Time
“For all the points of the compass, there’s only one direction and time is its only measure.” (Stoppard, 1966)
Unlike most computational systems, humans experience and must respond to their environment through time. One of the most basic forms of human and animal learning is to associate things that happen close together. Everyday activities like speech, music, dance demand precise time control, while imagination, mental simulation, and memory (“mental time travel”) seem to require encoding of (and ability to regenerate events) in the right temporal sequence. What then is the role of time in the various cognitive processes and representations we have been discussing throughout this course?
Primary Readings
Everyone should read these and be prepared to discuss:
| Elman, J. L. (1990) | Finding structure in time. Cognitive Science, 14, 179-211. |
| Gong, T. & Bramley, N. R. (to appear) | Evidence from the future. Journal of Experimental Psychology: General, https://psyarxiv.com/f67nx/ |
Secondary Readings
The presenter should read and incorporate at least two of these:
| Grice, G. R. (1948). | The relation of secondary reinforcement to delayed reward in visual discrimination learning. Journal of Experimental Psychology, 38(1), 1.This article, published in 1948, explores the relationship between reinforcement and delayed reward in visual discrimination learning. The study involved training albino rats to discriminate between black and white stimuli under different conditions of delay of reward. The results showed that the number of trials required to reach the learning criterion varied based on the duration of the delay, and the type of secondary reinforcement had an effect on the learning process. |
| Bramley, N. R., Gerstenberg, T., Mayrhofer, R., & Lagnado, D. A. (2018). | Time in causal structure learning. Journal of Experimental Psychology: Learning, Memory, and Cognition, 44(12), 1880.This study examines how timing information influences judgments of causal strength and structure induction. Four experiments were conducted, revealing the importance of event order and temporal intervals in learning causal relationships. A Bayesian model accurately predicted participants’ judgments, highlighting the significance of time in understanding causal learning. Participants’ judgments were largely explained by event order, but there was some evidence of delay sensitivity and participants were able to distinguish between causal structures based on event timings alone. |
| Gong, T., Gerstenberg, T., Mayrhofer, R., & Bramley, N. R. (2023). | Active causal structure learning in continuous time. Cognitive Psychology, 140, 101542.The article discusses active causal structure learning in continuous time, exploring how people learn about causal structure and make intervention choices. The study finds that participants’ accuracy depends on the informativeness and evidential complexity of the data they generate. Participants optimize their intervention choices to maximize expected information and minimize inferential complexity. The paper also discusses the computational challenges of active causal learning and the role of metacognitive awareness in successful learning. The findings provide insights into understanding how people learn about causal structure in dynamic environments. |
Questions under discussion
- How veridical is our perception of the time course of events and actions?
- What problem do RNNs try to solve and how well do you think they achieve this?
- How might time help us encode non-temporal things like relations?
