|
|
|
Neural circuits weight the optionsThe Anatomy of Bias: How Neural Circuits Weigh the Options Jan Lauwereyns MIT Press Introduction: The most important section of this book deals with the choice problem. How does the brain cope when it has to deal with a conflict between two different stimuli, plans or strategies, for instance the conflict between a short-term reward or the deferred enjoyment of a larger but longer-term reward. These situations are evaded in the rather superficial discussion that passes for the mainstream consensus on freewill, which typically refer to timing decision for trivial pre-planned actions. The author discusses effort-related decisions from the point of view of the influence of neuromodulators, speculating as to whether there are signals to inhibit a discarded courses of action, or whether there is simply stronger dopamine support for the favoured course of action. He also refers to a study that showed activation in emotional processing areas of the brain when moral decisions were involved. This carries with it the possibility that concepts such as fairness could create a dopamine-based bias in favour of particular decisions. These decisions also require the operation of some kind of weighting mechanism once again probably involving neuromodulators. What is only hinted at here is that the impact of neuromodulators such as dopamine is often registered in subjective consciousness, so what is sometimes being balanced is two subjective impulses. Bayes: Bayesian probability is discussed in the early part of this book. This is derived from the ideas of Thomas Bayes, where unequal weight can be given to different sources of information. The likelihood of something being true is weighted according to its probability prior to a pre-existing proposition being made, or particular evidence being presented. This previous level of probability is referred to as 'the prior'. Bayesian probability involves the combination of a particular piece of evidence and the likelihood of a particular hypothesis related to this evidence. The outcome of our thinking on this is referred to as the 'posterior probability'. This means that our beliefs about the world can be updated by combining new evidence with our 'priors'. The priors colour or help to interpret new information. Priors may allow us to draw more accurate conclusions than the newest evidence by itself. Another way of looking at this is to think that decision taking benefits from observing how often things happen in the real world. The priors can be described as bias, but it is argued that such bias is rational if it corresponds with the statistical regularities of the environment. The author looks first at the research of R.H.S. Carpenter (1.-4.). Carpenter argued on the basis of a number of his studies that response times with eye movements were several times slower than what should be expected. He suggested a central decision making process injecting random variability. The element of randomness was suggested to have an adaptive advantage in making it more difficult for predators or prey to predict an organism's future actions. Some of the Carpenter research showed that response time was quicker when the type of stimuli being delivered was a likely one, than when it was an unlikely one. The activity level of neurons was enhanced for stimuli with a high level of prior probability. The effect of a particular stimulus is seen to be greater if it is pre-coded as being likely. This also gives rise to a proneness to false alarms or wishful thinking. When such a system deals with a stimulus leading to a reward or a reinforcer, it would also require an error detection signal, both to deal with stimuli that produced unexpected rewards, and with stimuli that did not produce expected rewards. The positive error of a stimuli that produced an unexpected reward could allow a new reward association to be formed, while the negative error of a reward not being produced could lead the subject to reduce or eliminate the response to a stimuli. Studies by Schultz and colleagues (5. & 6.) show that dopamine neurons can quickly adapt to respond to a visual stimulus that is related to a reward. Other studies show that the dopamine neurons are influenced by both the size and probability of rewards. Neurons in the caudate nucleus, which is part of the basal ganglia, appear to have access to reward information, and the basal ganglia region is associated with motivation for action. The basal ganglia get excitatory input from dopamine neurons and from the cortex, and send inhibitory output to the thalamus, which sends it on to the cortex. Hikosaka and colleagues observed anticipatory behaviour in neurons that became active before the visual stimulus had been presented. This anticipatory activity was context dependent in that it happened in some trials but not in others. This behaviour seemed to derive from a relationship between spatial position and the availability of a reward. Neurons appeared to respond to particular combinations of spatial position and reward. The caudate neurons in particular were suggested to have 'desire fields' or relations between inputs that they were more likely to respond to. The anticipatory activity was finally concluded to be a bias mechanism allowing a quicker response to a positive stimuli. It favoured the hypothesis that a particular visual target was associated with reward. Studies showed that response time was much quicker if there had been anticipatory activity. This is seen as pointing to a reward-orientated bias system that leaves the brain at risk of false alarms or wishful thinking. From an adaptive point of view, it is argued that a good number of false alarms is outweighed by the risk of a single serious warning or important reward being ignored. In further examining the brain mechanisms involved here, the superior collicus is identified as the brain region that initiates eye movements. Inhibitory activity in a region known as the substantia nigra driven by the neuromodulator, GABA, prevents the superior collicus from reacting too easily. One suggestion is that the caudate acts to remove inhibitory activity in the substantia nigra. This then allows dopamine based signals to be sent back to the caudate area, to allow enhanced anticipatory activity, and thus lift the inhibitory blockade on eye movement. Thus anticipatory activity in the caudate can bring the relevant neurons in the cortex close to the threshold for action. It is also suggested that dopamine may be involved in altering synaptic weightings in learning. A study by Reynolds, Hyland & Wickens (7.) showed dopamine input strengthening the synapses between neurons in the cortex and the basal ganglia. This is taken to mean that a dopamine signal evoked by one laboratory trial could lead to an amplification of the caudate response in the next trial. In a study by Nakamura & Hikosaka (8.) electrical stimulation of the caudate applied after an eye movement reinforces future eye movements. The loop between cortex, caudate and substantia nigra could support a self amplifying readiness to respond to external signals. The process described here explains how the brain reacts more quickly to wished-for or reward objects, but it is also suggested to also work for the quick reaction to fear or phobia objects. Other studies described by the author suggest that mammals use prefrontal regions such as the orbitofrontal and the anterior cingulate to generate predictions of error based on recent decisions, thus building a greater flexibility into behaviour. The Choice Problem: The most interesting section of this book deals with the choice problem. How does the brain cope when it has to deal with a conflict between two different stimuli plans or strategies, for instance the conflict between a short-term reward or the deferred enjoyment of a larger but longer term reward. Similarly there can be a conflict between habitual responses and the prospect of getting a reward for a non-habitual response. These situations are evaded in the rather superficial discussion that passes for the mainstream consensus on freewill, which typically refer to timing decision for trivial pre-planned actions or at best a single instance of giving in to an unwished for impulse. The author discusses effort-related decisions from the point of view of the influence of neuromodulators, speculating as to whether there are signals to inhibit a discarded course of action, or whether there is simply stronger dopamine support for the favoured course of action. It is also suggested that the neuromodulators, serotonin might be involved in this type of decision process. A study by Greene et al (9.) showed strong activation in emotional processing areas of the brain when moral decisions were involved. This carries with it the possibility that concepts such as fairness could create a dopamine-based bias in favour of particular decisions. These decisions also require the operation of some kind of weighting mechanism once again probably involving neuromodulators. What is only hinted at here is that the impact of neuromodulators such as dopamine is often registered in subjective consciousness, so what is sometimes being balanced is two subjective impulses. Relative to this, the author gives the further view that knowing, understanding or remembering something is more like a feeling than the result of a completely detached rational process. It is suggested that these processes are only partly dependent on computation, and that the brain may have evolved some short cuts based on neuromodulators. References:- 1.) Carpenter, R.H.S. (1981) - Oculomotor procrastination - In:- Eye Movements: Cognition and Visual Perception (pp. 237-46) - Lawrence Erlbaum 2.) Carpenter, R.H. S. (1999) - A neural mechanism than randomises behaviour - Journal of Consciousness Studies, 6, pp. 13-22 3.) Carpenter, R.H.S. (2004) - Contrast, probability and saccadic latency: Evidence for independence of detection and decision - Current Biology, 14, pp. 1576-80 4.) Carpenter, R.H.S. & Williams, M.L.L. (1995) - Neural computation of log likelihood in control of saccadic eye movements - Nature, 377, pp. 59-62 5.) Schultz, W. et al (1993) - Response of dopamine neurons to reward - Journal of Neuroscience, 13, pp. 900-913 6.) Schultz, W. et al (1997) - A neural substrata of prediction and reward - Science, 320, pp. 1638-43 7.) Reynolds, J. et al (2001) - Reward related learning - Nature, 413, pp. 67-70 8.) Nakamura, K. & Hikosaka, O. (2006a) - Facilitation of saccadic eye movements by post-saccadic stimulation in the caudate - Journal of Neuroscience, 26, pp. 12885-95 9.) Greene, J. et al (2001) - Emotional engagement in moral judgement - Science, 293, pp. 2105-8 |
| Email: quantum-mind@blueyonder.co.uk Copyright © Quantum Mind 2012. Web design by 7Soft.co.uk. |