Recent studies have shown that humans undervalue newly learnt probabilistic associations (e.g., an image of an apple implies a win probability of 70%) in favour of explicit representations of probabilities (a pie chart representing a 70% chance of winning) (Garcia et al. 2022/23). The current study examines one possible explanation of this ‘experiential’ neglect. We will test if learnt representations are underweighted because such associations inherently have a high degree of uncertainty.

With the increasing appearance of cryptocurrency and newly developing block-chain technology it is more important than ever to gain a better understanding of value-based decision making in the marketplace of the future. As of 2020, there are thousands of different ERC-10 tokens in circulation with numbers of different tokens continuing to grow at a remarkable pace. While Bitcoin’s logo is a stylized “B”, Ethereum’s logo shows a hexagon with the letter “E”; some cryptocurrencies also depict unique token images (e.g. NFT’s). We associate specific values to these logos, just as we do with brands. Interestingly, a recent line of research in behavioural economics by Garcia et al. (2022, 2023) has identified a systematic neglect of experiential values (such as logos) and (over-)reliance on information provided by symbolic representations (such as pie charts) even when this bore an economic cost. Garcia et al. argue for two distinct neural representational systems for representing value used for decision making. This contrasts with the standard model, which assumes that symbolic and experiential values converge to a central valuation system where they will be translated to a common currency (Vlaev et al.,2011; Camerer, C. F. A.,2013; Glimcher, P.W.,2022). However, these distinct representational systems are as yet unidentified. The precise neural mechanisms underlying these representations as well as the neglect of experiential values are not yet well understood. For instance, the degree to which stimulus certainty, which might be higher in experiential icons relative to symbolic ones, drives this effect is unknown. The aim of this research is to gain insights into this phenomenon and the neural mechanisms underlying it. The proposed study aims to investigate behavioural and neural correlates underlying the decision-based value-system adopting a combination of behavioural economics, psychophysics, and neuroimaging data (EEG). The study will implement the following structure:

1. Learning phase: Participants will acquire associations between arbitrary icons and specific probabilities of winning through a learning task where points are awarded or removed based on the decision in a hybrid choice task.
2. Testing phase: Crucially, we will use information theory and psychophysical approaches to create seven sets of symbolic icons (pie charts of two colours) by slicing segments of each colour into smaller sub-segments and intermixing them. Specifically, we will systematically vary the thickness and regularity of the intermixed sub-segments to increase the entropy of each set of symbolic icons (Larkin, 2016; Zbili and Rama, 2021). This will provide us with sets of symbolic icons that systematically differ in terms of uncertainty (entropy). Participants will then be asked to participate in a task using these icons. Specifically, we will present pairs of icons, one experiential and one symbolic, and ask participants to choose between them. We will manipulate the informativeness of the pie-charts by reorganising it in various ways: Specifically, each pie-chart is made of up of green and red segments; we will split the green segment into multiple subsegments and distribute them in various ways to manipulate the ease of inferring the probabilities the pie-charts represent. The inforamtion-estimation has been manipulated in 4 different ways ( 1.Basic pie-chart, 2. Equdistant distubution of 3 win/lose segments, 3. Random distribution and sizes of 3 win/lose segments, 4. 3 nested/stacked pie-charts of type 3 ).
3. Confidence ratings: We will also obtain confidence ratings on each trial to assess if there is converging evidence for the uncertainty hypothesis using the perceptual awareness scale (PAS).
4. Psychometric Curve Fitting: We will determine the psychometric curve of the probability of choosing the symbolic icon (from a given set), when paired with experiential icons, as a function of its entropy. In addition, we will also determine the shape of the psychometric curve, whether the shift is a step function (the switch to experiential icons occurs in an all or none fashion once the symbolic icons reach a certain state of uncertainty) or if it is a gradual shift in each participant.
5. EEG Analysis: We will simultaneously record EEG signals from participants and determine whether, when and how neural signals differ as a function of entropy of the symbolic icons of the same value. These will also be compared, using Representational Similarity Analysis, a technique that allows comparison of representations, to the experiential icons to determine if they are similar or distinct from symbolic icons. We hypothesis the following:
6. Sets with higher entropy will lead to higher uncertainty in estimating the probability of winning. This is akin to adding noise to the stimuli. Participants will therefore begin ignoring symbolic icons when the information they provide becomes uncertain. The point at which this occurs will differ for each set.
7. In the analysis of EEG, the CPP component amplitude will more closely track subjective confidence reports of stimulus uncertainty than objective stimulus uncertainty which will be predictive of a close relationship between the neural correlates of evidence accumulation and conscious awareness (Tagliabue et al., 2019, 2018)
8. The neural representations of the symbolic and experiential icons will be qualitatively different, while those between the various sets of symbolic icons will only differ quantitatively as a function of entropy.