Create an Account CourseStreet Log in  Connect with Facebook
Home Blog
 

NRSC 2100 Blog

A GROUP WEBLOG FOR NRSC 2100 SUMMER NRSC 2100

Showing entries tagged surprise.  Show all entries

October 19, 2011

Well that's Surprising...ly Negative


Have you ever been surprised to be let down? Or in other words, have you ever expected a certain outcome only to be surprisingly disappointed? Well if you have, ladies and gentlemen, then do not fear; for your dorsal anterior cingulate cortex is functioning properly! And what's that? There's unified model for the long disputed function of the dorsal anterior cingulate cortex? That's right! Both of these birds were hit by the same stone recently when Alexander and Brown produced a computational model "tour de force" to illustrate how negative surprise signals drive dACC (dorsal anterior cingulate cortex) and mPFC (medial prefrontal cortex) responses.
Many theories have been concocted as to what the dorsal anterior cingulate cortex may be responsible for, such as error detection, error likelihood prediction, and conflict monitoring primarily, and even more such as reinforcement-guided decision making, negative reinforcement learning signals, and action value prediction error. Could the dACC be responsible for all of this in the brain? Well, Alexander and Brown's model seems to narrow our spectrum a bit and put an end to this controversy.
While their model agrees with previous theories that the dACC and mPFC predict action-outcome situations, it is uniformly different in the sense that these regions are responsible for multiple predictions for action-outcome situations in parallel, and then these predictions are scaled to their probability of their occurrence. When the predicted outcome doesn't happen, learning rates are modified in order to update action-outcome predictions to the degree necessary to learn from mistakes and find a better solution.
Another important point of this model's representation of multiple predictions of action-outcomes is that different ongoing predictions could account for heterogeneity of neural responses usually observed in single-unit studies. So basically, the dorsal anterior cingulate cortex and medial prefrontal cortex can encode different outcomes simultaneously for the same situation that are being encoded in different groups of neurons! Pretty impressive eh?
So let's just recap. The dorsal anterior cingulate cortex analyzes a particular action, predicts an outcome for this action, and if the action-outcome prediction is negated, then the dACC modifies learning rates so that the brain can learn from its mistakes. And the dACC and mPFC can do this multiple times at once!
So while Alexander and Brown's model is reasonable and presents much more concise data, it is obviously provoking new questions and controversy. Seeing as how consequences of positive and negative surprises are the same according to this new model, what makes a negative surprise more significant or important than a positive surprise? If the dorsal anterior cingulate cortex is responsible for negative surprise predictions and reactions, what is responsible for positive surprise monitoring? As for these questions, we shall see what new models of these mysterious brain regions are presented and what will be discovered for the tasks we perform in daily life. Regardless of what is discovered in the future, we'll all be surprised!

main article:http://www.nature.com/neuro/journal/v14/n10/full/nn.2932.html
Posted by      Mark A. at 4:19 PM MDT
  Christina Uhlir  says:
Mr. Alsberg,

Could you kindly explain the mechanism by which the "tour de force" operates?
Posted on Sun, 23 Oct 2011 2:20 PM MDT by Christina U.




 Copyright © 2007-2016 Don Cooper, Ph.D.. All rights reserved.
  Feed — Subscribe: RSS