Follow the talks of our Keynote speakers in these recordings :
Academic Analytics – Analysis and Mining of Educational Data to Support
Sabine Graf, Athabasca University, Canada
Offerings of online courses and even entire online programs are continually increasing in numbers and popularity. To deliver online courses to students, learning systems are used which collect and store huge amounts of data about how students behave, learn and perform in such courses. However, although such data can provide a wealth of relevant and actionable information for people involved in the learning and teaching process (e.g., educators, learning designers, students, etc.), such data are very rarely used. In this talk, I will focus on how such data can be utilized to support teaching. I will discuss the kind of information that can be extracted from such data and the potential of this data to support the teaching process. I will also introduce some of the systems, tools and plugins that have been designed, developed and evaluated within my research group. Those systems, tools and plugins use different approaches to support teaching. One of those approaches is to use artificial intelligence and data mining techniques to mine data about students’ behavior and actions in a course and provide specific, relevant and timely information to educators. Another approach is to use academic and learning analytics techniques to guide educators through extracting and analyzing relevant information from the huge amounts of data stored in learning systems and letting them conduct their own investigations into any questions they are interested in getting more information about. For both approaches, examples of systems, tools and plugins will be shown during the talk.
Machine learning for online education
Irwin King, The Chinese University of Hong Kong
With the advent of online education, students from all over the world can access quality education contents from anywhere and anytime. Due to this great convenience, we are able to collect a massive volume of learning activity data. In this talk, we introduce how machine learning techniques can improve teaching and learning for educators and students using the collected learning activity data. First, I will present a few machine learning models and algorithms for peer assessment. Peer assessment offers a scalable way to grade tens of thousands of assignments in Massive Open Online Courses (MOOCs) by asking each student to grade a subset of his/her peer’s assignment. Since these peer grades are very noisy, we need robust machine learning techniques to recover the ground truth scores or an overall ranking of all the assignments. Next, I will present machine learning models for knowledge tracing. Knowledge tracing is a task of tracing evolving knowledge states of students with respect to different concepts as they engage in a sequence of learning activities. The goal is to personalize the practice sequence to help students learn knowledge concepts more efficiently. Lastly, I will share our KEEP (Knowledge and Education Exchange Platform) education project for teaching and learning in Hong Kong and beyond.
The usefulness of search results and task outcome
Prof. Dr. Pertti Vakkari, University of Tampere, Finland
In evaluating search systems there is a growing trend to complement the established effectiveness indicator topical relevance by the usefulness of search results. Usefulness refers to the contribution of search results to a larger task generating information search. Task outcome signifies the end-product of a task performance process. It is evident that the whole search process contributes to task outcome. The aim of my presentation is to give a systematic account of the characteristics and results of studies focusing on the usefulness of search results or task outcomes in larger tasks containing searching. I summarize how the usefulness of search results, and search outcomes are defined and operationalized in empirical studies. I categorize various types of usefulness and outcomes and various predictor types used in studies. Finally, I summarize which factors significantly predict the usefulness of search results as well as search outcomes.