Broadly speaking, I am interested in understanding how we reason, learn and take decisions and how those indvidual decisions, informed by learning, shape collective behaviour at the level of the population. My current focus lies at the interface of physics, cognitive science and computer science. A specific area of interest is related to understanding how cooperation can evolve and spread in scenarios of social conflict. I am interested in developing mathematical and computational models of cooperative decision making (that incorporate learning through different learning mechanisms) and supplement them with computer simulations to understand how distinct modes of decision-making in individuals affect cooperation levels in the population. My goal is to incorporate realistic modes of decision-making to gain insights into how we can tackle critical social problems that rely on cooperative decision-making.
That is what many people asked when Geoffrey Hinton was awarded the Nobel Prize in Physics in 2024. Jokes apart, even though the theoretical framework of these topics were developed independently by mathematicians, physicists and cognitive scientists, they were often inspired by statistical mechanics models, graphical models and physics of stochastic processess. Some of these topics also fall under the category of complex adaptive systems that many Physicists work on. They are complex because they have many distinct interacting agents who can interact in different ways and the system as a whole can exhibit emergent behaviour that cannot be predicted on the basis of individual interactions. They are adaptive because these agents can learn and evolve.
Because these are fascinating areas of research that grapple with some of the critical questions in complex systems and cognitive science: How do we learn, reason and take decisions? Would you not like to understand how the mind works? If you are open to exploring new areas and working on unconventional problems that builds on your knowledge of statistical physics and interest in mathematical/computational modeling, you should certainly consider this area. You will get valuable exposure to an exciting new and rapidly developing area of science. .
If you do good work and master the technical skills needed to work in this area, your chances of getting a PhD or a postdoc position in this field is high. Even if you choose to move to a different field, here is what you should keep in mind. The key questions that potential PhD and Postdoc mentors ask while evaluating your CV are the following. Have you developed a strong foundation in your major subject? Have you acquired some technical skills and are you capable of using those skills to solve open research problems? In other words, do you know how to learn and apply the acquired knowledge to solve problems. The exact topic of research is not as relevant as it may seem to you now. There are plenty of examples of scientists trained in Physics, who are well-known for their seminal work in other areas (Francis Crick, Venki Ramakrishnan are just two such examples of scientists with PhD's in Physics who went on to win the Nobel prize in other disciplines.) Working in these emerging fields can increase your competitiveness for phd, postdoc positions and even non-academic jobs, as expertise in interdisciplinary research is highly valued.
You will be learning about theoretical foundations of AI like MDP's, reinforcement learning, graphical models, evolutionary game theory etc. but not about LLM's or algorithm development for machine learning. Our primary goal will be to develop and analyse mathematical and computational models to better understand human decision-making and reasoning in different situations. One potential direction would be to understand how cooperation evolves in different scenarios of social conflict using the framework of evolutionary game theory. If you come up with exciting new ideas independently, you will definitely get a chance to pursue them.
You need to enjoy both math and coding as you will be expected to develop and analyse mathematical and computational models. If you are majoring in Physics and enjoy thinking about problems in statistical physics or complex systems you should consider exploring this area. If you are majoring in Math or Biology and enjoy thinking about questions in statistical inference/graphical models or cognitive science you should also consider exploring this area.
Yes, indeed! Here are links to 3 recent articles published in a PNAS special features edition. These are perspective articles that highlight some of the key questions and goals in this field of research. I strongly urge you to take a look at these articles. If you have time to read just one, read the first one.