Analysis of The Applications of Cognitive Computing

Education and learning stand out among many application areas of cognitive computing due to their practical appeal as well as their research challenge. There are various applications of Learning Analytics, Educational Data Mining and Cognitive systems present, which can enhance humans learn as well as Personalized learning and can vastly improve the quality of learning and education.

Introduction

Cognitive computing is a mixture of computer science and cognitive science – that is, the understanding of the human brain and how it works. By means of self-teaching algorithms that use data mining, visual recognition, and natural language processing, the computer is able to solve problems and thereby optimize human processes. Humans learn from a very small number of examples. Similarly, we can devise a computational framework for the same task if we are allowed to use a large number of examples. More specifically, we can “train” a machine learning algorithm with a large number of labeled examples, represented by various features. . In some applications these features are handcrafted while in others they are automatically discovered by the algorithm itself. The important requirement here is a large amount of data, termed as Big Data.

Educational Data Mining (EDM) and Learning Analytics (LA) are two major areas of cognitive computing dealing with education and learning. ApplicationsHere are some applications of cognitive computing and to education and learning. Many of these may have appeared in the EDM and LAK conferences, which are the premier venues for publishing research on EDM and LA, respectively.

As an application of cognitive computing, we can also consider pedagogy, technology, human judgment, social factors, and various contextual elements.

Intelligent Tutoring Systems

Since the 1970s, ITS have been at the forefront of artificial intelligence research with a diverse array of application areas ranging from physics and mathematics to adult education and nurse educationOne of the most influential cognitive architectures behind ITS is Anderson’s Adaptive Character of Thought-Rational (ACT-R) theory. The central tenet of the ACT-R theory is that human cognition is the result of interactions among numerous small, indivisible units of knowledge in certain ways. The ACT-R theory gives the details of how these knowledge units interact. ACT-R is not an abstract theory of human cognition. It is rather a concrete framework similar to a programming language. Closely connected to improving an ITS is evaluating its adaptive tutoring feature, which is very much applicable to ITS for programming. Traditionally, using evaluation schemes based on machine learning, such as the Performance Factors Analysis (PFA) cognitive model.

Clustering and Student Modeling

Clustering is a common technique in EDM for aggregating student data in order to examine student behavior. This technique improves the stability of clustering for noisy data. In addition, It can be incorporated into any ITS as a black box. Another central problem in EDM is student modeling. One such work is on modeling learning curves. Using data from Duolingo (2016), Streeter (2015) has used probabilistic mixture models to capture the learning curves of language learners. Streeter’s research work generalizes knowledge tracing and offers an elegant probabilistic model for modeling learning curves. The parameters of the model have been learned using the well-known expectation – maximization algorithm. Based on the large-scale Duolingo dataset, the mixture model outperforms many of the previous approaches, including popular cognitive models like Additive Factor Model (AFM) and PFA.

Predicting Student Performance

Predicting student performance based on various factors has been a popular line of research in EDM. Traditionally, student performance has been measured in specific disciplines or topics, such as algebra, programming in Java, or even as specific as learning fractions in mathematics. Tomkins et al. (2016)’s research answer some of the questions on domain-specific student performance using their case study of a high school computer science MOOC. The MOOC is a separate course with its own evaluation, but the students taking the MOOC ultimately take the computer science Advanced Placement (AP) exam. As a result, there are two performance measures: one coming from the MOOC and the other from the AP exam. It has been empirically observed that a student’s score in the AP exam is a better predictor of the student’s future success than the student’s performance in the MOOC. One such factor is coaching. Many of the students received coaching while taking the MOOC, while others studied independently. The ones that were coached, showed a greater level of activity in the MOOC’s forum through questions, answers, and other contributions. The coached students performed better than the independent students in the MOOC. However, on the actual AP exam, the independent students scored higher.

Affect Detection and Student Engagement

Affective states are closely related to learning and cognition. We react differently to different experiences during our learning process. Assessing affective states of students is now gaining traction within both the EDM and LA communities. Students’ affective states are closely tied to their learning and therefore to their performance. Based on over one thousand middle-school students’ interactions with a math tutoring system, San Pedro et al. (2015) find that the fine-grained variations of affect over time ultimately impact a student’s test score. For affect detection, the authors used the data of student interactions with the math tutoring system. This can be largely attributed to the affective state of delight for which the video-based detector is very much the superior between the two.