As a part of my Physics education, I have had the pleasure to attend a large number of courses. Typical trajectory of a physics course can be summarized as follows: we start off by picking up necessary math skills (for example, vector algebra and calculus for Newton’s laws of motion), then we use these skills to unpack the concise statements of physical laws (for example, what does F=ma mean?) and at the end, we apply these laws to specific problems to improve our understanding of the laws (for example, if an object of 5 kg is subjected to 4 N of force, what’s the acceleration?).

I attended a Machine Learning course recently which was different from traditional courses in more than one ways. It was based on flipped classroom concept; we did simulations in class, we had discussions in class and we read research papers as part of assignments. One can argue that the unusual strategy suited the course because it was on Machine Learning, which is still in infancy unlike Electrodynamics. However, I believe that the same strategy can be applied to graduate courses teaching well-established fields. Although flipped-classroom approach has been applied to undergraduate courses in many universities, it’s hard to find a Physics graduate courses, which has been taught in non-traditional approach.

The two most important skills, which I learnt in that course, was to ask big questions and read research papers critically. I am going to explain below what exactly I mean by the above terms and why I think these skills are crucial for learning how to do research. These two skills have never been taught to me in any other course even though the goal of any graduate course should be to prepare graduate students for their life as a researcher.

Why learning to ask big questions is important for a budding researcher? Asking the big questions helps one to understand what’s the whole point of a certain subject (for example, what is Machine Learning and why it’s hard?). This is the first step that one can take to get a sense of where the community currently stands. Once we know what is the current body of knowledge, we need to ask second question: how can we expand the boundary of current knowledge?

To make it more explicit, let me give examples of big questions that can be asked and explored in graduate courses:

- In a
**Quantum Mechanics**course, we can start by asking what’s a quantum system – does quantum systems need to be necessarily small and discrete? When does a system stop behaving like a quantum system and starts becoming like a classical system? How do you design experiments that measure the quantum behavior of the world? Stern-Gerlach experiment and Feynman’s double slit thought experiment are good examples to illustrate the subtlety of quantum experiments. - In a
**Statistical Physics**course, we can ask what’s the recipe of studying complex physical systems according to Statistical Physics? What is ergodicity and when does it fail? Why Anderson argues More is Different? In other words, why reductionist principle can fail to explain an emergent phenomenon like Bose-Einstein condensation? How Quantum Statistical Physics is different from Classical Statistical Physics? - In a
**Quantum****Many-Body****Physics**course, we can start by asking why many-body systems are hard to simulate on classical computers? If one starts from non-interacting limit, can we use perturbation to understand the effect of interaction in a many-body system? Would we ever discover a general principle that goes beyond mean-field theory? What’s the effect of infinite dimensional phase-space on physics of many-body systems? Anderson Orthogonality Catastrophe is a perfect example illustrating this point. - In a
**Quantum Computing**course, we can ask what is a quantum computer? Is a quantum computer necessarily better than a classical computer? Why Feynman argued that a quantum computer can help us simulate a quantum world? - In a
**Machine Learning**course, we can ask what is machine learning? What is the recipe of studying complex systems according to machine learning? Why machine learning is hard? - In a
**Biophysics**course, we can ask what defines life? What measurements can be done to distinguish between a living and non-living beings? What is Physics of Life – would we ever discover some new fundamental force in living beings?

The whole point of these questions is that we are spending time in class thinking about what are the most important questions for the field, how does the subject traditionally answer them and is there a better way to go about solving them?

When we start asking these questions, it’s natural to ask how scientists before us tried to answer them. This brings me to the second skill: reading research papers critically. Before I go into details, let me list some resources which I found useful in improving my paper reading skills. I recently got to know there is a systematic method called QALMRI to learn how to read papers, which people follow in cognitive science. Also, I have found Terry Hwa’s reading guide useful.

It’s important to note that merely assigning classic papers to read without any proper class discussion would not help. A class discussion would help students to see what they missed in their first reading. Further, reading classic papers can also help students write better papers in future.

I have seen that almost all Professors know classic papers in their field, but curiously they rarely assign these papers as part of course assignments. I don’t know why they don’t. However, they definitely assign a lot of problems. They seem to be convinced that physicists learn best when solving problems and calculating physical quantities.

Sometimes, calculating something doesn’t necessarily leads to better understanding. You can solve problems by following some algorithm without any deep understanding; you need to stop and ponder why something is working (or not working). Don’t get me wrong, it’s definitely helpful to solve textbook-style problems to gain some intuition. But I feel since we physicists are so good at calculating things (using spherical cow approximations), we forget that starting point of any research is asking the right kind of questions. Also, we sometimes ignore the fact that there are other ways of acquiring knowledge – asking big questions, whose answers can’t be necessarily calculated at the current moment and learning from someone else’s calculation/simulation/experiment by reading their papers. Along with problem solving skills, I am arguing here that above mentioned skills should be also taught in classrooms.

Graduate courses that emphasize asking big questions and reading papers critically are needed to train next generation of physicists, who don’t only know how to solve problems but also to recognize which problems are important. This philosophy-centric teaching approach is truer in spirit to the goals of *philosophiae doctor *(PhD) degree.