Let Them Eat Code

Education in the digital age
Salman Khan came to education by accident when tutoring his cousins via YouTube using a digital blackboard and colored pens. Among the millions of videos shared every day on the website, his were noticed and became massively popular. He conveys his love of all things mathematical as the warm, funny, intelligent mentor everyone would love to have.

In 2009, he founded the Khan Academy, a non-for-profit organization supported by major donors such as the Gates Foundation and Google. It offers a completely free library of educational videos along with practice exercises. The material covers a range of topics, from calculus to physics, finance and history, in 10 languages. The institution developed analytics software to track user progress, predict her/his prior knowledge and fine-tune the presentation of the content. 3.5 million students a month use the well-structured series.

In the summer of 2011, Stanford University took the experiment in a different direction. The initiative to offer three free higher education classes was announced via a single email sent to a distribution list of 1,000 recipients. Renowned professors Peter Norvig and Sebastian Thrun taught the Artificial Intelligence class. The two other, Introduction to Databases by Jennifer Widom and Machine Learning by Andrew Ng were of equal quality even though they have not received the same publicity. All professors offered the same course on campus to full-time students.

The interest was unprecedented. 160,000 students registered from 190 countries for the AI course (I will focus on it for the purpose of this article). New units were posted weekly with practice quizzes. Captions were provided for most videos and translated into 13 languages, thanks to volunteers. For students on the advanced track (a basic option was also available), assignments were due, via selecting radio buttons and entering numbers in text fields, as well as a mid-term and a final exam. After the deadlines, a server-side script run and within minutes, students could see their result. On-campus students had additional programming assignments (not possible with the site current set-up but considered for future courses). 30,000 students remained actively engaged until the end of the term and took the final exam. The online students who completed the course received a statement of accomplishment signed by the professors, not the university. 

The class was rich in content and challenging. Each unit was the introduction to a deeper field of study, such as problem solving, probability, machine learning, hidden Markov network, game theory, computer vision, robotics and natural language processing. Norvig and Thrun recorded themselves explaining concepts while writing key elements on paper real-time. The material, completely new to most students, was presented in small chunks. In many instances, quizzes and homework assignments required reflection to answer open-ended questions, a very appropriate approach to Artificial Intelligence which is, after all, “the discipline that you apply when you want to know what to do when you do not know what to do” (unit 1-10).

Stanford did not invent the concept of massive open online courses (MOOC) but the reputation of its professors as well as the topics taught contributed to its popularity. Other US institutions have made their courses available in the past but this is the first time that remote viewers are invited to submit homework and take exams for a university-level course, therefore becoming active students. Such initiative already exists in other countries like the Open University in England.

What is the secret sauce?
Free education is of course the first appeal to most remote students. Khan Academy offers a complement for many households who cannot afford private tutoring. The Stanford’s tuition of $40,000 per year, is out of reach for many American students and unthinkable in most countries around the world.

The reputation and credentials of the educator establishes trust. Stanford is known as one of the best computer engineering school in the world. Being taught by the head of Google Research, formerly head of Computation Sciences Division at NASA Ames, along with a research professor of Computer Science at Stanford known for creating a car that drives itself, is a powerful motivation to perform better. More importantly, teachers demonstrate passion for their field of work, through their life dedication or in the way they communicate to students. Salman Khan describes math as “the language for thinking” and demonstrates what an expert thinker he is along with the great fun he gets from working with numbers.

But this is not the whole story. There are many testimonies using the Khan method succeeding where the traditional academic model failed. It is too early to determine the academic results of Stanford online students compared to their campus classmates, although the exponential outreach is evident. Salman introduced a new form of teaching, borrowed by Stanford, radically different on some very essential points.

Watch any of the videos for yourself. The format is simple and similar to the classroom experience. There is a blackboard or a piece of paper. The teacher presents the material while writing the explanation down. It feels like it is happening real time and unfolding before your eyes. The intimacy of the computer screen gives the sense that the course is made just for you. It is quite different from watching the recording of a past class attended by someone else.

In a classroom, you may ask your teacher for clarification once but, if you still don’t understand, you will most likely not dare ask again or if you do, she/he may respond in an abrupt fashion, concerned not to run out of time or bore the rest of the class. Online, you learn at your own pace. You can rewind the video as many times as you need to. Speed is not necessary synonym with intelligence, more to the point, absorbing complex material slowly, or slower than others, is not equivalent to lacking abilities.

Learning is cumulative knowledge. If you did not do well in school at a given time, you probably missed some of the founding blocks on which other areas are built. The domino effect may resolve in performing poorly in a whole area and perhaps abandoning further studies all together. A well-structured online lesson is modular and presented in small bits. If complementary or additional information is needed, you can pause and jump to another section or search outside sources. Online access provides an opportunity to fill out those missing blocks and for many to be educated later in life.

In addition to the lectures, the format includes frequent practice exercises and quizzes. They do not have the stigma of an exam. In fact they are game-like and making mistakes is part of the play. Aristotle, one of our oldest professors said, “what we have to learn to do, we learn by doing”. Indeed, the active act of problem solving is empowering and triggers stronger memory retention. From it, comes the joy of learning.

From the perspective of the teacher, the class is more engaging. If students watch lectures at home, class time can be dedicated to tutoring and in-depth conversations. The group experience takes on a different form with peers teaching one other, making them more involved and encouraged to think. Time will tell if professors are willing to relinquish a heavily monitored classroom and trusting of students’ own pace and curiosity. And if students have the motivation.

Technology and social networking
This new form of online teaching is one example of the tremendous possibilities of technology and connectivity and the impact of our largest public institution: the Internet.

As Mr. Khan colorfully stated, “YouTube is not just for cats playing the piano”. Its mission statement is to give everyone a voice. The video-sharing service claims that the equivalent of 8 years of content is uploaded every day for an audience of hundreds of millions of users all around the world. Both Khan and Stanford uses YouTube embedded videos. It is unfortunately an issue for countries like China because of government’s censorship.

Spontaneous study groups erupted in various countries and cities, via the use of Meetup, a free service to organize local groups.  The forums were an important factor to provide a broader sense of community and re-create the social bonding students experience in a physical classroom. They helped many stay engaged when the class became difficult. 

An independent AI study group was established on Reddit, the free service to create communities around specific areas of interest. “Redditors” entries are constantly sorted and displayed on the page based on others’ feedback (upvoting or downvoting) and the total vote count. When students posted questions or requests for clarification, many jumped in to discuss, provide help or suggest additional reading. The group behavior self-regulated quickly. Sharing and comparing grades was acceptable but bragging out of turn was not. One student posted having finished his mid-term exam in just an hour, while most took several. He was promptly scolded and voted 27 points down. There were numerous complaints early on, on the audio quality, the occasional connectivity problem with the servers (there were glitches throughout the term with under-capacity servers, not prepared for the class success) or the difficulty of the assignments, but a relative low number of trolls. Around the time of the mid-term, complaints tempered off, perhaps because the unhappy quit the class. Online, one can walk out of the room unnoticed.

Stanford provided its own Q&A community called Aiqus. Its goal was to facilitate posting questions. The student-moderators are defined by their quantity of karma points, acquired by asking relevant questions (viewed by many) or providing useful answers. They controlled that posts were questions only within the honor code (no direct answer to homework questions). Stanford was very responsive and made adjustments when needed, whether it concerned the ambiguity of a question, the quality of the recording or the method of teaching. It is offering new classes this semester again.
  
Impact on future education
This is perhaps the first chapter in a profound and universal change in education. Online quality education challenges private institutions and the very definition of the classroom. The United States are currently experiencing a devaluation of education while it is gets more expensive every year. Universities may need to bring their tuition down, reflect on the definition of elitism and consider a more equitable system.

Families are no longer willing to pay for a college degree if there is a cheaper alternative of the same quality, in particular if credentials no longer guarantee employment. In the work place, companies may adopt new standards to access the skills of potential employees.

Most importantly, this may stimulate curiosity and thirst for knowledge, not only for students but also for people of all ages around the globe.

See you in class.

Véronique Brossier
(Stanford Artificial Intelligence online student 2011)