Competitive feedback loops in binary decision making at neuronal level

From supermemo.guru
Jump to navigation Jump to search

This reference is used to support claims made by Dr Piotr Wozniak as part of an article series on memory, learning, creativity, and problem solving.

In binary decisions, we choose between two options of different value. Neuronal systems that allow of dynamic decisions with fluctuating value assessment may have a form of a simple feedback loop that may involve mutual competitive lateral inhibition between neurons involved in the decision. Feedback and stabilization makes it possible to express the history of valuations, and provide for faster decisions in well researched contexts.

When decisions keep bringing the same rewarding outcome, the synaptic configuration can be simplified by pruning. Unused options are eliminated and we end up with deterministic outcomes and fast thinking.

On a large scale, more complex networks, or multiple networks may be involved in decision making. In the context of schooling, the student is often forced to choose between the valuations of her learn drive and the valuations coercively imposed by the school system. The resulting war of the networks may determine a spectrum of school behaviors from rebellious (i.e. guided by the learn drive) to helpless (i.e. submissive towards the system of school rewards and penalties)(see: school penalties and the learn drive). Metaphorically speaking, a dog caged for long periods of time may not be aware that its cage has been opened.

When brain control systems are overridden (e.g. in dieting, coercion, interrupting sleep with alarms, etc.), war of the networks may result in pruning that makes the control system dysfunctional. For example, the outcome of coercive dieting may produce a control system where eating is always the right option leading to obesity.

Wolfram Schultz explains the role of the dopamine error signal in this process:

Within a simple comparator model of decision-making, the global influence of the dopamine error signal derived from chosen value may serve to update input decision variables for specific options via eligibility traces selectively stabilized by activity from the chosen option

Figure: Neuronal feedback loops in competitive decision making. Input patterns determine the input value (e.g. as computed by knowledge valuation network). Inhibitory neurons IA and IB decrease the chance of firing in the competitive decision neuron (Decision A or Decision B). History of prior decisions will determine synaptic stabilization, which will favor decisions that used to bring higher past rewards. See also: War of the networks, Competitive feedback loops in binary decision making at neuronal level, and Learn drive at school


Quoted excerpts come from the following reference:

Title: Neuronal Reward and Decision Signals: From Theories to Data

Author: Wolfram Schultz

Date: 2015

Backlink: War of the networks

Link: https://www.ncbi.nlm.nih.gov/pmc/articles/PMC4491543/

Figure 45 link: https://www.ncbi.nlm.nih.gov/pmc/articles/PMC4491543/figure/F45/