provided positive feedback^[20]. Emotional and motivational aspects of learning were found to be significantly enhanced when the APA exhibited personality (i.e., extraversion type) that complemented the learner's own^[21]. Bian et al. also reported that the motivation effect is crucial for understanding the influence of an APA's properties, namely, personality and feedback strategy, on learning^[7].

Although the role of motivation in explaining the persona effect has been highlighted in certain studies, different motivation types have not been considered. According to SDT, intrinsic and extrinsic motivations are distinct. Intrinsic motivation refers to doing something because it is inherently interesting or enjoyable, rather than for some other consequence. When intrinsically motivated, a person is moved to act for the fun or challenge rather than because of external prods, pressures, or rewards. Extrinsic motivation is a construct that pertains to an activity completed to attain some separable outcome and instrumental value^[22]. It is defined as the desire to engage in behavior for external reasons^[23]. The properties of an APA can influence its motivation antecedents. Some of these factors contribute to intrinsic motivation, and some contribute to extrinsic motivation, which may lead to different learning outcomes. Over three decades of research^[24] has shown that the quality of experience and performance can drastically differ when an individual behaves for intrinsic versus extrinsic reasons. For example, Dinçer and Doganay examined the effects of multiple APAs on learners' academic motivation and noted that intrinsic and extrinsic motivation should be distinguished^[25]; however, they focused only on intrinsic motivation, whereas the present study addresses both. Therefore, the distinction between the paths of the intrinsic effect and extrinsic motivation effect (based on SDT) may be crucial for clarifying the persona effect produced by the APA's personality and emotional feedback.

2.3　SDT

SDT distinguishes between different types of motivation based on the varied reasons or goals that prompt an action. Intrinsic motivation refers to doing something because it is inherently interesting or enjoyable, whereas extrinsic motivation refers to doing something because it leads to a separate outcome^[22].

Cognitive evaluation theory (CET) and organismic integration theory (OIT), two sub-theories of SDT, explain intrinsic and extrinsic motivation in detail^[24]. CET describes the conditions that elicit and sustain intrinsic motivation and argues that interpersonal events and structures (e.g., feedback and rewards) conducive to perceived competence during an action can enhance intrinsic motivation for that action. Perceived competence represents the degree to which individuals perceive themselves as skilled at a given activity. The "perceived competence proposition" expects "… a close relationship between perceived competence and intrinsic motivation such that the more competent a person perceives him- or herself to be at some activity, the more intrinsically motivated he or she will be at that activity"^[22]. In addition to a need for competence, CET proposes that humans have a need for autonomy, referring to the need to experience the initiation and regulation of behavior as self-determined^[26]. The "perceived autonomy proposition" states that events that increase a person's perceived autonomy while performing a certain behavior will increase intrinsic motivation for that behavior, whereas events that reduce perceived autonomy will reduce intrinsic motivation^[22].

In laboratory studies, perceived competence has typically been manipulated by varying the type of performance feedback received while controlling for actual performance. Participants who are offered positive verbal feedback (e.g., "You performed very well") demonstrate increased perceived competence and greater subsequent enjoyment of the task than participants who are offered negative or no feedback. Extensive research has substantiated this finding^[27,28]. Therefore, an APA's feedback strategy may influence students' perceived competence during learning. In addition, it can significantly affect positive emotions, which may influence students' autonomy in learning. In summary, the APA's feedback strategy shapes two antecedents of intrinsic motivation: perceived competence and perceived autonomy (Figure 1a).

OIT details various forms of extrinsic motivation^[22]. The least autonomous form is called an external regulation. Such behaviors are performed to satisfy an external demand or to obtain a contingency of an external reward. Externally regulated behavior is experienced as externally controlled or alienated and perceived as having an external locus of causality^[24].

It was possible that choleric APAs led to better learning performance in Bian et al.'s previous study^[12] because it enhanced students' extrinsic motivation to learn. The literature supports this assertion. For instance, Nass and Moon found that compared with submissive computers (i.e., those that used more equivocal language and expressed low confidence during a task, characteristic of phlegmatic APAs), dominant computers (i.e., those that used strong, assertive language and expressed high confidence during a task, characteristic of choleric APAs) were not evaluated more negatively by participants; rather, participants evaluated the dominant computers as more competent^[29]. Vasile concluded that a medium to high impulsivity (characteristics of choleric APA) in a teacher, if such impulsivity was functional, could help students increase their knowledge level^[30]. Therefore, APAs with different personality types might pose various levels of deterrent to participants, thus affecting external regulation (i.e., extrinsic motivation; Figure 1b). Here, such a deterrent referred to a type of invisible external pressure induced by the APA's personality characteristics; a deterrent compelled learners to remain on the learning task. Therefore, an APA's personality type might be an influential factor in learners' extrinsic motivation.

3 A possible explanation model

A previous study^[12] showed that the emotional feedback of an APA influenced both learning experience and performance, and positive feedback was designed to improve learning. In comparison, the personality of an APA had a considerable influence on learning performance, and choleric APA led to better performance than phlegmatic APA. Reviewing the existing studies showed that both the agent's personality and feedback had a potential influence on the individual's motivation. Section 2.3 details the views of SDT and an analysis of its potential role in explaining the effects of APA. Therefore, based on SDT this study is focused on exploring the influence mechanism of the APA's personality and feedback strategy on learning from the perspective of the motivation effect. Although the role of motivation in explaining the persona effect has been highlighted in the literature^[19], different motivation types have not yet been considered. Based on this previous work^[12], in this study, it was assumed that the APA's personality type (choleric/phlegmatic) and feedback strategy (positive/negative) were typical and representative; thus, a preliminary theoretical model was established based on these two classifications.

Based on the aforementioned review and analysis, in this study, it was hypothesized that an APA's feedback strategy affected intrinsic motivation-related factors; whereas personality type, affected extrinsic motivation-related factors. This distinction served as the basic hypothesis to explain the effect of an APA's personality type and feedback strategy on flow experience and learning performance, from which the authors proposed a theoretical model (Figure 2).

This model suggested that learners' positive emotions and self-efficacy in learning were two antecedents of intrinsic motivation, and feedback strategies could directly affect the two antecedents. The two factors (positive emotion and self-efficacy) represented the level of perceived autonomy and perceived competence, respectively, which were two key conditions for intrinsic motivation^[24,26]. Flow was an intrinsically enjoyable state, and self-determined forms of motivation facilitate flow^[16]. Intrinsic motivation had a direct impact on learning performance, which had been validated^[31]. Consequently, positive feedback improved both flow and learning performance.

According to a previous study^[7], the APA's personality type did not significantly affect experiences related to intrinsic motivation. Thus, the effect of personality on learning performance might not be accomplished through the path of the intrinsic motivation effect, but through the path of the extrinsic motivation effect, especially given the role of external regulation. Specifically, choleric APAs were more neurotic and extraverted than phlegmatic APAs; they were active but aggressive and impulsive^[7]. Schneider examined the relationship between neuroticism and deterrent appraisal and found that they were positively correlated, which might induce higher external regulation and extrinsic motivation^[32]. In contrast, phlegmatic APAs were more passive, stable, and mild, and could not effectively enhance external regulation and extrinsic motivation. Improvements in extrinsic motivation could enhance learning performance^[33,18], but it was not a significant predictor of flow because it was controlled by external conditions and lacks autonomy. Non-self-determined forms of motivation contributed little to flow states, which explained why APAs with different personality types could affect learning performance but not flow, and it also explained why choleric APAs could lead to better performance. The proposed model attempted to explain the mechanism of APA's personality (choleric/phlegmatic) and feedback strategy (positive/negative), but related assumptions had to be empirically examined in the context of human-agent interaction.

The aim and purpose of this study was to explore the role of the motivation effect in exploring the influence mechanism of the APA's personality and feedback strategy. Two main tasks were performed: (1) an empirical study was conducted to repeatedly test the reliability of previous results on the effect of the APA's personality and feedback strategy; (2) another empirical study was conducted to explore the mechanisms of these effects.

4 Empirical study 1

4.1　Purpose and hypothesis

The aim of this study was to evaluate the reliability of the previous results, which was a prerequisite for further evaluation of the effect mechanism. According to the previous results of Bian et al. obtained from VR tai chi training, the APA's feedback strategy (positive/negative) significantly affected both flow and learning performance, but the APA's personality type (choleric/phlegmatic) only affected learning performance rather than flow experience^[12]. Combining these results with our explanation model, two hypotheses were proposed:

H1: The APA's feedback strategy significantly affected both flow and learning performance, and a positive feedback strategy led to higher levels of flow experience and better learning performance compared with the negative feedback strategy.

H2: The APA's personality type had a significant effect on learning performance rather than flow experience, and choleric APAs evoked higher levels of learning performance than phlegmatic APAs.

4.2　Design

A 2×2 factorial mixed design (i.e., personality types: choleric/phlegmatic; feedback strategies: positive/negative) was adopted here. The feedback strategy was a within-subject factor; whereas, personality type was a between-subject factor. The dependent variables included flow experience and learning performance. The presentation order of the two feedback strategies was counterbalanced.

4.3　Participants

A total of 52 undergraduate students (44 men and 8 women) participated in the test. Participants' ages ranged from 18 to 21 years (M=19.32, SD=0.73). They were all college students from a university in Shandong Province, China, and were recruited by giving notice in their classes. They were unaware of the experiment's true purpose. Only those who had no previous experience learning tai chi were chosen to participate in the formal study. The reason for this was to avoid interfering with the experimental results. They were not paid for their participation, but they would receive some small gifts as rewards.

4.4　Manipulations of APAs' personality and feedback types

According to Bian et al.'s previous paper, APAs with choleric and phlegmatic personality types were manipulated by controlling the physical appearance, behavioral features, and verbal features of APAs here (Table 1)^[12]. The feedback strategies were manipulated by designing head and upper-body movements^[12]. Positive feedback included nodding, clapping hands, thumb up, and two thumbs up. Negative gestures included shaking the head, waving hands, crossing hands, and pointing fingers.

APAs' personality parameters (according to Bian' previous work^[12])

Personality	Characteristics	Appearance manipulation	Verbal features	Behavioral features
Choleric (neurotic extravert): high E score and high N score	Active, restless, aggressive, impulsive, hot-tempered, touchy	Asymmetrical faces, frowning eyebrows, staring eyes, prolonged direct eye gaze, wide eyes	Strong, confident, powerful words and phrasing	Higher movement speed, larger movement range
Phlegmatic (stable introvert): low E score and low N score	Passive, peaceful, controlled, careful, even-tempered, calm	Symmetrical face, peaceful face, no facial expression, formal clothes	Calm, less direct and confident phrasing	Lower movement speed, smaller movement range

4.5　Experimental task and procedure

The participants were asked to learn tai chi in a virtual training studio. Upon logging into the system, each participant was immediately assigned to one APA (choleric or phlegmatic; Figure 3). The details of the APAs and the training system can be found in Bian et al.^[12]'s previous study. First, the participant learned four forms of tai chi (20min). The APA used one of the two feedback strategies to instruct the participants during learning. Second, the participants' learning outcomes were recorded as videos. Third, the participants then completed a questionnaire. In the second 20min session, the participant learned another four forms of tai chi using the same procedure. The difficulty between the two tasks was the same. The only difference was that the agent used a different feedback strategy to instruct the learners. The remaining procedures were the same as described earlier.

4.6　Measures

4.6.1　Flow experience

Flow was measured using the Flow Short Scale. The scale included 10 items (e.g., "The right thoughts/movements occur of their own accord"). In the original questionnaire, each item was assessed on a seven-point Likert scale. In a previous study by Bian et al.^[12], it was adapted to a five-point Likert scale from 1 (strongly disagree) to 5 (strongly agree). As in the previous study, this study also used a five-point Likert scale. The final score was calculated by totaling the scores of each item; thus, the range of score was 10‒50. This scale has been proven to be a dependable instrument^[34]. The internal consistency reliability of the scale was good (α=0.874).

4.6.2　Performance

The expert evaluation method was used to evaluate learning performance. The participants' performance on these strokes of tai chi was evaluated according to five appraisal criteria, which included high technical quality, good coordination, appropriate strength, good spiritual state, and appropriate rhythm^[35]. After watching the recorded performance of each participant, three experts scored each stroke on a five-point Likert scale from 1 to 5. If a stroke was totally abandoned by the participant in the video, it was scored as 0. For each expert, the total score of performance he gave was calculated by totaling the scores of each stroke; thus, the range of score was 0‒100. The final performance score was the average score of the three experts' scores. The rating scores of the three experts were highly consistent (α=0.813).

4.7　Results

To explore the effect of the APA's personality type and feedback strategy on learners, two 2×2 repeated-measures analysis of variance (ANOVAs) was performed using flow experience and performance as dependent variables. Levine's test for homogeneity of error variances supported the equal variance assumption for the analyses. Table 2 lists the descriptive statistics.

Mean values of variables related to extrinsic motivation under different conditions

Personality types	Measures	Positive feedback strategy M (S.D)	Negative feedback strategy M (S.D)
Choleric APA	Flow experience	37.10(3.97)	36.99(3.99)
	Performance	63.24(13.59)	61.03(14.22)
Phlegmatic APA	Flow experience	37.19(5.69)	35.47(5.22)
	Performance	57.75(15.71)	52.61(14.39)

The results revealed a significant main effect of feedback strategy on flow experience, F _(1,50)=7.943, p=0.0007, η² _p =0.142. Compared with the negative feedback strategy, the positive feedback strategy was associated with a higher level of flow experience. The main effect of personality type on flow experience was not significant, F _(1,50)=4.404, p=0.679, η² _p =0.004. The interaction effect of personality type×feedback strategy on flow experience was not significant.

The results also revealed a significant main effect of feedback strategy on performance, F _(1,50)=6.178, p=0.016, η² _p =0.114. Moreover, results on the effect of the APA's personality type revealed a tendency towards significance, F _(1,50)=3.709, p=0.060, η² _p =0.072. Compared with phlegmatic APA, choleric APA was associated with a higher level of performance. However, the interaction effect was not statistically significant.

Therefore, these results were in line with previous results that also showed a significant effect of emotional feedback on learning experience and performance, but personality type only had a considerable influence on learning performance. These results supported the proposed theoretical models H1 and H2. A deeper and more comprehensive examination of the model might help clarify the effect mechanism of the APA's personality type and feedback strategy. Next, empirical study 2 evaluated the theoretical model in detail.

5 Empirical study 2

5.1　Purpose and hypothesis

Following the study's results, Study 2 further evaluated the theoretical model to clarify how APAs' personality type and feedback strategy affected learning. Based on the model, the direct effects of intrinsic and extrinsic motivation on learning performance and the direct effect of intrinsic motivation on flow have been verified^[31,36]. Therefore, this study focused on examining the influence of APAs' personality types and feedback strategies on intrinsic and extrinsic motivation (Figure 4).

According to this model, APAs with different personality types might induce different "deterrence" appraisals, which might lead to various levels of extrinsic motivation. Thus, the following two hypotheses (Hs) were proposed:

H1: The APA's personality type had a significant effect on "deterrence" appraisal, and choleric APAs led to higher levels of deterrence reactions than phlegmatic APAs.

H2: The APA's personality type had a significant effect on extrinsic motivation, and choleric APAs led to higher levels of extrinsic motivation than phlegmatic APAs.

According to the proposed model, APAs' feedback strategies affected students' positive emotions and learning self-efficacy, which further influenced intrinsic motivation. Therefore, H3, H4, and H5 were proposed as follows:

H3: The APA's feedback strategy had a significant effect on the learner's positive emotions; a positive feedback strategy led to more positive emotions than a negative strategy.

H4: The APA's feedback strategy had a significant effect on learners' self-efficacy; a positive feedback strategy led to a higher level of self-efficacy than a negative strategy.

H5: The APA's feedback strategy had a significant effect on the learner's intrinsic motivation; a positive feedback strategy led to higher intrinsic motivation than a negative strategy.

5.2　Design

This study conducted an experiment involving a 2×2 factorial within-subjects design (i.e., personality type: choleric/phlegmatic; feedback strategy: positive/negative). Personality type and feedback strategy were within-subject factors; dependent variables included indicators of the intrinsic motivation effect (emotional valence, self-efficacy, and situational intrinsic motivation) and extrinsic motivation effect (deterrent appraisal and situational extrinsic motivation). To avoid interference from the APA's gender, each participant was matched with an APA of the same gender. To eliminate a potential sequential effect, the order in which the experimental conditions were presented was counterbalanced.

5.3　Participants

A total of 29 undergraduates (20 men and 9 women) were recruited to participate in the experiment. Participants' ages ranged from 20 to 27 years (M=23.06, SD=2.14). The sampling method was the same as that used in Study 1. They were unaware about the study' purpose and had no previous experience in learning tai chi to avoid potential interference. They were not paid for their participation, but they would receive some small gifts as rewards.

5.4　Apparatus and materials

Here, a wireless Biopac MP150 physiological data acquisition system loaded with AcqKnowledge 4.3 software was used to acquire physiological signals. Participants' cardiovascular activity was recorded by applying bipolar EL 504 cloth base electrodes to the left and right sides of the chest with a 3cm×30cm Electro Lead (BN-EL30-LEAD3) connected to a Biopac BioNomadix RSP and electrocardiogram amplifier. Skin conductance was recorded using a BioNomadix electrodermal activity transducer connected to a BioNomadix two-channel electrodermal activity amplifier. Recorded data were imported into AcqKnowledge 4.3 for calculation^[12].

5.5　Experimental task

The experimental program was completed using E-prime 2.0 and presented on a large projection screen. The program depicted the forms of tai chi using individual 3D images. These images were continuous in content but were presented one-by-one. Participants needed to carefully observe and memorize each form and imagine themselves learning tai chi by following the images. The presentation speed of the images was controllable: each participant could click the mouse to advance to the next image as needed. At certain time points during the learning task, two questions were presented sequentially to evaluate the quality of learning. However, the main purpose of the questions was not to measure achievement, but to create a situation in which participants received feedback. After answering the questions, an APA appeared on the screen to provide feedback, and the study's aim was to examine participants' social responses (especially physiological) to the APA. To avoid interference generated by movement, participants were not allowed to imitate the tai chi movements displayed on the screen.

5.6　Procedure

After arriving at the lab, each participant was given a chest belt and surface electrodes before being provided with experimental instructions. Electrodes used to measure SCL were connected to the forefinger and middle finger of the left hand; electrodes used to measure HR were connected to the chest. Prior to the experimental task, each participant completed a 5min baseline measurement. Subsequently, they learned how to perform the task through a sample exercise program. Subsequently, each participant began the formal experimental task. The system assigned an APA as a virtual coach, and the coach provided a simple self-introduction. Then, the participant performed two phases of tai chi learning under the agent condition. The APA used two different strategies to provide feedback information in the two phases. Immediately after completing the task, participants completed a questionnaire.

After resting for 5min, the participant was reassigned to another APA with a different personality and subsequently learned the remaining tai chi positions. This process was the same as that described earlier. During the task, each participant's physiological data were recorded continuously.

5.7　Measures

Indicators of the intrinsic motivation effect included emotional valence, self-efficacy for learning, and intrinsic motivation. Those related to the extrinsic motivation effect included "deterrence appraisal" and extrinsic motivation.

5.7.1　Emotional valence

Emotional valence was evaluated using an 11-point Likert scale (Table 3).

The item of measuring emotional valence

Negative Neutral Positive
-5	-4	-3	-2	-1	0	1	2	3	4	5

5.7.2　Self-efficacy for learning

Self-efficacy refers to one's perceived capability to perform a course of action to attain a desired achievement. It is defined in terms of specific skills or domains^[37]. Here, self-efficacy for learning was measured using three items^[7]. Participants rated them on a five-point Likert scale from "1" (totally disagree) to "5" (totally agree). The scale was dependable (α=0.931).

5.7.3　Situational intrinsic motivation

Intrinsic motivation was measured using two items from the intrinsic motivation subscale of the Situational Motivation Scale (SIMS)^[38]. Because the use of SIMS was situational, this study outlined a situation for these items, such as, "I continue to learn under the instruction of the virtual coach, because I think that this activity is interesting". Participants rated them on a five-point Likert scale from "1" (totally disagree) to "5" (totally agree). The scale was highly dependable (α=0.678-0.712).

5.7.4　Appraisal of deterrent

Because an individual's responses to an APA were usually mindless, the respondent might deny that they responded socially to an APA^[29]. Therefore, deterrent was measured using physiological data. According to previous results^[39,32], changes in heart rate (ΔHR) and changes in skin conductance level (ΔSCL) were chosen as physiological markers of deterrent appraisal.

ΔHR=mean HR-baseline HR

The mean HR represented the average HR of all recorded fragments under one experimental condition. The heart beats faster under stress; thus, when a participant perceived a deterrence, the HR accelerated^[39].

ΔSCL=mean SCL-baseline SCL

The mean SCL represented the average SCL of all recorded fragments under one experimental condition. The skin of the palm sweats more under stress, independent of temperature; thus, its conductance increased. When the participant perceived a deterrence, the SCL rose^[39].

5.7.5　Situational extrinsic motivation

Situational extrinsic motivation was measured using two items from the external regulation subscale of the SIMS^[38]. For example, "I continue to learn under the instruction of the virtual coach, because it is something that I have to do". Items were assessed on a five-point Likert-type scale ranging from "1" (strongly disagree) to "5" (strongly agree). The scale was dependable (α=0.801-0.878).

5.8　Results

5.8.1　Effects on variables related to extrinsic motivation

Data from one participant were eliminated because the surface electrode fell off during the experiment. Finally, data from the remaining 28 participants were analyzed.

To explore the effect of personality type and feedback strategy on deterrence reactions, two 2×2 repeated-measures analysis of variance (ANOVAs) were performed using ΔHR and ΔSCL as dependent variables. Table 4 lists the descriptive statistics.

Mean values of variables related to extrinsic motivation under different conditions

Experimental conditions	ΔHR M(SD)	ΔSCL M(SD)	Extrinsic motivation M(SD)
Choleric APA+	1.19(3.84)	3.02(5.16)	6.21(2.68)
positive feedback
Choleric APA	2.59(4.82)	3.07(4.99)	6.69(1.98)
+negative feedback
Phlegmatic APA	0.64(3.45)	2.71(4.63)	6.45(1.96)
+positive feedback
Phlegmatic APA	1.16(3.98)	2.39(4.61)	5.86(1.75)
+negative feedback

The results revealed a marginally significant main effect of personality type on ΔHR, F _(1,27)=3.759, p=0.064, η² _p =0.131. Compared with phlegmatic APA, choleric APA was associated with a higher level of ΔHR (Figure 5). The main effect of the feedback strategy on ΔHR was significant, F _(1,27)=4.404, p=0.046, η² _p =0.150. Compared with the positive feedback strategy, negative strategy feedback led to a higher ΔHR. The interaction effect of personality type×feedback strategy on ΔHR was not significant, F _(1,27)=0.956, p=0.338, η² _p =0.037.

The results of ΔSCL indicated that the main effect of the APA's personality type was marginally significant, F _(1,27)=4.134, p=0.052, η² _p =0.133. The choleric APA increased participants' ΔSCL more than the phlegmatic agent did (Figure 6). However, neither the main effect of the feedback strategy, F _(1,27)=0.250, p=0.621, η² _p =0.009, nor the interaction effect, F _(1,27)=0.746, p=0.395, η² _p =0.027, was significant, supporting H1.

Another 2×2 repeated-measures ANOVA was conducted to examine the effect of the APA's personality type and feedback strategy on extrinsic motivation. The results of the difference tests showed that the main effect of personality type was significant, F _(1,27)=4.327, p=0.047, η² _p =0.134. Specifically, the choleric APA induced stronger extrinsic motivation than the phlegmatic APA (Figure 7), supporting H2. However, neither the main effect of feedback type, F _(1,27)=0.441, p=0.512, η² _p =0.016, nor the interaction effect, F _(1,27)=1.665, p=0.207, η² _p =0.056, was significant.

5.8.2　Effects on variables related to intrinsic motivation

To explore the effects of personality type and feedback strategy on variables related to intrinsic motivation, a set of repeated-measures ANOVAs were performed. Table 5 lists the descriptive statistics.

Mean values of variables related to intrinsic motivation under different conditions

Experimental conditions	Emotional valence M(SD)	Self-efficacy for learning M(SD)	Intrinsic motivation M(SD)
Choleric APA+ positive feedback	2.31(2.21)	7.31(3.01)	7.24(1.79)
Choleric APA+ negative feedback	‒0.90(2.43)	5.55(1.88)	4.55(2.18)
Phlegmatic APA+ positive feedback	2.17(1.69)	7.79(1.59)	7.62(1.63)
Phlegmatic APA + negative feedback	‒0.79(2.59)	6.00(1.79)	5.00(1.75)

The results on emotional valence suggested that participants had negative emotional experiences under the negative feedback condition and positive emotional experiences under the positive feedback condition (Figure 8). Also, the values of emotional valence under different conditions were significantly different, F _(1,27)=34.141, p<0.001, η² _p =0.549. These results supported H3. Nevertheless, the main effects of the APA's personality type, F _(1,27)=0.003, p=0.958, η² _p =0.000, and the interaction effect, F _(1,27)=0.203, p=0.656, η² _p =0.007, were not significant.

The results on self-efficacy revealed a significant main effect of feedback strategy, F _(1,27)=17.153, p<0.001, η² _p =0.380, with participants in the positive feedback condition demonstrating higher self-efficacy (Figure 9); thus, supporting H4. Neither the main effect of the APA's personality type, F _(1,27)=2.910, p=0.099, η² _p =0.094, nor the interaction effect, F _(1,27)=0.005, p=0.946, η² _p =0.000, was significant.

The results on intrinsic motivation also revealed a significant main effect of feedback strategy (Figure 10), F _(1,27)=49.328, p<0.001, η² _p =0.638, supporting H5. Neither the main effect of APA's personality type, F _(1,27)=2.865, p=0.102,η² _p =0.093, nor the interaction effect, F _(1,27)=0.018, p=0.894, η² _p =0.001, was significant.

6 Discussion

This study further explored the underlying mechanism of APAs' typical personality type (choleric and phlegmatic) and feedback strategy (positive and negative) on students' flow experience and learning performance. The authors hypothesized that the distinction between intrinsic and extrinsic motivation would play a significant role in clarifying this question, based on which a theoretical model was proposed. Subsequently, two empirical studies were conducted to evaluate the main principle of the model, and the results verified the hypotheses and supported the model.

6.1　The effects of APA's personality

This study verified that the APA's personality type affected only learning performance but not flow experience because the effect of personality type on learning performance worked through the path of the extrinsic motivation effect rather than the intrinsic motivation effect.

None of the effects of personality type on the variables related to intrinsic motivation were significant. Therefore, the basis of Bian et al.'s previous study to predict the effect of APAs' personality was not appropriate (i.e., APAs with different personality types influenced positive emotions and led to different learning performance)^[7]. The key point was not that the choleric APA might evoke negative emotions, but that it posed a kind of deterrence/pressure to learners. This was similar to positive external pressure, which promoted extrinsic motivation for learning. Specifically, different personality types induced various levels of deterrence appraisal (i.e., a condition of external regulation), leading to differences in extrinsic motivation^[24]. The results on participants' physiological indicators of deterrence reaction and subjectively perceived extrinsic motivation suggested that the choleric APA evoked a higher level of deterrence reaction and external motivation, supporting the hypothesis of the extrinsic motivation effect.

These results were consistent with those of several previous studies. Li et al. developed two virtual interviewers with distinct personalities and deployed them in a real-world recruiting event^[40]. They found that users would confide in and listened more to a virtual interviewer with a serious, assertive personality (similar to choleric APA). These conclusions were in line with Vasile's finding in an actual education situation: as long as the responses were functional, teachers with moderate to high impulsivity would help improve students' knowledge^[30]. Moreover, when collaborating with a strict person, one tended to be more careful about rules, and a higher level of extrinsic motivation would be simulated. The media equation theory suggested that the social behavioral and perceptual processes observed in human-human interactions could be applied to human-agent interaction, which have been verified^[9,41]. Therefore, all these were as evidence supporting the present results. Moreover, the question of why the APA's personality type had no significant effect on flow could also be explained by these results; i.e., extrinsic motivation did not contribute to flow.

However, there were also some inconsistent results. According to an existing study^[42], if the APA displayed friendly actions (similar to phlegmatic APA), students in response would be more motivated. Moreover, Hanna and Richards explored the impact of virtual agents with different personality traits on taskwork and teamwork and found a significant difference between the groups of participants according to the agent's agreeableness/antagonism personality (related to phlegmatic/choleric)^[14]. Agreeable agents positively influenced the development of taskwork and teamwork. This did not seem to be in line with this study and previous results mentioned. Actually, this fit with the stereotype that a gentle and friendly teacher was more likely to encourage motivation because this trait seemed to be more "positive." However, most results showed that APAs with strict, serious, and assertive personalities led to stronger influence and learning results. The reason for this might lie in the distinction between different motivations. APAs' friendly actions might inspire intrinsic motivation, while APAs' serious actions might inspire extrinsic motivation. Therefore, in terms of the influence on learners, which motivation effect played a more prominent role, might depend on the following point: whether the personality expression was strong enough to dominate the cognition of learners.

6.2　Effects of APA's feedback

The model also assumed that the APA's feedback strategy affected learning performance along with flow experience because the feedback strategy operated through the path of the intrinsic motivation effect. The results indicated that the APA's positive feedback strategy led to significantly greater positive emotions, higher self-efficacy, and a higher level of intrinsic motivation. These results supported the hypothesis of intrinsic motivation effect. This meant that the APA's feedback strategy influenced learners' perceived autonomy and perceived competence, two main conditions of intrinsic motivation^[24]. Intrinsic motivation could be influenced by the APA's feedback strategy, which was consistent with some existing findings, such as those suggesting that APAs' positive feedback could serve as an effective form of motivational feedback to enhance intrinsic motivation^[15]. Intrinsic motivation was a vital contributor to flow and learning performance^[31,16]; thus, the APA's positive feedback could lead to more favorable flow and learning performance.

This study provided a detailed theoretical explanation in Section 3. The empirical results supported the theoretical model and interpretation, which will not be discussed here.

6.3　Motivation effect

Although this study examined the effects of the APA's personality and feedback in detail, the two paths of the motivation effect were more important. The types of personality and feedback investigated here were limited, and the motivation effect was merely explored based on these two characteristics. However, the two paths of the motivation effect proposed in the model were universal and could also be used to analyze and verify the impact of the other characteristics of the APA on users.

These two motivation types have been distinct for a long time; however, neither have they been used to explain the mechanism of the persona effect nor to analyze how the characteristics of APAs motivate or persuade users. Studies from the perspective of the motivation effect could help understand and develop motivational and persuasive technology, both of which were fields of concern in human-computer interaction.

Computer systems are increasingly imbued with motivational design with the aim of positively engaging users towards engagement with the task they are attempting to accomplish through using the system^[43‒46]. The end goal of motivational design is not only motivation but also the accomplishment of a level of behavioral change^[47]. Motivational design requires the command of several disciplines, such as (motivational/social/behavioral) psychology and design beyond software development^[48,49]. Sundar et al. proposed a theoretical framework for motivational application design based on SDT^[50]. Their ideas were similar to our models but are not set in human-agent interaction situations. This study was consistent with these studies to a certain extent and could complement each other.

Persuasive technology referred to technology that was designed to change users' attitudes or behaviors through persuasion and social influence. It has been heralded as a new paradigm using technology to change people's behavior in human-computer interaction^[51]. Halko and Kientz summarized the persuasive strategies and methods used in the field of psychology to modify behaviors^[52]. One of the categories, similar to the idea of this study, was the internal and external motivators. The former referred to persuading the user through external motivators (e.g., giving pressure or a reward for completing a task), while the latter referred to persuading the user through internal motivators (e.g., the good feeling a user would have to achieve the goal). Therefore, the current work on the motivation effect could help clarify how to use persuasive technology to change learners' attitudes and behaviors in human-agent interaction and provide a reference for persuasion design.

6.4　Implication for VR community

The present study deepened scholarly understanding of motivation and persona effects. It provided a theoretical reference for designers of APAs to rationally use their advantages to develop motivating virtual learning environments to enhance the VR learning experience and outcomes. On the one hand, understanding the effect of the two characteristics of APA on learning and the underlying mechanism could help designers understand how people perceive APA, which could provide a reference for them to design more motivational APAs in VELE. On the other hand, according to the theoretical model and empirical results, designers could not only understand the role of APAs' emotional feedback and personality type, but also pay attention to other factors relevant to the path of the motivation effect. In summary, this work provided theoretical and empirical references to optimize the VR training/learning environment.

6.5　Limitations and future directions

Personality and feedback represent two critical APA characteristics for improving knowledge and acquiring skills^[9,53], and the classification of the two characteristics was typical. However, this study only provided a preliminary framework. This was a further study based on the results of Bian et al.^[12]. To maintain the continuity and comparability with the previous study, the experimental design of this study needed to be consistent with the previous study. For this reason we only divided the feedback (positive and negative) and personality of APA (choleric and phlegmatic) into two types. However, neutral and several levels of positive/negative feedback, as well as more types of personality types, should be considered. Future research will focus on these questions and further develop an explanation model.

Another limitation was that individual differences were not considered in the model. There might be interactions between APA and user characteristics during the human-agent interaction in a virtual learning environment. The effect of APAs' characteristics might be different for different individuals, or the effect of different APAs might have different effects on the same individual. For instance, a study investigated the influence of the match in the two personality traits between virtual agents and humans on the development of a shared mental model (SMM) and performance. They found that when the collaborating human and agent had matching agreeableness personality traits, there was a positive influence on the SMM between them and better performance outcomes. Therefore, the applicability of the model should be further evaluated in human-agent interactions when considering individual factors.

Finally, the current research outcome was not enough to develop a quantitative model of the influence of the APA's personality and feedback strategy. On the one hand, our proposed theoretical explanation was not widely proven for sufficient applications. On the other hand, the level of personality characteristics and emotional feedback were not further distinguished and quantified. The development of a quantitative model of the influence of an APA's personality and feedback strategy is a direction for future research.

7 Conclusions

In this study, SDT was applied to explain the persona effect, and a theoretical model was proposed to clarify how APAs' personality and feedback types affect users' flow experience and performance through paths of different motivation effects, which appeared reasonable based on the results. An APA's feedback strategy affected learners through the intrinsic motivation effect, whereas the APA's personality type affected learners through the extrinsic motivation effect. The present study deepened scholarly understanding of the motivation effect and provided evidence to refine the theoretical development of the persona effect.

Reference

Kim Y, Baylor A L, Shen E. Pedagogical agents as learning companions: the impact of agent emotion and gender. Journal of Computer Assisted Learning, 2007, 23(3): 220–234 DOI:10.1111/j.1365-2729.2006.00210.x

Dunsworth Q, Atkinson R K. Fostering multimedia learning of science: Exploring the role of an animated agent's image. Computers & Education, 2007, 49(3): 677–690 DOI:10.1016/j.compedu.2005.11.010

Kim C. The role of affective and motivational factors in designing personalized learning environments. Educational Technology Research and Development, 2012, 60(4): 563–584 DOI:10.1007/s11423-012-9253-6

Lin L J, Atkinson R K, Christopherson R M, Joseph S S, Harrison C J. Animated agents and learning: Does the type of verbal feedback they provide matter? Computers & Education, 2013, 67: 239–249 DOI:10.1016/j.compedu.2013.04.017

Beale R, Creed C. Affective interaction: How emotional agents affect users. International Journal of Human-Computer Studies, 2009, 67(9): 755–776 DOI:10.1016/j.ijhcs.2009.05.001

Terzis V, Moridis C N, Economides A A. The effect of emotional feedback on behavioral intention to use computer based assessment. Computers & Education, 2012, 59(2): 710–721 DOI:10.1016/j.compedu.2012.03.003

Bian Y L, Yang C L, Gao F Q, Li H Y, Zhou S S, Li H C, Sun X W, Meng X X. A framework for physiological indicators of flow in VR games: construction and preliminary evaluation. Personal and Ubiquitous Computing, 2016, 20(5): 821–832 DOI:10.1007/s00779-016-0953-5

Hu C, Walker M A, Neff M, Fox Tree J E. Storytelling agents with personality and adaptivity. In: Intelligent Virtual Agents. Cham: Springer International Publishing, 2015: 181–193 DOI:10.1007/978-3-319-21996-7_19

McRorie M, Sneddon I, McKeown G, Bevacqua E, de Sevin E, Pelachaud C. Evaluation of four designed virtual agent personalities. IEEE Transactions on Affective Computing, 2012, 3(3): 311–322 DOI:10.1109/t-affc.2011.38

10.

Ogawa Y, Kikuchi H. Assigning a personality to a spoken dialogue agent by behavior reporting. New Generation Computing, 2017, 35(2): 181–209 DOI:10.1007/s00354-017-0012-4

11.

Read S J, Miller L C. Virtual personalities: a neural network model of personality. Personality and Social Psychology Review, 2002, 6(4): 357–369 DOI:10.1207/s15327957pspr0604_10

12.

Bian Y L, Yang C L, Guan D D, Xiao S, Gao F Q, Shen C A, Meng X X. Effects of pedagogical agent's personality and emotional feedback strategy on Chinese students' learning experiences and performance: a study based on virtual Tai Chi training studio. In: Proceedings of the 2016 CHI Conference on Human Factors in Computing Systems. San Jose California, New York, NY, USA, ACM, 2016, 433–444 DOI:10.1145/2858036.2858351

13.

Dekker T W G. Personality in embodied conversational agents: Effects on user experience. Proc. of 17th Twente Student Conference on IT, 2012

14.

Hanna N, Richards D. The impact of virtual agent personality on a shared mental model with humans during collaboration. AAMAS, 2015, 1777–1778

15.

Mumm J, Mutlu B. Designing motivational agents: The role of praise, social comparison, and embodiment in computer feedback. Computers in Human Behavior, 2011, 27(5): 1643–1650 DOI:10.1016/j.chb.2011.02.002

16.

Shiban Y, Schelhorn I, Jobst V, Hörnlein A, Puppe F, Pauli P, Mühlberger A. The appearance effect: Influences of virtual agent features on performance and motivation. Computers in Human Behavior, 2015, 49: 5–11 DOI:10.1016/j.chb.2015.01.077

17.

van der Meij H, van der Meij J, Harmsen R. Animated pedagogical agents effects on enhancing student motivation and learning in a science inquiry learning environment. Educational Technology Research and Development, 2015, 63(3): 381–403 DOI:10.1007/s11423-015-9378-5

18.

Yılmaz R, Kılıç-Çakmak E. Educational interface agents as social models to influence learner achievement, attitude and retention of learning. Computers & Education, 2012, 59(2): 828–838 DOI:10.1016/j.compedu.2012.03.020

19.

Chen C H, Chou M H. Enhancing middle school students' scientific learning and motivation through agent-based learning. Journal of Computer Assisted Learning, 2015, 31(5): 481–492 DOI:10.1111/jcal.12094

20.

Lim H C, Stocker R, Barlow M, Larkin H. Interplay of ethical trust and social moral norms: Environment modelling and computational mechanisms in agent-based social simulation (ABSS). Web Intelligence and Agent Systems: an International Journal, 2011, 9(4): 377–391 DOI:10.3233/wia-2011-0225

21.

Liew T W, Tan S M. Virtual agents with personality: Adaptation of learner-agent personality in a virtual learning environment. In: 2016 Eleventh International Conference on Digital Information Management (ICDIM). Porto, Portugal, IEEE, 2016, 157–162 DOI:10.1109/icdim.2016.7829758

22.

Deci E L, Ryan R M. Intrinsic motivation and self-determination in human behavior. Springer Science & Business Media, 2013

23.

Lepper M R, Corpus J H, Iyengar S S. Intrinsic and extrinsic motivational orientations in the classroom: Age differences and academic correlates. Journal of Educational Psychology, 2005, 97(2): 184

24.

Lindebaum D. Rhetoric or remedy? A critique on developing emotional intelligence. Academy of Management Learning & Education, 2009, 8(2): 225–237 DOI:10.5465/amle.2009.41788844

25.

Dinçer S, Doğanay A. The effects of multiple-pedagogical agents on learners’ academic success, motivation, and cognitive load. Computers & Education, 2017, 111: 74–100 DOI:10.1016/j.compedu.2017.04.005

26.

Ryan R M, Deci E L. Intrinsic and extrinsic motivations: classic definitions and new directions. Contemporary Educational Psychology, 2000, 25(1): 54–67 DOI:10.1006/ceps.1999.1020

27.

Reeve J, Deci E L. Elements of the competitive situation that affect intrinsic motivation. Personality and Social Psychology Bulletin, 1996, 22(1): 24–33 DOI:10.1177/0146167296221003

28.

Vansteenkiste M, Deci E L. Competitively contingent rewards and intrinsic motivation: can losers remain motivated? Motivation and Emotion, 2003, 27(4): 273–299 DOI:10.1023/a:1026259005264

29.

Nass C, Moon Y. Machines and mindlessness: social responses to computers. Journal of Social Issues, 2000, 56(1): 81–103 DOI:10.1111/0022-4537.00153

30.

Vasile C. Impulsivity dynamics in Romanian teachers’ personality. Procedia-Social and Behavioral Sciences, 2012, 69: 2101–2107 DOI:10.1016/j.sbspro.2012.12.172

31.

Fortier M S, Vallerand R J, Guay F. Academic motivation and school performance: toward a structural model. Contemporary Educational Psychology, 1995, 20(3): 257–274 DOI:10.1006/ceps.1995.1017

32.

Schneider T R. The role of neuroticism on psychological and physiological stress responses. Journal of Experimental Social Psychology, 2004, 40(6): 795–804 DOI:10.1016/j.jesp.2004.04.005

33.

Sträfling N, Fleischer I, Polzer C, Leutner D, Krämer N C. Teaching learning strategies with a pedagogical agent. Journal of Media Psychology, 2010, 22(2): 73–83 DOI:10.1027/1864-1105/a000010

34.

Bian Y L, Zhou C, Chen Y Q, Zhao Y S, Liu J, Yang C L. The role of the field dependence-independence construct on the flow-performance link in virtual reality. In: Symposium on Interactive 3D Graphics and Games. San Francisco CA USA, New York, NY, USA, ACM, 2020, 1–9 DOI:10.1145/3384382.3384529

35.

Shi B, Hu Y G, Shi F A. Study on the appraisal criteria of university students' 24-stroke Taiji. Journal of Nanyang Teachers’ College (Natural Sciences Edition), 2003, 2(3): 3–5

36.

Pintrich P R, de Groot E V. Motivational and self-regulated learning components of classroom academic performance. Journal of Educational Psychology, 1990, 82(1): 33–40 DOI:10.1037/0022-0663.82.1.33

37.

Hung C Y, Sun J C Y, Yu P T. The benefits of a challenge: student motivation and flow experience in tablet-PC-game-based learning. Interactive Learning Environments, 2015, 23(2): 172–190 DOI:10.1080/10494820.2014.997248

38.

Guay F, Vallerand R J, Blanchard C. On the assessment of situational intrinsic and extrinsic motivation: the situational motivation scale (SIMS). Motivation and Emotion, 2000, 24(3): 175–213 DOI:10.1023/a:1005614228250

39.

Meehan M, Insko B, Whitton M, Brooks F P. Physiological measures of presence in stressful virtual environments. ACM Transactions on Graphics, 2002, 21(3): 645–652 DOI:10.1145/566654.566630

40.

Li J Y, Zhou M X, Yang H H, Mark G. Confiding in and listening to virtual agents: the effect of personality. In: Proceedings of the 22nd International Conference on Intelligent User Interfaces. Limassol Cyprus, New York, NY, USA, ACM, 2017, 275–286 DOI:10.1145/3025171.3025206

41.

Reeves B, Nass C. The media equation: How people treat computers, television, and new media like real people. Cambridge, United Kingdom: Cambridge university press, 1996

42.

Borjigin A, Miao C Y, Lim S F, Li S Y, Shen Z Q. Teachable agents with intrinsic motivation. In: Lecture Notes in Computer Science. Cham: Springer International Publishing, 2015, 34–43 DOI:10.1007/978-3-319-19773-9_4

43.

Jung J H, Schneider C, Valacich J. Enhancing the motivational affordance of information systems: the effects of real-time performance feedback and goal setting in group collaboration environments. Management Science, 2010, 56(4): 724–742 DOI:10.1287/mnsc.1090.1129

44.

Landers R N, Bauer K N, Callan R C. Gamification of task performance with leaderboards: a goal setting experiment. Computers in Human Behavior, 2017, 71: 508–515 DOI:10.1016/j.chb.2015.08.008

45.

Oinas-Kukkonen H. A foundation for the study of behavior change support systems. Personal and Ubiquitous Computing, 2013, 17(6): 1223–1235 DOI:10.1007/s00779-012-0591-5

46.

Santhanam R, Liu D, Shen W C M. Research note—gamification of technology-mediated training: not all competitions are the same. Information Systems Research, 2016, 27(2): 453–465 DOI:10.1287/isre.2016.0630

47.

Orji R, Vassileva J, Mandryk R L. Modeling the efficacy of persuasive strategies for different gamer types in serious games for health. User Modeling and User-Adapted Interaction, 2014, 24(5): 453–498 DOI:10.1007/s11257-014-9149-8

48.

Deterding S. The lens of intrinsic skill atoms: a method for gameful design. Human-Computer Interaction, 2015, 30(3/4): 294–335 DOI:10.1080/07370024.2014.993471

49.

Morschheuser B, Riar M, Hamari J, Maedche A. How games induce cooperation? A study on the relationship between game features and we-intentions in an augmented reality game. Computers in Human Behavior, 2017, 77: 169–183 DOI:10.1016/j.chb.2017.08.026

50.

Sundar S S, Bellur S, Jia H Y. Motivational technologies: a theoretical framework for designing preventive health applications. Persuasive Technology. Design for Health and Safety, 2012, 112–122 DOI:10.1007/978-3-642-31037-9_10

51.

Page R E, Kray C. Ethics and persuasive technology: An exploratory study in the context of healthy living. NIMD2010, 2010

52.

Halko S, Kientz J A. Personality and persuasive technology: an exploratory study on health-promoting mobile applications. Persuasive Technology, 2010, 150–161 DOI:10.1007/978-3-642-13226-1_16

53.

Shute V J. Focus on formative feedback. Review of Educational Research, 2008, 78(1): 153–189 DOI:10.3102/0034654307313795