Summary: Experts trained computer models to predict daily events, revealing that responding to confusion, rather than just forecast errors, improved knowing. This challenges the idea that occasion understanding is based solely on surprises and suggests that the brain may employ two methods. However, studies on remembrance show identifying occasion boundaries predicts better recall, particularly for older adults.
The purpose of continuing research is to enhance memory by enabling better understanding of these limitations. Studies may improve our understanding of mental processes and lead to strategies to treat memory loss caused by age. This job emphasizes the complex connections between memory storage and occasion classification.
Important Facts:
- Doubt improves understanding of common events, according to system designs.
- Memory engagement is strongly correlated with occasion boundary identification.
- Older people often struggle with celebration control, linked to memory decline.
Origin: WUSTL
Lifestyle is a series of smaller activities: making morning coffee, letting the dog out, opening a computer, letting the puppy back in. Put them all together, and you have a whole evening.
Our brains are dedicated to analyzing and interpreting the activities that characterize our daily lives, according to Jeff Zacks, the Edgar James Swift Professor of Arts and Sciences and head of the Department of Psychological & Brain Sciences.
Zacks argued that understanding the world requires knowing where occasions begin and end.
This crucial aspect of human consciousness is explored by Zacks and other experts in Arts & Sciences and the McKelvey School of Engineering in two novel papers.  ,
Before making predictions about what will happen next, Zacks led a review that trained computer models to watch more than 25 hours of video of people carrying out plain, day-to-day tasks like cooking and cleaning.
The investigation came to a surprising conclusion: The computer models were most exact when they responded to doubt. The model may reset and reevaluate the scene when it was most uncertain of what would happen next, which improved its understanding nevertheless.
Co-authors of the investigation, which will be published in an impending version of , PNAS Nexus,  , include Tan Nguyen, a grad student in Zacks’s Dynamic Cognition Laboratory, Matt Bezdek, a senior scholar in the laboratory, Aaron Bobick, the James M. McKelvey Professor and professor of the McKelvey School of Engineering, Todd Braver, the William R. Stuckenberg Professor in Human Values and Moral Development, and Samuel Gershman, a Harvard scientist.
The human mind was originally suggested by Zacks to be particularly sensitive to the minor occurrences that characterize our lives. He suggested that every time someone made a prediction error, or something they did n’t anticipate, they would reevaluate the scene.
The previous idea was questioned by the finding that the powerful computer model valued uncertainty more than prediction mistakes. ” We’re doing research around”, Zacks said. When confronted with fresh data, we “revisit principles.”
Surprises also issue, and there’s no need to completely throw out the concept of projection problem, Nguyen said. ” We’re starting to think that the brain uses both mechanisms”, he said. ” It’s not a case of either/or. Each model has the potential to contribute to improving our understanding of human cognition.
Maverick Smith, a postdoctoral researcher at Dynamic Cognition Lab, is also examining more deeply how event comprehension and memory interact. Working with former WashU postdoc Heather Bailey, who is now an associate professor at Kansas State University, Smith co-authored a review article for Nature Reviews Psychology, gathering increasingly compelling evidence that long-term memory is closely linked to the ability to logically and accurately determine the location of one event at the end and the beginning.
According to Smith,” There are many individual differences in the ability to determine when events start and end,” and those differences can significantly influence how much people will remember later.
We hope to be able to develop a tool that can aid in the segmentation of events.
Smith relies on video to learn how events are processed, just like Zacks. Instead of a person cooking and cleaning, his videos show a person shopping in a store, setting up a printer, or doing other mundane tasks.
When a particular event begins or ends, viewers in various experiments press buttons. Smith then tests the participant’s memory of the videos with a series of written questions.
According to Smith, older people are more prone to having trouble processing events, which may contribute to age-related memory loss. He said,” Maybe there is a way we can intervene to help them better recall the events in their lives.”
Zacks, Nguyen, Smith, and other members of the Department of Psychological &, Brain Sciences have ambitious plans to further their understanding of the brain’s ability to process and remember events.
The Zacks team is using fMRI brain imaging to monitor the response of 45 study participants to real-time videos of everyday events. ” We’re studying the actual neural dynamics of these cognitive processes”, Zacks said.
Eye movements are tracked in a different ongoing study, giving us fresh information about how we perceive the world. ” When people watch an everyday activity, they spend a lot of time looking at and thinking about people’s hands”, Zacks explained.
By facilitating the identification of the gaps between events, Smith is currently conducting video-based experiments to see if he can improve the memory of study subjects, including older people and those with Alzheimer’s disease. Ultimately, he would like to understand how event observations are stored and maintained in long-term memory.
According to Smith,” some people are undoubtedly better than others at segmenting events into relevant chunks.” ” Can that ability be improved, and can that lead to improvements in memory? Those are the questions we’re still asking”.
About this news about neuroscience research and memory
Author: Leah Shaffer
Source: WUSTL
Contact: Leah Shaffer – WUSTL
Image: The image is credited to Neuroscience News
Original Research: Open access.
” Modeling human activity comprehension at human scale: Prediction, segmentation, and categorization” by Jeff Zacks et al. PNAS Nexus
Abstract
Modeling human activity comprehension at human scale: Prediction, segmentation, and categorization
Humans form sequences of , event models—representations of the current situation—to predict how activity will unfold. The cognitive system switches between active event models and different mechanisms to control the behavior stream’s segmentation.
Recurrent neural networks for short-term dynamics and Bayesian inference over event classes for event-to-event transitions have been combined to create a computational model that learns knowledge about event classes ( event schemas ).
Event schemas are used to build a number of event models in this architecture. Through 18 hours of naturalistic human activity, this architecture was practiced in one pass. To verify whether each variant’s human segmentation and categorization were compatible with another 3,5 h of activities, we conducted another 3. 5 h of testing.
The architecture developed segmentation and categorization that were similar to human-like performance after learning to predict human activity.
The active event model’s prediction produced a lot of uncertainty, and the other transitioned when the active event model produced a lot of prediction error.
Despite not receiving any feedback on segmentation or categorization, the prediction uncertainty variant provided a slightly closer match to human segmentation and categorization.
These findings support the ability of event model transitioning to account for two crucial aspects of human event comprehension: prediction uncertainty or prediction error.
