In the research tradition called “contextual behavioral science” (Zettle, Hayes, & Barnes-Holmes, 2016) it is argued that a large part of cognitive phenomena are made possible due to a type of operant behavior known as “arbitrarily applicable relational responding”. Relational Frame Theory (RFT; Hayes, Barnes-Holmes, & Roche, 2001; Roche & Dymond, 2013) is a contextual behavioral account of language and cognition. RFT aims to develop a unified account of language and cognition and have been showed to account for as diverse topics as language development, the emergence of a self, human suffering, intelligence, problem solving, etc. The fundamental thesis of RFT is that language and cognition are all instances of arbitrarily applicable relational responding (AARR). According to this perspective, relating means responding to one event in terms of another. While both non-humans and humans are able to respond relationally, only humans seem to able to do this arbitrarily. For example, a human being can be presented with three similar coins and being told that “coin A is worth less than coin B, which in turn is worth less than coin C”. The fact that a human being in some context would immediately pick coin A, is to RFT an example of AARR in which stimuli are arbitrarily related along a comparative dimension of worth.
NARS (Non-Axiomatic Reasoning System; Wang, 2006, 2013) is a project aiming to building a general purpose intelligent system. An assumption in NARS is that the essence of intelligence is the principle of adapting to the environment while working with insufficient knowledge and resources. Accordingly, an intelligent system should rely on finite processing capacity, work in real time, be open to unexpected tasks, and learn from experience. NARS is built as a reasoning system, using a formal specification “non-axiomatic logic” (NAL) to define its functionality. NAL is designed incrementally with multiple layers. At each layer, NAL and its internal language Narsese are extended to have a higher expressive power, a richer semantics, and a larger set of inference rules, so as to increase the intelligence of the system. The reasoning process in NARS uniformly carries out many cognitive functions that are traditionally studied as separate processes with different mechanisms, such as learning, perceiving, planning, predicting, remembering, problem solving, decision making, etc.
The primary aim of this work is to investigate if NARS can do AARR with gradually increasing complexity, and under which conditions this is made possible. Potential applications are for example describing and exploring mental health phenomena within an artificial general intelligence framework.