How to Teach

When considering the effect of technological progress on the future of education, we frequently encounter two extreme reactions: nothing has changed; everything has changed. When MOOCs (pre-recorded online courses) came in, for example:

• Some academic leaders — including the academic leader of the institution where I was teaching — pronounced that they were not an important development since good teaching requires personal interaction. Pure hype. Move on, nothing to see. (Fortunately, some of us did not heed the injunctions.)
• Others declared that traditional education was doomed; everything would go through Coursera, EdX, and consorts. Who in the 21st century wants to be sitting in a packed auditorium listening to boring lectures?

Both attitudes are absurd. Regardless of technology, the basic issues of teaching — the flow of knowledge and skills from more experienced to less experienced, and the reverse flow of lessons gained by the teacher in the process — remain; but we cannot pretend that we are in 1960 and should continue to teach like our grandparents and learn like our parents. The question is not whether to embrace new technology, but how.

What follows is my current answer, developed over many years and practiced in the last decade or two; I hope it has value for others. It is a preview of a talk meant for the forthcoming FISEE 2023 (Frontiers In Software Engineering Education) workshop in Villebrumier (near Toulouse, France, 22-25 January 2023). The event, like its predecessor in 2018, is intended to provide a forum for discussions of "Technology for education and education for technology" (the title for this year), with keynotes by Carlo Ghezzi from Milan and Armando Fox from Berkeley, all in a pleasant environment conducive to in-depth interactions. It is devoted to exactly the kind of issues discussed here and anyone interested in joining can request an invitation (see the form on the workshop's site at www.laser-foundation.org/fisee/2023/). Proceedings articles will be produced after the workshop for a new FISEE volume of Springer LNCS.

Some context

The approach I am describing here, the Trio system, has been developed over many years, with the help of many people, and particularly the experience of teaching the  Introductory Programming course of a computer science department of a major university 14 years in a row, an experience also applied in other institutions and for the past three years for a slightly different course at Constructor University. I hope the analysis and principles can be helpful to others.

I am focusing on the case of a wide-audience course that has to undergo a few iterations. The reason is that it requires a significant investment  in preparation time, which may be hard to justify for courses given just once, on a very specialized topic, or to a small group of students. I should add that this article makes no claim of proving any results; it cites no empirical study and its claims to success are of the kind that does not fare well with referees, as in "it makes a lot of sense" and "the students seem to like it" and "I find teaching much more exciting this way," not the kind of argument that will ingratiate you with, say, SIGCSE (I can imagine the reviews, in fact I could write them myself!).

With all these caveats, though, I strongly believe that for the circumstances mentioned the Trio approach is the way to teach today.

Are MOOCs a threat?

The Trio way is also a way to avoid the looming horror story for teachers and students alike, which one can illustrate with this case about which I recently heard from colleagues.

They had for a few years been successfully teaching, as external instructors, a data science course for a technical educational institution that does not have staff in that field. Everyone was happy, but this year they received notice that their contract was not being renewed: the institution had found a MOOC on Coursera that addressed the need, so the students would just be told to take it.

Such a decision seems ludicrous, and ultimately suicidal for the institution in question: how do you convince young people to go to a university if the end result is that they will be told to watch some videos online, produced by some company a continent away?

On the other hand, this case also brings to light the importance of taking advantage of MOOC-style technology.

The MOOCs, however, should be your MOOCs. If you just refer students to some other institution's products, you lose any value and credibility as an institution.

The Trio system

So here is what I do. Over the past 15 years, I have recorded MOOC after MOOC after MOOC. On introductory programming, on software engineering, on agile methods, and currently on requirements engineering based on my latest book. As a result, I have or will have material covering most of what I need for wide-audience courses. The material has to evolve, of course, so I know I will continue to record stuff forever, but that's life.

The consequence is that I no longer, for these courses, come in and teach a lecture the old way. Well, actually I do so occasionally: in the first lecture or two of the semester, to get things started; once in a while after that, to introduce some variety, or cover some topic that was not recorded;  or in response to a specific student requests, but those cases are exceptions. Instead, each class unit consists of three parts, hence the name "Trio system." The first part is a self-paced video-watching session; the second, an interactive Q&A session; the third, a lab session.

That third component is traditional: the lab session (sometimes called "ex") conducted by a teaching assistant, is devoted to explanations and exercises. That interactive part has always been and will remain essential.

The first component, the video-watching, requires no participation from the instructor (other than having recorded the video, of course): the students are told every week that by a certain day and time they must have watched the videos of the corresponding lectures. There is, in fact, a time slot reserved for that purpose, like a regular lecture slot, but the students can take care of the watching at any convenient time before the second component.

That second component takes place in the slot that would normally have been the lecture time. I come in, wait for questions, and answer them. That is all. I do not show prepared material, as in a traditional lecture. Of course, I have lots of ready material accumulated over the years or (as in the good old days) prepared hastily at dawn in anticipation of new questions, but the discussion is 100% driven by student questions.

Experiences with the Trio

I have honed this practice over the years. Almost as soon as I started teaching large courses, I included a few sessions that we called "Socratic lectures." I would announce that I would not cover a certain topic in class but expected the students to read such-and-such a chapter of the textbook (I am fortunate that for most of my classes I am able to use my own books, but this would work with another author's textbook as well) or such-and-such slides, and I would only take questions.

That was long before we started hearing the buzzword  "flipped classroom," but with a different name it was the same idea.

It made for some fun, at least second-degree, because the first few times the students were completely puzzled, upset even, since the setup did not correspond to any of their expectations and practices. I would say "Good morning, I am going to take your questions now," and no one would say anything. It takes some stoicism to continue peering at a group of 300 people who are not saying anything, and them peering at you. Five minutes, 10 minutes, 15 minutes — nothing.

And then, a miracle would happen: someone in the audience would raise a hand with a timid question and from then on the floodgates would open and the discussion would even run out of time. Once the inhibitor has been removed, there is no limit to how many questions come out.

Back then the "Socratic lectures" would happen once in a while, but now they are the norm. The initial reluctance still exists, but it dissipates very quickly. (As noted above, the first lecture or two are traditional-style, since students have not yet watched material that can trigger questions.) Then the questions quickly start coming.

It helps that I no longer expect questions to be all verbal; students can ask them in advance of the session on a chat channel (Slack), removing some of the obstacles for the more timid participants and making it possible to phrase questions carefully.

Needless to say, I repeatedly make it clear that there is no such thing as stupid questions, genuinely try to answer every single one (right away if possible, although sometimes I acknowledge the question and postpone the answer to the next session), and never consciously put down a student. (That does not mean kowtowing to students. As anyone who teaches at university level knows, part of the job consists of dispelling wrong preconceptions — the old Harvard/Annenberg "Private Universe" video has retained all its value —  so it is OK to tell a student that he or she is mistaken, but always in a respectful manner.)

Liberating the hidden questions

Opening up to students' questions how much misunderstanding can subsist after a lecture.

In a traditional setting, all these pent-up interrogations have no outlet; sure, almost every teacher asks, once in a while, "Are there any questions?", but generally there is no time to handle more than one of two at best, and not in depth. With the lecture time turned entirely into a question-and-answer session, hidden doubts get out, uncertainties are aired, misconceptions come to light and get corrected. The pedagogical effect is much deeper than in a traditional setting.

The teacher benefits, too. Part of the experience of teaching is learning. For a standard topic that you are teaching repeatedly, however, there is a point of diminishing returns. Not that you necessarily get bored (for my part, I love teaching and never experience boredom), but after a few iterations I do not learn much more by just teaching a topic.

You also stop improving; if you are teaching, say, algorithms, there are only so many ways to present Quicksort. It is unlikely that you are going to get significantly better after 3 or 4 times.

A discussion session, however, is never twice the same. Some questions will recur — although less and less as you take stock of them and include the answers in your recorded material — but others will surprise you, almost every time. Occasionally, a truly original question will lead you to exploring an aspect of the topic that you had not carefully considered before, and come back with a detailed answer, a kind of micro-lecture, on the spot of in the next session.

I can vouch only for myself, but can confirm that with the Trio system I continue to learn in every single session.

The primary beneficiaries should, of course, be the students. As noted at the outset, I do not at this point have systematic empirical evidence, but I am convinced that the pedagogical effectiveness of Trio-based courses is significantly improved. The students get to understand the material in much more depth, to ask questions that might otherwise remain unanswered, and to get rid of many misconceptions. They also interact more with their peers and see who else is having the same doubts and interrogations.

Some prerequisites

Clearly, the Trio system with its emphasis on  unpredictability of interactions can only work for a teacher who is comfortable with the subject matter and not afraid to face students. In principle we should all be in that situation, but undeniably some people get put in teaching positions without having enough of the necessary self-confidence. In such cases, the apparatus and pomp of traditional teaching act as a protection: you hide behind your slides and make sure to talk all the way up to bell time to avoid any embarrassing question. I am not casting the first stone: not everyone was born to teach, even the most self-assured instructors may be asked to take on a class outside of their comfort zone. Here we are talking about courses taught in good conditions, on a topic you master.

Note, in any case, that in the Trio world there is really no reason for being scared. For every lecture n (except the last in the semester, which is usually special anyway), there is a lecture n + 1. You can postpone an answer to next time; and if you mess up, you can provide a correction next time. Students will be less upset by the mistake than pleased by the honesty. If you do not actually mess up (after all, on a topic you master, that should not happen often), you can always come back to an improvised answer and make it better. So the experience does not need to be stressful.

The main obstacle is elsewhere: the sheer amount of work needed to produce a MOOC. It is not just a matter of standing in front of a camera and recording your usual spiel; you need to rethink the material thoroughly, divide it into short segments (typically 6-8 minutes, which does not come easily to verbose speakers like me), follow various rules that the MOOC experts will be glad to teach you, have impeccable slides (since any typo or other defect will remain for the whole world to see), and aim for impeccable delivery. The task is particularly hard if, like me, you are a nervous person and tend to stumble just because you know you should not because you are recording. I have gone through hundreds of false starts and rejected takes (although less now ,as I am finally becoming more relaxed).

The effort of devising and recording a MOOC is far more than what it usually takes to prepare slides and possibly a few lecture notes. In addition, you need some technical assistance. To refer to something else of which I also have practice, the effort for a MOOC is not unlike that of writing a textbook.

Indeed, the approach has worked well for me, since I am usually able to offer the combination of my own MOOC, my own textbook, slides, and other course material (such as exercises) developed over some time. Others may have to resort to external sources for some of these elements, such as someone else's textbook (or even someone else's MOOC, but I would recommend caution here). The Trio system will still work and deliver its benefits.

In general, you will have to develop and record your own video lectures. Not all courses justify that effort, but for those that do, my experience has shown the result to be rewarding.

Precursors, from Socrates on

The benefit of teaching through direct student interaction is, of course, not a new idea. The term "Socratic" takes us back to a great pedagogue of the 5th century BC who taught, among others, Plato and Xenophon. Closer to us, it is not surprising that elite institutions in several countries have long practiced interactive discussion as a key teaching technique, from Oxford tutorials to the "colles" of French Classes Préparatoires and "Petites Classes" at Polytechnique (where, in my time, the main math lecture was taught by a Fields medal winner and these group lab sessions by mere full professors from top universities).

With the Trio system, any institution can use these techniques, and interactive Q&A becomes the main form of teacher-student encounter, not a supplementary tool.

For applicable courses I would not, in spite of the effort that the Trio system implies, go back to the traditional way. Not everyone will agree, and new techniques will come in to recast the question once more — a discussion that I will be excited to have at the Villebrumier workshop.

Acknowledgment: my MOOC experience and the development of the techniques presented here owe a lot. over many years, to the ideas, prompting, and assistance of Marco Piccioni.

Bertrand Meyer is a professor at the Constructor Institute (Schaffhausen, Switzerland) and chief technology officer of Eiffel Software (Goleta, CA).