#10918 | AsPredicted

As Predicted:Effects of expected value and uncertainty on exploration (#10918)


Created:       05/15/2018 06:46 AM (PT)

This is an anonymized version of the pre-registration.  It was created by the author(s) to use during peer-review.
A non-anonymized version (containing author names) should be made available by the authors when the work it supports is made public.

1) Have any data been collected for this study already?
No, no data have been collected for this study yet.

2) What's the main question being asked or hypothesis being tested in this study?
Question 1: What are the effects of expected reward value and uncertainty on people’s exploratory choice behavior?
Question 2: (How) do the effects of expected reward value and uncertainty on exploration differ between young adolescents (12-15 years old) and adults (18-30 years old)?

3) Describe the key dependent variable(s) specifying how they will be measured.
We will measure exploratory choice behavior using a gambling task in which participants make repeated choices between two stimuli (boxes) associated with different, but partly overlapping, reward (number of points) distributions (see the Conditions section for details). After each choice, participants obtain a specific number of points sampled from the reward distribution associated with the chosen stimulus. Participants’ goal is to earn as many points as possible.
The dependent variable is choice strategy: Exploitation vs. exploration. Choices for the stimulus with the highest and lowest expected reward value will be classified as exploitative and exploratory, respectively. We will derive trial-to-trial estimates of the (relative) expected value of the two stimuli using a computational model (see below).

4) How many and which conditions will participants be assigned to?
The gambling task consists of eight blocks of twenty trials, preceded by one practice block (data from the practice block are not analyzed). The reward-generating distributions associated with the two stimuli vary across blocks but remain constant within each block. The difference between the means of the two distributions is 10 or 20, in half of the blocks each, in random order. The standard deviation of the reward-generating distributions is 8 in all blocks.

5) Specify exactly which analyses you will conduct to examine the main question/hypothesis.
We will use multilevel logistic regression to examine the effects the two options’ relative expected value and uncertainty, and age group (adolescents vs. adults), on choice strategy (0=exploit; 1=explore):
First-level regressors are (i) the difference in expected value between the two options (highest – lowest), (ii) the difference in estimation uncertainty for the two options (uncertainty lower-value option – uncertainty higher-value option), and (iii) their interaction. Trial-specific expected value and uncertainty estimates are derived from a computational model (see below; we will obtain expected value and uncertainty estimates using the best-fitting computational model). Group (adolescents vs adults) will be included as second-level regressor that is allowed to interact with the first-level effects.

We will also fit two computational model to the observed choice data. In Model 1, learning is modeled using a Kalman filter. Choice behavior is modeled using a softmax rule supplemented with an ‘uncertainty bonus’ mechanism that modulates the value of each option as a function of its uncertainty (Daw et al., 2006). Model 2 is a reinforcement learning model with a constant learning rate. As in Model 1, choice behavior is modeled using a softmax rule. The value of each option that is entered into the softmax rule is modulated as a function of the option’s uncertainty, which is assumed to be a (non-linear) function of the number of observed outcomes from that option. We will fit the models to participants’ trial-to-trial choices, using hierarchical Bayesian parameter estimation, separately for the adolescent and adult group. To examine evidence for differences between the two age groups, we will compare the two groups’ posterior distributions of the population mean for each parameter.
Model comparison: We will also fit a reduced version of each model that does not capture an effect of uncertainty on choice behavior (φ=0 in Model 1; ω)=0 in Model 2. We will compare the performance of the resulting 4 different models, separately for the adolescent and adult group, using the DIC.

6) Describe exactly how outliers will be defined and handled, and your precise rule(s) for excluding observations.
Participants will be healthy adolescents (VWO/gymnasium students; 12-15 years old) and adults (university/HBO students, 18-30 years old). People with psychiatric or neurological disorders, and people who used alcohol or recreational drugs on the testing day will be excluded.
Participants who always choose the same option (i.e., never explore) on more than 25% of the blocks will be excluded. Participants who do not complete all blocks will be excluded as well

7) How many observations will be collected or what will determine sample size?
No need to justify decision, but be precise about exactly how the number will be determined.

We aim to collect data from 35 participants per group (excluding outlier participants, who will be replaced) before October 31st 2018. However, if we have only tested between 25 and 35 participants per group by October 31st 2018, we will end data collection and hence have a sample size between 25 and 35. If we have tested less than 25 participants per group by October 31st 2018, we will continue data collection until at least 25 participants per group have been tested.

8) Anything else you would like to pre-register?
(e.g., secondary analyses, variables collected for exploratory purposes, unusual analyses planned?)

Secondary analyses:
• We will repeat the regression analysis using 4 regressors separately coding for the expected value and uncertainty of the lower- and higher-value option (instead of one relative uncertainty and one relative expected value regressor).
• We will repeat the computational-modeling and regression analyses while including only the first 6 choices in each block, to facilitate comparison with previous related studies (e.g., Cogliata-Dezza et al., 2017).
• We will conduct an additional regression analysis on choice strategy with two first-level regressors coding for the linear and quadratic effects of trial.
• We will examine potential effects of sex (male vs. female), and its interaction with age group, by adding these variables as second-level regressors to the regression analysis.
• We will examine whether individual differences in intolerance for uncertainty, as measured by the Dutch version of the Intolerance of Uncertainty Scale–Short version (IUS-12), predict individual differences in exploratory behavior and/or the effects of expected value and uncertainty on exploratory behavior. We will repeat the regression analysis including participants’ IUS-12 scores (range from 12 to 60, with higher scores indicating greater intolerance for uncertainty) as a second-level regressor.