Arbitrarily Applicable Same/Opposite Relational Responding with NARS (2505.07079v2)

Abstract: Same/opposite relational responding, a fundamental aspect of human symbolic cognition, allows the flexible generalization of stimulus relationships based on minimal experience. In this study, we demonstrate the emergence of \textit{arbitrarily applicable} same/opposite relational responding within the Non-Axiomatic Reasoning System (NARS), a computational cognitive architecture designed for adaptive reasoning under uncertainty. Specifically, we extend NARS with an implementation of \textit{acquired relations}, enabling the system to explicitly derive both symmetric (mutual entailment) and novel relational combinations (combinatorial entailment) from minimal explicit training in a contextually controlled matching-to-sample (MTS) procedure. Experimental results show that NARS rapidly internalizes explicitly trained relational rules and robustly demonstrates derived relational generalizations based on arbitrary contextual cues. Importantly, derived relational responding in critical test phases inherently combines both mutual and combinatorial entailments, such as deriving same-relations from multiple explicitly trained opposite-relations. Internal confidence metrics illustrate strong internalization of these relational principles, closely paralleling phenomena observed in human relational learning experiments. Our findings underscore the potential for integrating nuanced relational learning mechanisms inspired by learning psychology into artificial general intelligence frameworks, explicitly highlighting the arbitrary and context-sensitive relational capabilities modeled within NARS.

Arbitrarily Applicable Same/Opposite Relational Responding with NARS

In the paper presented by Johansson et al., the authors explore the computational modeling of same/opposite relational responding within the Non-Axiomatic Reasoning System (NARS). This domain of research is integral to advancing the capabilities of AGI, particularly in replicating nuanced human-like symbolic reasoning as described by Relational Frame Theory (RFT). The research is centered on the concept of Arbitrarily Applicable Relational Responding (AARR), which is fundamental in human cognition facilitating processes, such as language understanding, abstract reasoning, and symbolic manipulation.

The paper introduces an extension to NARS termed acquired relations, enabling the system to derive both symmetric (mutual entailment) and novel relational combinations (combinatorial entailment) from minimal explicit training. Through structured Matching-to-Sample (MTS) procedures, NARS effectively learned and generalized relational rules, demonstrating its capability to emulate phenomena seen in human relational learning experiments.

Core Findings and Methodology

The research delineates a meticulous experimental methodology structured into three phases:

1. Explicit Pretraining: Relational frames for mutual entailment and combinatorial entailment were explicitly trained to establish foundational capabilities in NARS. Mutual entailment involved symmetry-based training (e.g., if $X \rightarrow Y$, then $Y \rightarrow X$), while combinatorial entailment focused on transitivity-based training (e.g., $X \rightarrow Y$ and $Y \rightarrow Z$ infer $X \rightarrow Z$).
2. Relational Network Training: This phase utilized novel stimulus sets, employing Matching-to-Sample tasks within SAME and OPPOSITE contexts. The training facilitated the formation of internal acquired relations and relational hypotheses, preparing NARS for relational generalization in subsequent unreinforced testing phases.
3. Derived Relational Testing: The crucial testing phase assessed NARS' ability to spontaneously generalize relational understanding to novel stimulus pairs without explicit training or reinforcement. NARS successfully achieved perfect accuracy, offering robust evidence of its internalization of relational mechanisms.

Results

The authors report notable results whereby NARS consistently achieved perfect accuracy (100%) across all phases. Internal confidence metrics regarding mutual and combinatorial entailments showed substantial growth, reflecting NARS' internalization of relational principles and successful generalization of relational responding. This computational model effectively demonstrates derived relational responding, where SAME relations were spontaneously inferred from previously trained OPPOSITE groups, showcasing the system’s flexibility in relational reasoning.

Implications and Future Directions

The findings from this paper signify a substantive contribution to computational cognitive architectures, showcasing the integration of psychological theories such as RFT into AGI systems. The results are indicative of NARS' potential to perform human-like symbolic reasoning across various contexts, which could enhance AGI functionalities in language comprehension, adaptive decision-making, and dynamic environment interaction.

However, the current paper confines itself to relatively simple and abstract relational structures. Future research should focus on extending this framework to more complex relational networks, possibly intertwining real-world sensorimotor data to improve ecological validity and practical applicability. Addressing challenges related to scalability and generalizability remains pivotal for AGI applications, ensuring robust performance amidst real-world uncertainties and complexities.

In conclusion, Johansson et al. have succeeded in computationally modeling same/opposite relational responding, advancing the interdisciplinary discourse at the intersection of cognitive psychology and artificial intelligence. This work fosters a pathway toward developing more sophisticated, context-aware AGI systems, emulating intricate human-like relational reasoning capabilities.