Machine Psychology and Artificial General Intelligence

Machine Psychology is an interdisciplinary framework integrating learning psychology and artificial intelligence to advance AGI research¹. By synthesizing operant conditioning with Pei Wang’s Non-Axiomatic Reasoning System (NARS), Machine Psychology proposes a model where AI systems adapt dynamically through structured learning².

Foundations of Machine Psychology

Machine Psychology builds on two core scientific traditions:

Learning Psychology: Behavioral adaptation via operant and relational learning paradigms³.
Artificial General Intelligence (AGI): Computational intelligence designed to generalize knowledge across domains⁴.

The research examines how AI can develop generalized identity matching, functional equivalence, and arbitrarily applicable relational responding (AARR)⁵. These cognitive mechanisms are foundational in behavioral psychology and are tested within NARS to develop flexible, adaptive intelligence.

The Role of Operant Conditioning in AGI

Operant conditioning is a powerful learning mechanism in psychology, where behaviors are shaped by consequences⁶. Within NARS, this principle enables the system to refine responses dynamically, similar to how humans learn through reinforcement⁷.

Experiments in Machine Psychology show that AI can develop structured responses to environmental contingencies, demonstrating key cognitive properties of intelligence⁸.

Empirical Studies in Machine Psychology

Recent empirical research in Machine Psychology has explored:

Stimulus Equivalence: AI learning relationships between different stimuli without direct training⁹.
Relational Framing: AI establishing new cognitive frames based on past experiences¹⁰.
Generalized Identity Matching: AI recognizing and applying abstract relationships across diverse inputs¹¹.

These studies suggest that AI can exhibit forms of abstract thought and reasoning, moving closer to AGI capabilities.

Conclusion

Machine Psychology represents a new paradigm in AGI research, leveraging behavioral principles to create adaptable, learning-driven AI systems. By integrating operant conditioning with the flexible reasoning mechanisms of NARS, this approach bridges the gap between human cognition and artificial intelligence.

References

Johansson, R. (2024). Empirical Studies in Machine Psychology. Linköping University Electronic Press. ↩
Wang, P. (2022, April). Intelligence: From definition to design. In International Workshop on Self-Supervised Learning (pp. 35-47). PMLR. ↩
De Houwer, J., & Hughes, S. (2020). The psychology of learning: An introduction from a functional-cognitive perspective. MIT Press. ↩
Goertzel, B. (2014). Artificial general intelligence: concept, state of the art, and future prospects. Journal of Artificial General Intelligence, 5(1), 1. ↩
Johansson, R. (2024). Machine Psychology: integrating operant conditioning with the non-axiomatic reasoning system for advancing artificial general intelligence research. Frontiers in Robotics and AI, 11, 1440631. ↩
Skinner, B. F. (1938). The behavior of organisms: An experimental analysis. BF Skinner Foundation. ↩
Hayes, S. C. (2001). Relational frame theory: A post-Skinnerian account of human language and cognition. Plenum Press. ↩
Hammer, P. (2022, April). Reasoning-learning systems based on non-axiomatic reasoning system theory. In International Workshop on Self-Supervised Learning (pp. 89-107). PMLR. ↩
Sidman, M. (1994). Equivalence relations and behavior: A research story. Authors Cooperative. ↩
Barnes‐Holmes, D., & Harte, C. (2022). Relational frame theory 20 years on: The Odysseus voyage and beyond. Journal of the experimental analysis of behavior, 117(2), 240-266. ↩
Johansson, R., Lofthouse, T., & Hammer, P. (2022, August). Generalized identity matching in NARS. In International Conference on Artificial General Intelligence (pp. 243-249). Cham: Springer International Publishing. ↩