Reflective Commonsense Thinking

Reflective commonsense thinking is a small field involving many cognitive sciences

The field of reflective commonsense reasoning is a research focus that necessarily exists as a small intersection of the cognitive sciences. For example, in the fields of artificial intelligence, we focus on layered cognitive architectures developed for distributed sets of perceptual, motor control, and knowledge-base resource commonsense reasoning tasks. There are not many cognitive architectures that currently exist and fewer of these are layered theories of mental control.

Goals of the field

The goals of the field of reflective commonsense thinking are:

	(1) to build models of reflective commonsense thinking,
	(2) to validate these models against human medical data,
	(3) to use these models to medically recognize human patterns of reflective commonsense thinking.

Comparable cognitive architectures

EM-1 is an example of a cognitive architecture that focused specifically on reflective commonsense thinking as the first sample implemention of Marvin Minsky's Emotion Machine (EM) theory. EM-1 established an example of critical reflection in the commonsense domain of building a table in a very specific social and physical context. SOAR is inspiring as a cognitive architecture because of it's ability to handle some sorts of impasses in simple planning problems, although SOAR fails in it's ability to reflectively inspect and control its own planning algorithms. ACT-r is another interesting architecture because it is optimized as a mapping from neurological and psychological (behavioral) human data to a human artificial intelligence algorithm.

Relevance to theoretical psychology

In the field of psychology, we focus on experimental models of reflective control, such as task switching, perceptual rivalry, and declarative to procedural knowledge transfer.

Relevance to theoretical neuroscience

In the field of neuroscience, we are inspired by current efforts to understand neurological data in terms of complex computational architectures attempting to explain complex human thought processes, such as commonsense reasoning; ACT-r is an example of a step in this direction.

Relevance to computer science

In the field of computer science, we focus on the development of formal languages to describe reflective processes, such as human thought but also including many other potential types of computational reflection.

Relevance to machine learning

In the field of statistical machine learning, we focus on the development of inference mapping tools that could be used to understand specific human beings commonsense reasoning processes in terms of active functional components of a reflective cognitive architecture.

Medical and scientific applications

The field of reflective commonsense reasoning has great application to many of the cognitive sciences, which brings better technologies to current mental health treatment methods.

Applications to the field of neuroscience

In the field of neuroscience, understanding neurological data in terms of a functional process description provides invaluable semantics for the ultimate goal of understanding a mechanical description bridging human behavior and human neuroscience.

Applications to the medical field of cognitive therapy

In the field of cognitive therapy, inference tools are developed to infer a patient's mental strengths in terms of a mental architecture, and subsequently, for example, computer games can be used to strengthen critical components of the overall architecture.

Current research projects

The field of reflective commonsense thinking requires basic engineering research in order to support the research in the cognitive sciences.

Building a reflective commonsense thinking simulation

The types of reflective processes that we are attempting to describe for simulation are very specific humanly idiosyncratic and complex processes of thought. We have found a number of key aspects to engineering the physical aspects of the simulation for reflective commonsense thinking.Certain aspects of the human thinking processes appear to be evident:

Causal Reflective Programming

Causal tracing in a reflective parallel processing language will allow all actions of a program to be monitored and critically reflected upon by other parts of the program. This new form of causal reflective programming allows for describing computational feedback processes; thinking of feedback as the tool used extensively in analog circuit design and linear control systems theory will similarly help this form of non-linear control theory.

Deliberation Layer

Inspired by Marvin Minsky's description of the deliberation layer in his Emotion Machine theory, we are building a robust adaptive goal-oriented GPS-like planning architecture taking advantage of traced process executions in order to learn and imagine goal-oriented execution plans.

Many heterogeneous processes execute in parallel, reflecting on one another as they run

The brain as a physical computational structure is not a uniform "blank slate" for computation. There is intricate structure to the human brain and which groups of neurons synapse on which other groups of neurons seems to develop very consistently between individual humans. Each of these groups of neurons performs a different computational task in coordination with only some other groups of neurons at any given moment. We are experimenting with language extensions for managing pools of heterogeneous processes.

Reflective Commonsense Thinking

neuromin.de

Field