Two-step value normalization explains process of re-estimation of option triggered by sold-out event
Poster No:
906
Submission Type:
Abstract Submission
Authors:
Minho Hwang1, Dongil Chung2
Institutions:
1Ulsan National Institution of Science and Technology, Ulsan, Ulsan, 2UNIST, Ulsan, NA
First Author:
Co-Author:
Introduction:
Choices are often influenced by the context that changes over time. For instance, changes in an item's availability (e.g., an item being sold-out) affect subsequent purchasing choices. Previous studies reported that the presence of an unavailable option leads to more random choices, even when it is unrelated to the choice [1]. Value normalization theories, which propose that the neural encoding of the value for each option is normalized by the total value of all available options, accounted for such impacts of the choice-irrelevant option [2]. However, it is still unknown whether choices under context changes can be explained by the same value normalization mechanism. Here, we proposed a Two-step value normalization model (TSVN) that offers a potential explanation for the impacts of changes in the available options, using a context-adaptive value normalization process. Furthermore, we tested whether the normalized values estimated from our proposed model were represented in neural and physiological data.
Methods:
In the current study, 36 participants (male/female = 20/16, age = 21.15 ± 3.11) were recruited and asked to make a series of gamble choices while their gaze information, pupil dilation, and electroencephalogram (EEG) data were recorded. During the gambling task, three gamble options were presented at the beginning of each choice trial. On some trials, one of the three options became unavailable ('Sold-out' trials) before the presentation of the choice cue, prompting participants to make their decisions from the remaining two options. On some other trials, only two options were displayed from which individuals were asked to choose ('Two-option' trials). Our TSVN model proposes that the sold-out of one option triggers the 2nd-step normalization, which only takes into account the 1st-step values of the two remaining options. Pupil dilation was calculated based on the diameter of the pupil during gaze at option. The average amplitude of EEG signals during each option presentation epoch was used to calculate the event-related potential (ERP) and, event-related spectral perturbation (ERSP) was computed for each frequency band to measure oscillatory responses induced by the sold-out cues.
Results:
To test the impacts of the sold-out, individuals' choices were compared between Sold-out and Two-option trials. Linear regression analysis revealed that the proportion of individuals' choices unpredicted by their risk preference (independently measured from a typical 2AFC gambling task) increased as a function of the sum of all three options' values (ß = -0.99, p < 0.001). In a formal model-comparison against other normalization models, our TSVN model best explained individuals' choices both with and without the occurrence of sold-out options. The TSVN predicts that the value of each option can change after the revelation of its unavailability, and pupil dilation data supported this prediction. Specifically, 1st-step value differences were only significantly represented before the sold-out, while 2nd-step value differences were only represented after the sold-out. EEG data also showed evidence of the TSVN. Specifically, the amplitudes of P300 and late positive deflection (LPD) in the parietal region were positively associated with the value of the sold-out option. Furthermore, spectral strength of the gamma band oscillation in the frontal region was negatively associated with the 2nd-step value differences.
Conclusions:
Our data confirms that the change of option availability significantly affects individuals' choices. Consistent with the value normalization theories, the extent to which individuals change their choices at sold-out was associated with the sum of all options' values. More importantly, both pupil and EEG data reflected value information aligning with predictions made by our TSVN model. Together, the current study provides a neurophysiological account explaining how the context change during choice valuation affects individuals' decision process.
Higher Cognitive Functions:
Decision Making 1
Modeling and Analysis Methods:
Classification and Predictive Modeling
Novel Imaging Acquisition Methods:
EEG 2
Keywords:
Computational Neuroscience
Electroencephaolography (EEG)
Modeling
Other - Value normalization
1|2Indicates the priority used for review
Provide references using author date format
[2] Louie, K., Khaw, M. W., & Glimcher, P. W. (2013). Normalization is a general neural mechanism for context-dependent decision making. Proceedings of the National Academy of Sciences, 110(15), 6139-6144.