For the capture of academic emotions, both subjective and objective methods can be used. Facial expressions are a powerful nonverbal communication method and provide a lot of information about subjective personal experience (e.g., mental state, interest, viewpoint, physiological state, emotions). The development of artificial intelligence technology adds credibility to the recognition of various human emotions through facial expressions [
40], indicating the feasibility of using objective methods to capture academic emotions. However, studies of academic mood have long-term dependence on self-reporting [
41,
42]. Self-reporting is a valuable data source, but the desire to analyze students’ academic emotions in real time during the learning process has created a demand for other forms of academic emotional analysis. As a result, facial expressions, often seen as emotional derivatives, have become a viable channel for such exploration. Since Ekman put forward the general concept of using facial expressions as a data source, people have become more and more interested in facial expressions, and with the development of facial recognition technology over the last 10 years, there has been some related research in the field of education [
43,
44,
45,
46]. Especially in scientific experiments, we find that when the experimental results are revealed, the dominant expression of students is first surprised, then negative, and the probability of knowledge change is higher [
47]. Surprise, sadness, and disgust are also key facial expressions used to predict the change in students’ knowledge based on conflicting scenarios [
48].
Cognitive imbalance is an uncomfortable state [
67], which can lead to negative academic emotions. Cognitive imbalance is a necessary condition for students to deeply understand learning [
68]. Negative academic emotions (such as anxiety and disappointment) play an important role in the learning process. Other recent research has examined the effects of confusion and frustration on learning. Research performance, confusion, and frustration bring about better learning effects for learners [
69,
70,
71]. However, this may only apply to certain activities. For instance, if students are confused when reading multimedia materials because they do not understand the content, they may not be able to resolve their confusion, which will cause them to feel frustrated, and finally, bored. This illustrates the importance of analyzing academic emotions in each situation, so that we can understand when learners are confused or frustrated, thus helping students to learn effectively. The difference in academic emotions in the study [
36] depends on the positive or negative outcomes associated with solving the puzzle.
However, two other studies [
38,
54] have shown that academic emotional guidance has no significant influence on learning effects. An intervention study [
38] suggested that academic emotional guidance is related to behavior, learning strategies, and cognitive regulation, while an observational study [
54] suggested that, after academic emotional guidance occurs, it can improve memory and mental load without affecting learning effects. The evidence in this review extends the knowledge gained from the previous literature and suggests that positive academic emotions may have an advantage over negative academic emotions in terms of learning effects. Frequent participation in active academic emotional activities may influence behaviors conducive to learning strategy regulation [
72]. Another intervention study [
59] used graduate students as subjects to improve performance by promoting the active and flexible use of learning strategies. In the process of learning, with the increase in the difficulty of learning materials, there will be more negative academic emotions, but students can adopt more adaptive coping strategies to defuse negative academic emotions.
Due to the learning environment being constructed by different intervention conditions (learning forms and learning materials), feature-based academic emotions include the tendency of individuals to make stable and consistent responses in a specific way. In addition, immediate academic emotional response constitutes state-based academic emotion [
73]. Cues in learning situations can affect this response and may fluctuate over time. However, feature-based and state-based academic emotions are often very similar, and need to be carefully distinguished [
73]. Finally, the research [
52] is on state-based academic emotions, focusing on students’ short-term academic emotional experience in a specific game situation rather than their general experience. The academic emotions guided by learning materials focus on what is experienced by students under the particular intervention conditions. However, in certain cases, individual characteristics may be more advantageous. Games have promoted students’ science achievement in the long term [
52]. Some studies have proposed a spiral emotional learning model, which includes a right-to-left academic emotional axis level positive price, which explains the correlation between academic emotions and scientific learning [
74].
The model consisted of a right-to-left horizontal academic emotional axis of positive valence, a top-to-bottom representational constructive learning, and a bottom-to-top learning vertical learning axis, and identifies four quadrants. At the origin, the third axis of knowledge is perpendicular to these two axes. Some studies have proposed that in the process of science learning, learners’ academic emotions will change with the learning process, and knowledge will be acquired with their movement in the quadrant and spiral up along the knowledge axis [
74]. In addition, available evidence suggests that the difficulty of learning materials may have different effects on the use of learning strategies and academic emotional state [
58,
59]. Due to the challenge of learning materials, it is difficult for students to self-regulate their learning. Therefore, in the game-based learning environment, students can autonomously learn and practice, which enables them to maintain a high level of motivation and participation [
75]. However, there is still some debate about which academic emotions may be more effective at improving learning effects. Therefore, from now on, we should make further efforts to clarify the differences and relationships between the two types of academic emotions. In addition, in subsequent studies, the characteristics of academic emotions may need to be considered as a covariable to control their effects.
In summary, the results of this systematic review indicate that, compared with negative academic emotions, positive academic emotions may be more effective at improving certain aspects of learning effects, especially in high school and college students. The research results not only contribute to an understanding of the different learning effects of academic emotions, but also have some practical significance. There is growing support for incorporating the regulation of students’ academic emotions into classroom teaching plans, perhaps through different learning materials, learning environments, teaching methods, and facial recognition tools, to study students’ academic emotions throughout the learning process as an effective way to improve learning effects [
52,
53,
60]. Gee [
76] argues that learning should not be divorced from experience, because it is always best when people identify and generalize patterns in a given environment through concrete experience over a long period of time. It is important to note that passive academic emotions may be more conducive to long-term learning and memory, and the beneficial effect on learning effects should not be ignored, even though it may lead to poor academic performance. Therefore, a series of academic emotions (positive and negative) will appear in science learning experience, which may contribute to the learning effect. In addition, the lack of a social component (person–person interaction) in our learning scenarios may lead to more academic emotions being expressed through facial expressions [
77]. Social interactions that occur during the learning process may have further positive effects on learning effects [
56].