For many years I’ve studied moral judgment, with occasional projects on related topics such as religion,  cooperation, and free will. While I expect to continue working on morality, my lab is starting to work on other topics.

My research has both a social science side and cognitive science side. As a social scientist, I'm increasingly interested in projects aimed at producing demonstrable real-world social benefits. Interests include applied work on cooperation and conflict resolution (building on ideas presented in Moral Tribes) and research related to effective altruism. As a cognitive scientist, I’m increasingly interested in studying what one might call “thinking”.

The Infrastructure of Thought

A psychologist with a new interest in “thinking” may sound a bit strange. Isn’t all of psychology about thinking? Yes, in a sense. But at the same time many of the core mechanisms of thought remain mysterious.

We know a lot about the inputs to thinking (perception); the outputs of thinking (motor processing); the maintenance, storage, and retrieval of information about which one can think (memory); and about the use of words to transmit the products of thinking from one brain to another (language). Likewise, we've learned a lot about how thinking is influenced by attention and cognitive control and how behavior is reinforced by reward and punishment. We've learned a lot about how we form concepts of things in the world, as well as more abstract things such as numbers. We have some understanding of how reasoning works, mostly based on how it fails. But these failures typically amount to a reliance on something other than full-blown reasoning--thinking "fast" instead of "slow". We don't understand the kind of thinking that enables us to recognize our failures of reasoning as failures. We have, in other words, a rather limited understanding of what David Hume and other Enlightenment philosophers called “the Understanding,” or what I am here calling “thinking.”

What, then, is “thinking” in this more narrow sense? One possibility is that there is no “there” there, that thinking is just the sum of perception, memory, language, concept-formation, attention, control, etc. But I suspect that there is a distinctive type of processing that sits at the intersection of all of these things, and that is what I mean by “thinking.” (Here I am echoing Jerry Fodor’s famous claim that there must be a Language of Thought, although I share only some of his views about what the hypothesized LOT must be like.)

If I say, “Yesterday I was serenaded by 30 pink elephants,” you immediately understand what I mean, despite the novelty and improbability of my claim. Your ability to immediately grasp my precise meaning implies that you have the capacity to rapidly and flexibly combine concepts according to precise rules, yielding thoughts that may be new to you and perhaps new to the universe. What's more, you can immediately use new thoughts as inputs to further thinking. (Were an even number of animals performing?) We take these familiar abilities for granted, but how the brain accomplishes these feats is a mystery.

This use of structured conceptual combination is known as “compositionality.” Compositionality is most often discussed as a feature of language, but there are reasons to think that it goes beyond language and that it may precede language in our natural history. First, if I ask you to imagine my being serenaded by 30 pink elephants, you can do it, which means that you can, within your mind, compose a visual scene with features that mirror the complex compositional features of the corresponding sentence. This suggests that compositionality extends to non-linguistic processing. Second, animals such as chimpanzees seem to do some pretty sophisticated thinking in the absence of anything like human language, suggesting that compositional thought may precede compositional language. Whether language is compositional because thought is, or vice versa, or both, human thinking certainly seems to be compositional, and we have only faint clues about how our brains compose and manipulate the complex contents of thoughts.

With all of this in mind, Steven Frankland and I have begun working on this problem, trying to understand how the brain puts concepts together to form thoughts. The results of our first fMRI experiments indicate that the brain, in its capacity for semantic composition, functions more like a silicon computer than many have supposed, representing and updating the values of abstract semantic variables. These results provide some clues about how neural tissue can represent complex thoughts.

New projects include research on deductive reasoning and logical connectives (led by Frankland), propositional attitudes (led by Regan Bernhard), composition in visual imagery (led by Dillon Plunkett), and related research using intracranial recordings (with collaborators at MGH). I have a related interest in artificial general intellgience and, more specifically, in understanding how machines might acquire concepts and use them in flexible ways, as humans do. Bridging my interests in morality and artificial intelligence, I'm starting to think about ethical principles and safety constraints for autonomous machines.

I welcome inquires from prospective students and collaborators with interests and technical skills in the areas described above, including both applied moral cognition and the neuroscience of high-level cognition.