The creation of online courses and content for e-learning has become a big industry, but good education is becoming more and more expensive. Hence, universities like MIT undertook to create OCW (Open Courseware) in order to make their high-quality education available to students all over the world who are unable to study at these universities. This was the beginning of the Open Educational Resources (OER) movement. In India, we have the National Program on Technology Enhanced Learning (NPTEL), through which video lectures by well known IIT professors are made available freely over the Web. In this article, I give information about a bold experiment in distributed learning by well known professors from Stanford University, and also share some of the opinions I have formed on the basis of my own experience with these courses.
From October to December 2011, Stanford professors offered free online courses on three topics: Introduction to Artificial Intelligence (basic and advanced track) https://www.ai-class.com/, Introduction to Machine Learning (basic and advanced track) http://www.ml-class.org/course/auth/welcome , and Introduction to Databases http://www.db-class.org/course/auth/welcome . All the course content was offered through small video clips uploaded on the course website. By using a login ID and password, participants could assess the content together with assessment exercises, programming assignments, and exams. More than 1,00,000 students enrolled from all over the world for the AI course, and around 70,000 students for the Machine Learning course. There was no restriction on participation based on educational qualification; however, some pre-requisite video links were given as a reference for those who wanted to prepare themselves for the class. In the Machine Learning course, the required mathematical concepts were covered in separate videos, but their study was optional rather than mandatory. Ages of participants ranged from 18 to 72 years.
Watching only video lectures, without interaction with instructors or assessment of work, is a passive kind of learning. Many students have a very short attention span; watching an hour-long video to learn one concept has been found to be not very useful. For getting a certificate through online courses, there needs to be a rigorous formative and summative assessment. The Stanford courses were designed according to teaching methods suggested by educational psychologists. For example, in the Artificial Intelligence class every module contained 2 to 3 units, and each unit was divided into 25 to 30 small video clips, each from 30 seconds to 3 minutes long (one week was given for completing a module). The clips introduced the concept by asking questions based on previous knowledge, then taught the concept, and finally assessed the student's performance. For basic track, participants needed to view the video clips and do review exercises; for advanced track, participants were expected to complete homework as well as mid-term and final exams. Formative assessment during the content video clips was supported by immediate feedback designed to improve the student's understanding of the material. At the end of each unit (there were 2-3 units per week), there was a summative assessment, each containing 6-9 questions. Interaction of participants with the course website was tracked. There were several forums available for discussing the course's concepts and participants' difficulties. Although assessment was for objective questions, explanatory videos (feedback) were also available for formative and summative assessments. A large team of assistants from the relevant departments of Stanford University was supporting the entire process round the clock during the 16 weeks of the course schedule, to make this experiment successful. Mistakes found in the videos were rectified, once brought to the attention of the team. Sufficient time was given for learning the content and completing the homework assignments. Sometimes the server crashed, but it would be quickly repaired, and submission dates accordingly extended. For AI, 23% of participants made it through and acquired a certificate of accomplishment from the Stanford professors who were involved in conducting the courses. The mark distribution for the advanced track was 30% for homework, 30% for the midterm exam, and 40% for the final exam. A total of 72 hours were given for completing the midterm and final exams. Participants could submit their solutions to the system many times before the deadline. For the Machine Learning advanced track, participants needed to do programming exercises in order to implement the taught algorithms. During this formative assessment, detailed feedback was provided so that students could learn from their mistakes. Review exercises were designed to check the student's understanding, and the student had a hundred chances to complete them. Participants were encouraged to get a 100% score in each review exercise; a certificate of accomplishment was given to those participants who scored above 80%.
Excellent as these courses were, I personally felt that there were two inadequacies. Firstly, I sometimes had difficulty understanding the accent of some of the professors. Secondly, some of the examples and games that were used may be said to be culture specific (I was unfamiliar with some of the games), but I was able to overcome this problem by searching the Internet. In preparing these online courses, the possibilities of information technology were used to the full: in the AI course, there used to be weekly live office hours during which professors could meet participants online and interact with them. It was not exactly a real-time synchronous learning experience, but it represented a good combination of synchronous and asynchronous e-learning. Some participants complained that in the forum of the Machine Learning course only questions which had high ratings received answers, which limited the interaction. There are various reasons why not all of the enrolled participants were able to finish, but there can be no doubt that anyone who managed to complete the course was hugely benefited by it. One needs commitment and self-motivation to complete such online courses with rigor. Other than actually studying with them in their regular university courses, this is the best way to learn from these experts in their fields. The exercises were quite original, and answers were not found on the Web, which is a key factor in maintaining the quality and objectivity of the assessment.
An introductory course on programming a Web search engine and a course on programming a self-driving car are going to be offered by Prof. Sebastian Thrun over about seven weeks beginning on 20th February 2012. For the search engine programming course, no prior knowledge of programming is needed. The website for enrolling in the courses is: http://www.udacity.com/ . The plan, in due course, is to create a virtual university which will exploit the full capacities of information technology and the World Wide Web to offer high quality education in an inexpensive way. Lecturers and teachers will also benefit from these courses, since they will offer them another arena in which to hone their own skills.
