استفاده از داده های حسگر برای مدل خلاقیت دانش آموزان در یک محیط دیجیتالی

While creativity is essential for developing students’ broad expertise in Science, Technology, Engineering, and Math (STEM) fields, many students struggle with various aspects of being creative. Digital technologies have the unique opportunity to support the creative process by (1) recognizing elements of students’ creativity, such as when creativity is lacking (modeling step), and (2) providing tailored scaffolding based on that information (intervention step). However, to date little work exists on either of these aspects. Here, we focus on the modeling step. Specifically, we explore the utility of various sensing devices, including an eye tracker, a skin conductance bracelet, and an EEG sensor, for modeling creativity during an educational activity, namely geometry proof generation. We found reliable differences in sensor features characterizing low vs. high creativity students. We then applied machine learning to build classifiers that achieved good accuracy in distinguishing these two student groups, providing evidence that sensor features are valuable for modeling creativity.

There is a general consensus that creativity entails a product, idea, or process that is novel and useful ( Amabile, 1996 and Mayer, 1999). Given this definition, it is not surprising that creativity is at the core of societal advancement. However, it is important to remember that creativity is present “not only when great historical works are born but also whenever a person imagines, combines, alters, and creates something new, no matter how small” ( Vygotsky, 2004). Educational activities therefore afford many opportunities for creativity. Unfortunately, students have become less creative rather than more in recent years, as indicated by a 2011 meta-review published in the Creativity Research Journal ( Kim, 2011). Certainly, creativity entails many challenges, such as persevering through impasses, attacking a problem from multiple perspectives, maintaining positive affect in the face of failure, dealing with uncertainty in open-ended problem solving, and being flexible in one’s approaches ( Amabile et al., 2005, Burleson, 2005, Csikszentmihalyi, 1990, Fasko, 2001, Gough, 1979, Hennessey and Amabile, 2009, Isen et al., 1987 and Mayer, 1989). Therefore, students need personalized, continuous support and training throughout the process of creative endeavors. However, today’s classrooms are not equipped to provide such support ( McCorkle, Payan, Reardon, & Kling, 2007). In particular, while personalized instruction has tremendous potential to improve student learning ( Cohen et al., 1982, Lepper, 1988 and VanLehn, 2011), affect ( Lepper, 1988, Picard, 1997 and Woolf et al., 2010), and metacognitive behaviors ( Bielaczyc et al., 1995, Chi and VanLehn, 2010 and Muldner and Conati, 2010), providing a human tutor for each student is simply not practical. An alternative approach, which does not suffer from this limitation, corresponds to a class of cyberlearning technologies referred to as Intelligent Tutoring Systems (ITSs). ITSs rely on Artificial Intelligence techniques to provide instruction that is tailored to a given student’s needs, thereby increasing the chances of student learning. Once implemented, ITSs can easily be deployed to provide the benefits of personalized instruction to any student equipped with a computer, a laptop, or a related digital device. ITSs have already successfully improved domain learning by tracking students’ problem-solving progress, providing tailored help and feedback, and selecting appropriate problems (Aleven et al., 2006, Arroyo et al., 2011, Koedinger et al., 1997, Self, 1998 and VanLehn et al., 2005). However, ITSs have also been criticized for over-constraining student activities and over-emphasizing shallow procedural knowledge, and therefore not properly addressing 21st century skills such as creativity and critical thinking (Trilling & Fadel, 2009). In particular, to date, very little work exists on using ITSs to support creativity. To fill this gap, our ultimate goal is to extend ITSs with personalized Intelligent Creativity Support (ICS) to scaffold creative endeavors in various digital environments. To provide personalized support in digital environments through ICS tools, the corresponding system needs information about the student that can then be used to tailor pedagogical interventions. This functionality is realized by a student model (also called a user model), which is the ITS component responsible for assessing student traits and behaviors as he or she is working on an instructional task. Typically, student models aim to collect information about students unobtrusively, without disrupting students’ work. In our case, a second requirement is that students are given freedom to explore and innovate, i.e., that their interaction with the system is not constrained so that creativity is not hindered. The latter requirement makes the modeling task especially difficult because it is well established that open-ended interaction results in a low-bandwidth situation for the model, as there is little direction information on the target states of interest ( VanLehn, 1988a). Further complicating the situation is the fact that there is limited knowledge of how to model creativity in a digital environment. To address these modeling challenges, one possibility is to provide the model with information about students’ physiological data captured by various sensing devices. The use of sensing devices for student modeling has gained a lot of attention lately because these do not require interrupting students or restraining their interaction with a system. Sensing devices are also becoming more ubiquitous and are moving out of the laboratory and into today’s classrooms (e.g., Arroyo, Cooper, Burleson, Muldner & Christopherson, 2009). The approach of using sensing devices has already been successfully applied to obtain information on student states like knowledge or affect as students interact with digital learning environments (Kardan and Conati, 2012 and Muldner et al., 2010). However, to date it has not been tested for modeling of individual student creativity. Here, we present our work exploring the utility of sensing devices for modeling creativity, including an eye tracker, an Electroencephalography (EEG), and a skin conductance (SC) bracelet. Specifically, our work addresses the following question: Can gaze, SC, and EEG information be used to create a student model that distinguishes between low creativity and high creativity students? To answer this question, we used the above-mentioned sensors to record data while students engaged in a creative problem solving task in a digital environment. Our analysis revealed reliable differences between low creativity and high creativity students. We then applied machine learning to generate empirical models of creativity – we present two models that achieved good accuracy in discriminating between the low and high creativity groups. We begin with an overview of creativity and related work on sensing devices for various modeling tasks. We then describe the study we conducted and our findings, concluding with a discussion of our results and some future work.