At any given moment, the mind needs to decide how to think about what, and for how long. The mind’s ability to manage itself is one of the hallmarks of human cognition, and these meta-level questions are crucially important to understanding how cognition is so fluid and flexible across so many situations. While it is important to understand how the mind works with respect to individual tasks, it is equally important to understand how the mind reasons on a meta-level across tasks. Under the assumption that thinking is analogous to computing, my aim is to develop a mathematical theory of how the mind manages its computational resources. Using machine learning methods including Bayesian models, reinforcement learning, and deep learning, I build normative models of how people should make meta-level decisions and evaluate these models using a variety of behavioral methods.
I investigate the problem of cognitive resource management by focusing in particular on the domain of mental simulation. Mental simulation is a phenomenon in which people can perceive and manipulate objects and scenes in their imagination in order to make decisions, predictions, and inferences about the world. For example, take a look at the image to the left, and imagine that the gray blocks are made out of heavy stone, and that the green blocks are made out of a light plastic. Which direction will this tower fall? How do you know? Many people report that they know the answer because they can visually imagine what it will look like when the tower falls. In other words, they solve the problem mentally simulating what will happen.
Importantly, if thinking is computation, then mental simulation is one particular type of computation analogous to having a rich, forward model of the world. Given access to such a rich model, how should the mind use it? How does the mind infer anything from the outcomes of its simulations? How many resources should be allocated to running which simulations? When should such a rich forward model be used in the first place, in contrast to other types of computation such as heuristics or rules? Understanding the answers to these questions provides broad insight into people’s meta-level reasoning because mental simulation is involved in almost every aspect of cognition, including perception, memory, planning, physical reasoning, language, social cognition, problem solving, scientific reasoning, and even creativity.
The ideas regarding meta-level reasoning that I have explored in cognitive science are highly relevant to other fields such as machine learning, robotics, and artificial intelligence, for two reasons in particular. First, I think that cognitive science can bring new ideas and perspective in thinking about the problem of intelligence: humans are, after all, the best example of intelligence that we have. Second, while AI need not mimic human intelligence, it does need to interact and communicate with people. It is important that AI is built with the quirks of human intelligence in mind, which requires that researchers in computer science work together with researchers in cognitive science, psychology, and other fields in the social sciences.