I am a research assistant and Ph.D. candidate at the MIT Media Lab working on computational models of reflective commonsense reasoning. Once models of reflective commonsense reasoning are built, applications include understanding mental illnesses of reflection in terms of correlations with neurological regions and information patterns of the brain. We hope that teasing apart meaningful quantitative dimensions of "spectrum" mental disorders, such as Schizophrenia and Autism will involve more advanced commonsense models of learning and reasoning. This work combines the expertise of many fields, including artificial intelligence, cognitive science, and neuroscience. Our current work is in understanding the brain as a computer program. We are interested in answering questions such as: | | "What types of computer programs are good explanations of the human brain?", | | | "How are computational systems that reflect on episodic memories implemented in the brain?", and | | | "What new types of computer languages can help us to explore neural analogues to computation?". |
I have previously worked under the late Push Singh, who had developed the first working demonstration of Marvin Minsky's Emotion Machine cognitive architecture. This implementation is interesting for many reasons. First, it was based on memories in the form of commonsense narratives, a humanly natural but complex form of memory. Also, the model is able to reflectively debug its own problem solving by using narratives of the mental processes themselves. Previous experiences are transfered from one mental problem solver to another through a structure mapping process called Parallel Analogy, or Panalogy. Dustin Smith is further developing this process of panalogy between multiple parallel mental processes in a calendar application that takes into account the user's goals. |
