Computer Model of a "Sense of Humour". I. General Algorithm (0711.2058v1)

Abstract: A computer model of a "sense of humour" is proposed. The humorous effect is interpreted as a specific malfunction in the course of information processing due to the need for the rapid deletion of the false version transmitted into consciousness. The biological function of a sense of humour consists in speeding up the bringing of information into consciousness and in fuller use of the resources of the brain.

• The paper presents a computational framework where humor emerges from processing anomalies due to accelerated information transfer and temporal constraints.
• It details an algorithm that generates and selects interpretive trajectories, triggering humor when later inputs invalidate initial assumptions.
• The study highlights potential AI applications by modeling neural humor responses and explains individual differences through variations in processing time and memory capacity.

Summary of "Computer Model of a 'Sense of Humour'. I. General Algorithm"

The paper presented by I. M. Suslov offers a theoretical framework for the development of a computer model capable of mimicking the human sense of humor. This endeavor explores the notion that humor emerges from specific malfunctions in information processing within the human brain. Suslov conjectures that the biological functionality of humor is streamlined information processing into consciousness, alongside optimizing brain resources. Notably, the paper delineates a general algorithm for simulating humor, suggesting its future applicability in neural networks and possible mechanistic facilitation of laughter.

Suslov initially situates his hypothesis within evolutionary biology, suggesting that the capacity to experience humor evolved as a mechanism for transmitting processed information more effectively into consciousness. By conceptualizing the humorous effect as a cognitive anomaly resulting from the presence of two or more incongruent interpretations of a stimulus, the model proposes that the collision of these interpretations induces the emotional response we recognize as humor.

The humor model employs an information processing framework. This involves associating symbols with sets of images, generating possible interpretative trajectories, and selecting the most probable one as processed information. Humor arises when this selection process undergoes an expedited phase to meet a critical time constraint ($\tau_{max}$)—leading to premature release into consciousness. This malfunction occurs when subsequent inputs retroactively invalidate the assumed interpretation, necessitating deletion and correction, thus generating a humorous effect.

Several established phenomena are explained through this model, such as the ineffectiveness of repetitive jokes and the influence of delivery on humor. These are attributed to the brain's predictive accuracy mechanisms and limitations in processing time. Additionally, individual differences in humor perception are attributed to variations in the $\tau_{max}$ parameter and operative memory capacity.

The paper also touches on the broader implications of these theoretical appraisals for artificial intelligence, positing the potential to emulate basic humorous responses computationally. While simple humor involving wordplay might be relatively straightforward to model, capturing complex humor would necessitate a robust understanding of human cognitive frameworks and associative mechanisms.

In conclusion, Suslov’s paper introduces a computational perspective on humor that implicates temporal processing constraints as central to the humor experience. It highlights paths for future AI research, where modeling cognitive processes like humor might enhance the development of more advanced and nuanced machine learning algorithms capable of engaging in human-like emotional interactions.