Periodic Reporting for period 1 - SOURCES (Sources of rationality: arousal and the use of rational and heuristic decision strategies)
Berichtszeitraum: 2018-08-15 bis 2020-08-14
This project aimed to advance a theoretical model of the impact of emotional arousal on decision processes, and test this model in experiments with the use of neuroscience methods. It was focused on the noradrenergic arousal system, which has its source in nucleus locus coeruleus (LC) in the brainstem and projects broadly to most of the brain. LC regulates brain arousal level, influencing the sleep-wake cycle and the stress response. The best available proxy for LC activity is pupil diameter measurable with eytracking (ET). Pupils change in diameter primarily due to lighting conditions, but they also dilate when we are emotionally aroused or when we are intensively processing information. The impact of this system on other brain processes can be quantified using electroencephalography (EEG), measuring brain electrical activity. In order to understand decision processes fully, however, one has to go beyond collecting empirical data and also build theoretical models, in order to appreciate the underlying brain dynamics.
The main objective of the project was achieved by developing a spiking neural network (SNN) model of complex decision making, in collaboration with Prof. Chris Eliasmith’s group at the University of Waterloo, Canada, the developers of the Nengo neural simulator, which helped them to create the largest functioning brain model to date. Our model was also developed in Nengo, as a large spiking neural network, containing 8000 simulated neurons with 7 000 000 connections. It simulates activity of brain parts crucial for complex decision making, most importantly the frontal cortex and the basal ganglia. The main simulated processes are evidence accumulation in the cortex and the final choice of the preferred action by the basal ganglia. These processes are controlled by inhibition mechanisms in the cortex, which, as empirical research shows, are influenced by the noradrenergic system. The model can simulate ‘patient decision makers’, who process large amounts of information before a choice or ‘impatient decision makers’, who make choices based only on one piece of information. The model was tested against behavioral data from our previous study and reproduced them accurately.
Unlike other models, SNNs can produce brain phenomena, such as synchronised oscillations of neural activity. Thus, they are in good position to link neural and behavoral data from experiments. To take advantage of this, we conducted an ET-EEG laboratory study on the impact of arousal on decision making, where we checked if arousal elicited by affective pictures impacts information processing prior to making a choice, and if this impact is reflected in pupil size and EEG. Our earlier studies show that emotional arousal and mental effort impact pupil size during complex decision making and that EEG reflects the degree of information processing prior to making a decision. With the current study, we hoped to replicate these separate findings in one dataset. Initial insights into our behavioral data highlight the importance of individual differences in anxiety for understanding the effects of experimental manipulations of arousal. Highly anxious individuals, when stimulated with arousing pictures, tend to decrease the amount of information processed before a decision, which replicates our earlier findings and points to attention narrowing under high arousal as the mechanism underlying simplified decision making. In contrast, individuals with low anxiety, when presented with arousing pictures, tend to increase the amount of information before a decision – this is a novel finding demanding explanation. The rich data from our ET-EEG experiment are currently being processed, and the further analyses should tell us how the momentary changes in arousal, indexed by pupil dynamics, translate to information processing in individuals high and low in anxiety.
To achieve the applied objective of the project, we conducted an ET usability study in cooperation with Bidfood NL company. In this study, participants interacted with Bidfood NL website to perform a task of wholesale ordering of multiple food items for a restaurant, in ‘time pressure’ and no ‘time pressure’ conditions. We wanted to know if information processing load imposed by the necessity to perform within time limit would translate to indices of mental effort – namely the subjective feeling of effort as well as pupil size increase.
Indeed, we showed that participants’ pupils were larger in the time pressure condition, indicating more processing effort, even though the overall amount of information to process was the same in both conditions. Pupil size during information processing was also positively associated with subjective measures of cognitive effort. We also showed that large pupil size was associated with a characteristic visual search pattern suggesting that high cognitive effort leads to shallow information processing. Last but not least, we showed that the subjective measure of cognitive effort predicted the liking of the website, suggesting that processing effort can negatively impact evaluations of websites. We translated these findings into practical advice for Bidfood NL and we are curently preparing a publication reporting the results.
Unlike other models, SNNs can produce brain phenomena, such as synchronised oscillations of neural activity. Thus, they are in good position to link neural and behavoral data from experiments. To take advantage of this, we conducted an ET-EEG laboratory study on the impact of arousal on decision making, where we checked if arousal elicited by affective pictures impacts information processing prior to making a choice, and if this impact is reflected in pupil size and EEG. Our earlier studies show that emotional arousal and mental effort impact pupil size during complex decision making and that EEG reflects the degree of information processing prior to making a decision. With the current study, we hoped to replicate these separate findings in one dataset. Initial insights into our behavioral data highlight the importance of individual differences in anxiety for understanding the effects of experimental manipulations of arousal. Highly anxious individuals, when stimulated with arousing pictures, tend to decrease the amount of information processed before a decision, which replicates our earlier findings and points to attention narrowing under high arousal as the mechanism underlying simplified decision making. In contrast, individuals with low anxiety, when presented with arousing pictures, tend to increase the amount of information before a decision – this is a novel finding demanding explanation. The rich data from our ET-EEG experiment are currently being processed, and the further analyses should tell us how the momentary changes in arousal, indexed by pupil dynamics, translate to information processing in individuals high and low in anxiety.
To achieve the applied objective of the project, we conducted an ET usability study in cooperation with Bidfood NL company. In this study, participants interacted with Bidfood NL website to perform a task of wholesale ordering of multiple food items for a restaurant, in ‘time pressure’ and no ‘time pressure’ conditions. We wanted to know if information processing load imposed by the necessity to perform within time limit would translate to indices of mental effort – namely the subjective feeling of effort as well as pupil size increase.
Indeed, we showed that participants’ pupils were larger in the time pressure condition, indicating more processing effort, even though the overall amount of information to process was the same in both conditions. Pupil size during information processing was also positively associated with subjective measures of cognitive effort. We also showed that large pupil size was associated with a characteristic visual search pattern suggesting that high cognitive effort leads to shallow information processing. Last but not least, we showed that the subjective measure of cognitive effort predicted the liking of the website, suggesting that processing effort can negatively impact evaluations of websites. We translated these findings into practical advice for Bidfood NL and we are curently preparing a publication reporting the results.
One way to study the mind is to build artficial systems that can mimic the brain. We went beyond traditional, fragmented models of decision making to a large SNN model which reproducing human data and simulating neural activity. This opens the possibility of integrating our model with similar efforts, into a large brain model performing complex cognitive tasks. In future, such models can be integrated with even more sophisticated models developed within large projects, such as the Human Brain Project.
Another method to understand the mind is to employ fine-grained measurement techniques that can inform us about momentary changes in brain state and link them to behavior. To this end, we went beyond the traditional ‘single method study’ by simultaneously recording eye movements and EEG, together with behavior on a complex decision task. With the rich data obtained, we will be able to answer the precise questions about the nature of brain-behavior link posited with our computational model. We have also stepped out of the laboratory to test the applicability of eyetracking in the real world. We showed that pupil size, not frequently measured in real life contexts, reflects information processing effort and predicts visual search indicative of shallow, unfocused information processing.
Our results can impact both the basic research and business. Due to the increase in computational power, large models of complex cognition are likely to become the norm rather than the exception in neuroscience, and our model is one example how it can be done. Also, using multiple measurement techniques should become the norm, due to the progress in recording hardware and software. Even though such simultaneous recordings generate a wealth of data which take long to process, this can be aided by the developments in machine data analysis. And last but not least, by showing the applicability of pupil measurement in web usability context, we demonstrate the usefulness of neuroscience to business.
Another method to understand the mind is to employ fine-grained measurement techniques that can inform us about momentary changes in brain state and link them to behavior. To this end, we went beyond the traditional ‘single method study’ by simultaneously recording eye movements and EEG, together with behavior on a complex decision task. With the rich data obtained, we will be able to answer the precise questions about the nature of brain-behavior link posited with our computational model. We have also stepped out of the laboratory to test the applicability of eyetracking in the real world. We showed that pupil size, not frequently measured in real life contexts, reflects information processing effort and predicts visual search indicative of shallow, unfocused information processing.
Our results can impact both the basic research and business. Due to the increase in computational power, large models of complex cognition are likely to become the norm rather than the exception in neuroscience, and our model is one example how it can be done. Also, using multiple measurement techniques should become the norm, due to the progress in recording hardware and software. Even though such simultaneous recordings generate a wealth of data which take long to process, this can be aided by the developments in machine data analysis. And last but not least, by showing the applicability of pupil measurement in web usability context, we demonstrate the usefulness of neuroscience to business.