Emotions and cognition are intertwined as evidenced by studies (Dolan & Vuilleumier,
2003; Kort, Reilly, & Picard,
2001; Linnenbrink,
2006; Phelps,
2004; Vuilleumier,
2005) investigating the interactions between emotion and attention, learning and memory. Anatomically, Gray, Braver, and Raichle (
2002) also points to the integration of emotional and cognitive processes in prefrontal cortex of the human brain. This anatomic link was played out in real life in the unfortunate events experienced by Elliot, a patient of Damasio (
1994). As a result of a tumor cut from his cortex near the brain’s frontal lobe, Elliot, despite having a high level intelligence, became incapable of making simple decisions in his daily life and this eventually led to failures in both his career and marriage.
The link between emotions and cognition prompted studies investigating into the relationship between emotions and the learning achievement of students. Pekrun, Goetz, Titz, and Perry (
2002) opined that academic emotions (which are linked to academic learning, classroom instruction and achievements) with the exception of test anxiety are largely neglected by educational psychology research. They embarked on a series of qualitative and quantitative studies to uncover the academic emotions experienced by students and the effects of these emotions on the students’ learning and achievements. Their findings revealed that students experience a wide repertoire of emotions in academic settings and both positive emotions e.g. hope, pride and relief and negative emotions such as anxiety and boredom are prevalent in learning. The positive emotions e.g. enjoyment of learning affect achievement positively by strengthening motivation and enhancing flexible learning whereas the negative emotions e.g. anxiety erodes motivation and draws attention away from the task, resulting in shallow learning. The effects that both positive and negative emotions have on learning were also corroborated by other researchers (Ashcraft & Kirk,
2001; Greene & Noice,
1988; Isen,
2000).
Intelligent Tutoring Systems are built with the objective of providing learners with the benefit of one to one tutoring automatically and cost effectively. It acts like a personal training assistant that continually assesses one’s knowledge through interactions with the system and builds a personalized model of one’s acquired knowledge for the provision of tailored instructions or assistance in the form of hints or demonstrations when one seems to require help to move on.
A common criticism of ITS is that they are devoid of emotional awareness and empathy and that limits their tutoring effectiveness (Lepper, Woolverton, Mumme, & Gurtner,
1993). Pivotal to the effectiveness of human tutors in one to one tutoring is the fact that human tutors are able to constantly monitor both the affect and cognitive ability of their tutees and respond with adequate supportive tutoring measures (Goleman,
1995). Analogous to the way human tutors detect and respond to the affect of the their tutees to sustain the engagement of the latter, Affective Tutoring Systems (Picard,
1997) adapt to the tutees’ affect autonomously to bring about enhanced learning outcomes. Most Affective Tutoring Systems (ATSs) incorporate affect sensing, tutoring strategies and learning progress tracking into a single environment. With the ability to sense the affect of students, ATSs can for instance, sense that students are frustrated and offer hints to resolve the impasse.
Although researchers acknowledge the role that affect plays in learning, ATSs on the contrary are rarely implemented (Thompson & McGill,
2012). This is often attributed to the fact that the implementation of ATSs encompasses cross-disciplinary knowledge spanning domains of education, psychology and computer science. Often, the implemented ATS is highly specialized and purpose-built for a particular tutoring subject area, making replication or customization for a different subject area difficult, if not impossible.
AutoTutor is a fully automated computer tutor that helps students learn Newtonian physics and computer literacy topics. It presents problems or questions to the students and engages them in a dialogue to collaboratively build towards a solution (S. D’Mello et al.,
2008). D’Mello et al. aim to incorporate learners’ affect into existing AutoTutor’s pedagogical strategies (e.g. by regulating their negative emotions) and their research efforts culminated in the development of Affective AutoTutor. Affective AutoTutor is one of the few ATSs that detects and responds to students’ affective states (D'Mello & Graesser,
2010). Affective AutoTutor uses facial cues, body postures and conversational dialogue feature to infer the affect of students. The tutoring actions are then derived from a set of production rules that dynamically assessed the cognitive and affective states of students to address negative emotions such as frustration and boredom.
Easy with Eve is an ATS that is built by Massey University, New Zealand (Alexander, Sarrafzadeh, & Hill,
2006). It infers the emotions of learners mainly through facial expressions analysis that was developed in-house. In addition, a case based reasoning program was developed to output a weighted set of tutoring actions and facial expressions based on a given sequence of interactions. The recommended facial expression would be expressed through Eve – an animated intelligent agent embodied within the system.
Empathic Companion is an embodied agent type system that detects and responds to the user’s affective state (Prendinger, Dohi, Wang, Mayer, & Ishizuka,
2004). It was developed in the context of a web-based job interview scenario with the objective of regulating the user’s negative emotions when faced with difficult job interview questions. It employs the use of a decision network for fusing the inputs from a Galvanic Skin Response (GSR) and an Electromyography (EMG) sensor and for translating the sensors’ signals into both the emotions and the agent decisions. The results indicated that Empathic Companion does reduce the frustration of its users.
MetaTutor is an adaptive ITS that is designed to encourage students to employ meta-cognitive self-regulated learning (SRL) in the tutoring context of human circulatory system (Azevedo, Johnson, Chauncey, & Burkett,
2010). MetaTutor’s focus is on SRL skills that are fostered by managing learning through monitoring and strategy use. This is achieved in MetaTutor through the use of pedagogical agents which respond with an evaluation of the student’s current level of understanding upon request.
Crystal Island is a 3-dimensional narrative-centered learning environment for the tutoring of eighth-grade microbiology (Rowe et al.,
2009). Sabourin, Rowe, Mott, and Lester (
2013) fed the survey scores and in-game progress of students within Crystal Island into a Dynamic Bayesian Model (DBN) for the identification of students’ off task behavior. The students’ actions within the environment were logged. The logs were then analyzed to extract the behaviors of students that diverge from the learning task. These serve as the off-task labels for the classification task. The objective of the study is to identify whether students use off-task behaviors to regulate their emotions e.g. to alleviate frustration.
From the above discussion, we can surmise that an ATS first infers the affect of the students through various input modalities and then responds with various tutoring actions that alleviate or regulate the students’ negative affect. The tutoring actions or responses vary though across each of the above implemented ATS and there seems to be no uniform tutoring response even for the same exhibited affect e.g. frustration. There is in fact a critical lack of research in the area of ATS that explores the theories and methods of integrating affect and learning within the learning process.
The effectiveness of the tutoring response is measured by the degree by which the students’ negative emotion is alleviated. To achieve this, the students will have to ‘learn’ to cope with or regulate the negative emotions. The negative emotions direct students to focus inwardly on the emotions and draw attention away from the learning tasks (Hascher,
2010). In Kort, Reilly, and Picard (
2001)’ s four quadrant spiral learning model, they postulated that external help such as scaffolding can sustain and motivate a learner to cope with and overcome a learning stage characterized by frustration and misconceptions. Within the field of psychology, this process of “coping” refers to efforts to master, reduce or tolerate the demands created by stress (Weiten, Dunn, & Hammer,
2011). The coping strategies can be further segregated into 3 main types – appraisal-focused, emotional-focused and problem-focused.
Appraisal-focused coping strategy involves changing the way one thinks about the problem. One can possibly dispute or challenge one’s irrational assumptions. The key is to replace the negative thinking with rational analyses, potentially reducing stressful situations to less threatening ones. Another technique is to de-stress with humor. Humor buffers the effect of stress, allowing people to bounce back from negative situations faster and also helps to promote social interactions and support.
Emotional-focused strategy involves releasing one’s pent-up emotions through expression of the emotions. The techniques include managing hostility towards others and being more forgiving. It can also possibly involve the use of relaxation techniques such as mediation. In addition, effort and task difficulty attributions (Weiner,
1972) can also help in mediating the negative emotions experienced by students. Failure to do well in a learning task can be attributed to insufficient effort (a controllable factor) instead of one’s intelligence which is less malleable.
Lastly, the problem-focused strategy involves the use of a systematic problem resolution process of seeking help and acquiring new skills for dealing with the problem e.g. time management skills. By exercising self-control through techniques such as operant conditioning (which involves use of rewards and punishments), one can also better cope with the stress.
In this study, appraisal-focused, emotional-focused and problem-focused strategies are utilized for the regulation of negative emotions experienced by the students.