The Paradox of Time in Dynamic Causal Systems.

Area of Science:

• Cognitive Science
• Psychology
• Artificial Intelligence

Background:

• Humans use temporal cues (order, delay, variability) to infer causality.
• Previous research explored learning in dynamic systems with rapid changes, suggesting resource-limited strategies.
• The role of time in continuous dynamic systems for causal inference remains less understood.

Purpose of the Study:

• Investigate the impact of temporal dynamics on causal inference in continuous systems.
• Examine if slower system dynamics reduce errors compared to rapid dynamics.
• Develop and validate a model explaining human causal inference in dynamic environments.

Main Methods:

• Participants interacted with dynamic systems where causes and effects unfolded continuously.
• System dynamics were manipulated to be either rapid or slow.
• Human error patterns in causal inference were analyzed and compared across conditions.
• A novel computational model, Causal Event Abstraction, was developed.

Main Results:

• Slowing the task reduced specific types of causal inference errors.
• However, slowing the task also led to an overall increase in error rates.
• The Causal Event Abstraction model accurately predicted observed human error patterns.

Conclusions:

• Human causal learning in dynamic systems involves abstracting continuous processes into discrete events.
• Temporal dynamics significantly influence the nature and frequency of causal inference errors.
• The Causal Event Abstraction model provides a framework for understanding human causal cognition in dynamic environments.