The Accuracy of Causal Learning Over Long Timeframes: An Ecological Momentary Experiment Approach

Abstract

The ability to learn cause-effect relations from experience is critical for humans to behave adaptively -to choose causes that bring about desired effects. However, traditional experiments on experiencebased learning involve events that are artificially compressed in time so that all learning occurs over the course of minutes. These paradigms therefore exclusively rely upon working memory. In contrast, in real-world situations we need to be able to learn cause-effect relations over days and weeks, which necessitates long-term memory. 413 participants completed a smartphone study, which compared learning a cause-effect relation one trial per day for 24 days versus the traditional paradigm of 24 trials backto-back. Surprisingly, we found few differences between the short versus long timeframes. Subjects were able to accurately detect generative and preventive causal relations, and they exhibited illusory correlations in both the short and long timeframe tasks. These results provide initial evidence that experience-based learning over long timeframes exhibits similar strengths and weaknesses as in short timeframes. However, learning over long timeframes may become more impaired with more complex tasks.

Causal learning with delays up to 21 hours

“…The reason is that one way in which standard causal-learning paradigms are artificial is that all the trials are presented in quick succession, whereas in the real world (e.g., learning if a medicine is working, or what factors influence sleep), the experiences are spaced out over much longer periods of time. We have found that people can learn true relations between a single cause and a single effect about as well when spaced out one trial per day as when presented rapidly within a few minutes (Willett & Rottman, 2021). Furthermore, in both short and long timeframe conditions, participants incorrectly inferred correlations that did not exist ('illusory correlation') when observing skewed datasets.…”

“…Our approach for studying causal learning in more real-world timeframes is to incrementally move towards more and more realistic learning situations. In our first study on learning over a long timeframe, we simply stretched out the standard trial-by-trial paradigm so that one trial occurred each day (Willett & Rottman, 2021). In the current study, we took the next step of introducing delays between the cause and effect.…”

“…In the current study, we took the next step of introducing delays between the cause and effect. We have also studied learning over long timeframes with two causes but with no delays (Willett & Rottman, 2020 ). We are intentionally taking small steps to understand how individual changes affect learning.…”

“…Furthermore, in both short and long timeframe conditions, participants incorrectly inferred correlations that did not exist ('illusory correlation') when observing skewed datasets. Follow-up research focused on people's ability to learn about two causes and one effect in a long timeframe setting (Willett & Rottman, 2020). Another study investigated reward-based learning, not causal learning, in both a single massed session vs. several sessions spaced out over time.…”

The Paradox of Time in Dynamic Causal Systems

“…It is an open question what relationship such radical clock-time shift has on the interactions between human cognition, intervention choice, event abstraction and causal learning. Recent work examining causal inference from observations spanning hours [ 26 ] and days [ 27 ] suggests people have at least as much difficulty identifying relationships and dealing with confounds and dependencies. In such settings it seems likely that processing bottlenecks are caused as much by the structure and limits of long term memory and retrieval as by limited online processing bandwidth.…”