Akademik Faaliyetler
Proje ekibi tarafından geliştirilen sistem bağlamında yapılan çalışmalar aşağıda sunulmuştur.
Başlık: Zeki Öğretim Sistemlerinde Hibrit Bir Modelin Tasarlanması ve Geliştirilmesi
Yayın Türü: Doktora Tezi
Kurum / Enstitü Adı: Hacettepe Üniversitesi Eğitim Bilimleri Enstitüsü
Yazarlar: Furkan Aydın
Özet
Bu çalışma kapsamında Zeki Öğretim Sistemi (ZÖS) öğrenci modelinin tasarlanması ve geliştirilmesi amaçlanmıştır. Bu hedef doğrultusunda öğrencilerin bireysel özelliklerine göre farklı öğretimsel destek ihtiyaçları söz konusu olabileceğinden hareketle çalışmada öğrencilerin gereksinimlerine dayalı bir ZÖS tasarımı nasıl olmalıdır sorusuna gelişimsel araştırma (developmental research) yönetimi ile cevap aranmıştır. Öğrenci modeli için ilk aşamada öğrencilerin ihtiyaçları incelenmiş ve ZÖS’ün bileşenleri alan yazının taraması ile birlikte ortaya konulmuştur. Yapılan çalışmada bir yazılım gerçekleştirme söz konusu olduğu için ortaya konulacak ZÖS’ün tasarım ve geliştirilmesine yönelik yazılım geliştirme modellerinden Hızlı Prototipleme Modeli temel alınmıştır. Çalışma kapsamında öğrenci modeline odaklanılsa da ZÖS’te yer alan tüm bileşenler işe koşulmuştur. Bu sebeple geliştirilen ZÖS’ün alan modelinde ayrık bilgi bileşenler yaklaşımı, öğretici modelinde ise işbirlikçi filtreleme yöntemlerinden Ağırlıklandırılmış Jaccard tekniği kullanılmıştır. Öğrenci modelinde ise Bayes, Katman ve Stereotip (BaKaSt) öğrenci modellerinin bir arada bulunduğu hibrit bir öğrenci modeli ortaya konulmuştur. Geliştirilen ZÖS, BaKaSt olarak adlandırılmıştır. BaKaSt’ın değerlendirilmesi amacıyla deneysel bir araştırma yürütülmüş olup araştırmaya 58 kişi deney, 46 kişi kontrol grubu olmak üzere toplamda 104 lisans öğrencisi katılmıştır. Kontrol grubunda öğrenciler soru çözerken karşılaştıkları zorluklarda öğretimsel desteği kendilerinin seçtiği sistemi kullanmışlardır. BaKaSt’ı kullanan deney grubunda ise öğretimsel destek sistem tarafından sunulmuştur. Yapılan deneysel işlemler sonucu BaKaSt’ı kullanan öğrencilerin alternatif sistemi kullanan öğrencilere göre akademik başarısının daha yüksek olduğu tespit edilmiştir. Ayrıca BaKaSt’ı kullanan öğrencilerin alternatif sistemi kullanan öğrencilere göre daha fazla öğretimsel destek aldıkları ve daha az yardım arama davranışlarında bulundukları belirlenmiştir.
Başlık: Zeki öğretim sistemleri üzerine yazılan tezlerin incelenmesi: Türkiye örneği
Yayın Türü: Makale
Dergi Adı: Eğitim Teknolojileri Kuram ve Uygulama
Yazarlar: Furkan Aydın, Halil Yurdugül
Özet
Bu araştırma Türkiye’deki yükseköğretim kurumlarında zeki öğretim sistemleri üzerine yapılan tezlerin incelenmesi ve araştırmacılara bu konu ile ilgili bir bakış açısı sağlamayı amaçlamaktadır. Bu amaç doğrultusunda alan yazın taraması “zeki öğretim”, “zeki web tabanlı”, “zeki öğretim sistemleri” anahtar kelimeleri ile Yükseköğretim Kurulu (YÖK) tarafından doktora ve yüksek lisans tezlerinin saklandığı tez veritabanında yapılmıştır. Araştırma kapsamında 2021 yılına kadar yayınlanan ve belirlenen ölçütlere uyan 25 tez incelenmiştir. Tezler önceden belirlenmiş on ölçüte göre analiz edilmiştir. Bu ölçütler ise a) yıl, b) eğitim seviyesi, c) içerik, ç) çalışma grubu, d) öğrenci modeli, e) öğrenci karakteristiği, f) karar vermede kullanılan algoritma, g) öğrenene sunulan geri bildirim ve ğ) tasarım etkililiğidir. Analiz sonucunda bir takvim yılı içerisinde en fazla tezin 2013 yılında yazıldığı, son beş yılda (2016-2020) ise dört tez yazıldığı tespit edilmiştir. Tezlerin dokuz tanesi doktora, 16’sı ise yüksek lisans düzeyindedir. En çok Matematik dersine yönelik ZÖS geliştiği görülmüştür. Çalışma grubu olarak en çok lisans öğrencileri tercih edildiği bir diğer bulgudur. Tezlerde ZÖS’ün en çok e-öğrenme kavramıyla birlikte ele alındığı; daha sonra sırasıyla öğrenme alanı, öğrenci modeli, yapay zekâ, uzman sistemler ve uyarlanabilir öğrenme kavramlarıyla birlikteliği tespit edilmiştir. Öğrenci modeli olarak en çok katman modeli kullanılmıştır. Öğrenci karakteristiklerinden en çok bilgi düzeyinin kullanıldığı görülmüştür. Karar vermede en çok kural tabanlı yaklaşımın kullandığı belirlenmiştir. Geri bildirim türü olarak en çok ipucunun kullanıldığı tespit edilmiştir. Bunlara ek olarak, tezlerin büyük bir çoğunluğunun, tasarım etkililiğinin değerlendirilmediği ya da raporlanmadığı belirlenmiştir.
Başlık: Finding traces of motivational beliefs in learning analytics supported massive open online courses.
Yayın Türü: Konferans Bildirisi
Konferans Adı: Association for Educational Communications & Technology (AECT) 2021
Yazarlar: Mustafa Tepgeç, Fatma Gizem Karaoğlan Yılmaz, Ramazan Yılmaz, Furkan Aydın, Sema Sulak, Halil Yurdugül
Özet
This study aims to reveal the links between self-reported psycho-educational variables and learner interactions in the Massive Open Online Courses (MOOCs). The study also aims to introduce the MOOCs system and its components, which were developed as part of a large-scale project. It is frequently emphasized the importance of establishing a link between log data captured in online learning environments and the factors affecting learning commonly used in the field of learning sciences (Er, Dimitriadis, & Gasevic, 2020; Mangaroska & Giannakos, 2019; Sedrakyan et al., 2020). This study will provide empirical findings for the gap in this issue. The study also offers attendees a theoretical perspective on what real-time log data means in instructional contexts, as well as insight into the design and instructional use of such environments, especially for teachers and designers.
MOOCs allow learners from all over the world to specialize in specific subjects and receive certifications, regardless of their location or educational institution. Learners can progress at their own pace without being constrained by a curriculum, use the materials indefinitely, study independently of time and location, and constantly test themselves in such environments. Moreover, learners can experience all of these for little or no charge. Learners, on the other hand, leave a massive amount of data in these environments. The field of learning analytics is concerned with intervening in the learning process based on these data. Two main information sources are frequently considered for data-driven interventions. One of these sources is previous learning experiences. In other words, it's the information from learners who have experienced similar learning conditions before. The literature is another source of data that can be used to develop instructional interventions. The design principles proposed in the literature can serve as the basis for instructional interventions. At this point, this study will provide empirical findings for the community of education who will evaluate the literature as a data source.
This research also aims to introduce the MOOC platform developed by the authors using HTML, PHP, JavaScript, and MySQL. Learners can select content types based on their preferences in the developed system. In addition to video lectures, the system offers textual presentations, e-books, infographics, and alternative videos as instructional sources. Learners can utilize descriptive analytics to see their current situation, as well as predictive analytics to evaluate their next learning experience. Learners' entire interaction data are stored, and the system presents the required information via learning analytics dashboard. To do this, the system employs both classification and clustering algorithms in this process. The system also functions as an intelligent tutoring system while the learners are taking the competency test. The system will direct the learner to this module if he fails the competency test without help. The goal of this module is to improve learners’ problem-solving skills by using scaffolding strategies like hints and worked examples. Detailed information about the system's features will be provided during the session. However, as previously stated, the study's main goal is to figure out how the learners' interaction data in this system corresponds with psycho-educational variables.
The study involved 156 students from three different state universities during the 2020-2021 academic year, 20 of whom were graduate students and 136 undergraduate students. Within the context of the Statistics course, students used the developed MOOC platform for 8 weeks. Learners’ interactions with the system were recorded and analyzed during this period. In addition, the scale developed by Pintrich, Smith, Garcia, and McKeachie (1991) was used to assess students' self-reported motivational beliefs. Within the scope of the study, the data of 90 students who responded to this scale were analyzed. The motivational beliefs dimension of the scale was used in the context of the study, but not the learning strategies dimension. In this context, intrinsic goal orientation, extrinsic goal orientation, task value, control of learning beliefs, self-efficacy for learning and test anxiety were examined.
The main goal of this study is to discover the relationship between student interactions and motivational beliefs in the online environment. Test anxiety and intrinsic goal orientation were found to have an impact on the success of competence tests as a result of the research. In other words, it has been discovered that learners' curiosity about the system and its contents, combined with their test anxiety, has an impact on their performance in the system. Furthermore, students' level of appreciation for the system's content was found to be related to their belief in control over learning, or in other words, the belief that their own efforts affect their learning outcomes. The duration of the learners' navigation in the system was found to be significantly affected by their task value, which is their perception of the system's importance and benefit, as well as the tasks it contains. According to the findings, while undergraduate and graduate students performed similarly on competency tests, graduate students' task value was significantly higher, while undergraduate students' test anxiety and extrinsic goal orientations were superior to graduate students. In addition to these variables specified in the study, the relationships of motivational beliefs with various variables were also evaluated.
Başlık: Learning analytics based feed-forward: Designing dashboards according to learner expectations and lecturer perspectives.
Yayın Türü: Konferans Bildirisi
Konferans Adı: Association for Educational Communications & Technology (AECT) 2021
Yazarlar: Mustafa Tepgeç, Fatma Gizem Karaoğlan Yılmaz, Ramazan Yılmaz, Sema Sulak, Furkan Aydın, Halil Yurdugül
Özet
Online learning has become increasingly prominent in educational settings over the past years, especially in the era of Covid-19 pandemic. However, online learning environments work in favour of autonomous learners, in other words, self-directed learners. Even if online courses in which contains are based on a well-structured instructional design, these environments are disadvantageous for learners who do not take responsibility for their own learning (Toro-Troconis, Alexander, & Frutos-Perez, 2019). Therefore, such learners need guidance in these environments. Studies emphasizing the guidance in online learning environments form the basis of learning analytics conceptualized for the first time in the previous decade.
Learning analytics provide valuable information for learners and instructors by combining and analyzing learners' historical data during the learning experience. The most common way of employing this information is in the form of learning analytics dashboards (LADs). Even though there have been various studies on the use of LADs in educational settings over the last decade, most of them lack theoretical basis. It is frequently emphasized that learning analytics capabilities are limited without theoretical framework and contextual interpretation of the collected data (Er, Dimitriadis, & Gasevic, 2020; Mangaroska & Giannakos, 2019; Sedrakyan et al., 2020). Therefore, bridging the gap between learning analytics and learning sciences are essential. The theoretical background in feedback design has shed light on this gap in recent years and the phenomenon has been called “learning analytics based feedback” (Lim et al., 2020; Pardo et al., 2019; Sedrakyan et al., 2020; Yilmaz, 2020). This study employs design principles of feedback as well as feed-forward(*) to design well-structured LADs. Feed-forward, unlike feedback, has received little attention in educational research. On the other hand, the feed-forward as an intervention strategy should be reviewed now, as the prediction of future performance with educational data mining approaches is more advanced.
Considering the theoretical perspective mentioned above, this study aims to determine what kind of information learners expect through LADs in online learning environments and to design LADs in accordance with these expectations. The study consists of four phases: (1) identifying learner needs and expectations in online learning environments; (2) designing LADs by taking into account data sources, which are the findings from the first phase and best practices in the field of learning analytics; (3) refining the design according to lecturers' views; (4) demonstrating LADs that was refined in the third phase and determining whether the design is appropriate for learners needs and expectations.
This study was designed as a qualitative, exploratory study to gather in-depth information for designing LADs. The study may be described as design-based research as it requires a comprehensive and iterative design and development process. First, for the purpose of determining the needs, an inclusive open-ended question was asked, and the learners' answers were evaluated through content analysis (first phase). Following the removal of two incomplete responses, the responses of 22 undergraduate and graduate students were considered for further analysis. The LADs’ draft data visualizations were then designed based on the learners' needs and best practices in the field of learning analytics (second phase). The draft data visualizations includes the information in the following sub-categories: (1) description and aim of data visualizations; (2) logs and metrics to be taken into account; (3) calculation methods required for the information to be presented; (4) right chart/graph types for data visualization. Then, for each draft data visualization, lecturers expressed their opinions on these categories as well as rated their potential contribution to learning outcomes on a scale of 1 to 5. During this phase of the study, six lecturers took part, four of whom were experts in educational data mining. Learners who took part in the first phase of the study gave their opinions about refined design in a similar manner (fourth phase) after the design was revised according to the lecturer's views. These learners, unlike the lecturers, evaluated only the first and fourth from the categories listed above. They also gave a 1 to 5 rating to the data visualizations chosen by experts for their suitability to their needs and expectations.
Due to the nature of the study, the outputs of each phase were considered as inputs to the next phase. It is critical to take into account the needs, expectations and views of all stakeholders to design useful LADs. Considering the learners’ needs and expectations, the findings related to feedback show that the LADs should provide these information to learners; a general overview of their progress in comparison to peers (norm-referenced feedback), which type of content (video, illustrations, etc.) is suitable for their characteristics, support for planning and organization of learning experience, in-depth information about their performance in the exams. In addition, the findings related to feedforward show that LADs should provide these predictions to learners; whether they will be successful in the relevant course before it takes place, whether sufficient time has been allocated for passing the course, when the course will be completed. 32 draft data visualizations were developed based on the learners’ expectations and the best practices in the field. 20 of them were classified as feedback intervention, while the remaining 12 were classified as feedforward intervention. 8 data visualizations were eliminated as a result of the lecturers’ evaluation, and a total of 24 data visualizations were revised and presented to the learners. The outputs of this study will contribute to instructional designers to gain insight into employing learning analytics according to learner characteristics, researchers to gain a perspective for the theoretical gap between learning analytics and learning sciences, and how teachers can make interventions in line with students' needs and expectations in online environments.
Başlık: Students’ preferences and views about learning in a smart mooc integrated with intelligent tutoring
Yayın Türü: Konferans Bildirisi
Konferans Adı: International Conference on Cognition and Exploratory Learning in Digital Age (CELDA) 2021
Yazarlar: Fatma Gizem Karaoğlan Yılmaz, Mustafa Tepgeç, Cennet Terzi Müftüoğlu, Sema Sulak, Muhittin Şahin, Furkan Aydın, Ramazan Yılmaz, Halil Yurdugül.
Özet
Massive Open Online Courses (MOOCs) have become widespread all around the world since their conceptualization, both in terms of the number of students enrolled and the number of courses available. Some issues or learner needs in these environments, such as accreditation, quality of assessment and scaffolding for left-behind learners, have also surfaced as a result of rise in popularity. Profiling learners, using scaffolding strategies, structuring the assessment in a dynamic and effective manner, monitoring performance, and providing feedback/feedforward based on learning analytics are all expected to be valuable solutions to these issues. In this study, it is aimed to describe student views on the MOOC platform, which has the above-mentioned features and was designed and developed according to the AGILE software development model. Participants of this case study consist of 53 undergraduate students from three different universities. The data was collected using a questionnaire and semi-structured form both of which were developed by the researchers. The findings obtained were considered under these themes: benefits, disliked aspects, preferences in different learning contexts, ease of use, features open to improvement. Findings based on both quantitative and qualitative interpretations are presented on each theme. The findings of this study are limited to student views. In the later stages, authentic usage circumstances can be presented by taking into log data as a data source.
Başlık: Using adaptive mastery testing in assessment management systems
Yayın Türü: Konferans Bildirisi
Konferans Adı: International Conference on Cognition and Exploratory Learning in Digital Age (CELDA) 2021
Yazarlar: Muhittin Şahin, Furkan Aydın, Sema Sulak, Cennet Terzi Müftüoğlu, Mustafa Tepgeç, Fatma Gizem Karaoğlan Yılmaz, Ramazan Yılmaz, Halil Yurdugül.
Özet
The use of technology for teaching and learning has created a paradigm shifting in learning environments and learning process, and also the paradigm shifting has also affected the assessment processes. In addition to these, online environments provide more opportunities to assess of the learners. In this study, the Adaptive Mastery Testing (AMT) system in Assessment Management System was designed and developed in which students can test themselves, recognize their strengths and weaknesses, and determine their learning objective based competencies. AMT environment is structured in accordance with the rapid prototyping software developing model. In this environment, there are questions for the students about four learning objectives, which are among the basic subjects of the Statistics course. AMT environment consists of presentation, assessment, domain and learner model. Pilot implementation was carried out with 98 undergraduate students. In order to evaluate of the environment; number of tests taken, number of correct answer, number of wrong answer and number of total answer data were used. According to the findings, it is seen that most of the students are masters in the learning objectives presented to them. In addition, it was found that half of the students took an average test to become a master. In other words, half of the learners participating in the study were determined as masters by the system in the first test they took for each learning objective.
Başlık: Smart MOOC Integrated with Intelligent Tutoring (SMIT)
Yayın Türü: Makale (inceleme sürecinde)
Dergi Adı: -
Yazarlar: Ramazan Yılmaz, Halil Yurdugül, Fatma Gizem Karaoğlan Yılmaz, Muhittin Şahin, Sema Sulak, Furkan Aydın, Mustafa Tepgeç, Cennet Terzi Müftüoğlu, Ömer Oral.
Özet
With an accelerating transition from Web 3.0 technologies to Web 4.0 technologies today, Learning Management Systems (LMSs) undergo a transition to new generation learning systems called LMS 3.0, powered by machine learning based on educational data mining. Further, the Massive Open Online Courses (MOOCs) are rapidly gaining popularity; such that over 1000 universities established their own MOOCs in 2018 alone and introduced more than 11.400 courses on these systems in one year. MOOCs are a type of LMS, but it seems that the influence of the instructor in these systems is minimal or simply lacks. These systems present the learning content and materials to all learners attending the course in the same way and fail to offer individualized instruction that recognizes the individual differences and needs of the learners. That said, it is reported that such problems can be eliminated by making the new generation intelligent learning systems. However, there are still both an ongoing search on how to make such systems intelligent and a conceptual discussion concerning them. Integrating an intelligent tutoring system (ITS) with learning analytics, this study seeks to design and present the framework of an intelligent tutoring system with open access that a) identifies the learning needs of learners through adaptive mastery testing and guide learners based on these needs, b) overcomes learning deficiencies, monitors learners’ interactions with content through learning analytics and offers suggestions, c) supports learning with dynamic Assessment processes and d) tests learners’ learning competencies.