Decoding fairness motivations - repository

<div>Repository of Data, Scripts and Files </div><div><b><br></b></div><div><b>Neural Data:</b></div><div><br></div><div><u>Files:</u></div><div><br></div><div><i>Neural_data_D1.zip </i>- the neural data from study 1</div><div><i>Neural_data_D2.zip</i> - the neural data from study 2</div><div><i>ROIs.zip</i> - Regions of interest extracted from neurosynth masks (Cognitive Control & ToM)</div><div><i>Searchlight_ROIs.zip</i> - Regions of interest extracted from the Searchlight results</div><div><br></div><div>Neural data and regions of interest (ROI) used in the analyses. </div><div>Specifically, <i>Neural_data_D1.zip</i> & <i>Neural_data_D2.zip </i>contain t-statistics maps obtained from a single trial GLM for multivariate pattern analysis (see details below). </div><div>Further, onset files for the decisions phase of each trial are included. </div><div>In addition, Region of interests in the<i> ROI.zip</i> file are regions of interests obtained from neurosynth https://neurosynth.org/. </div><div>The ROIs in the <i>Searchlight_ROIs.zip</i> were extracted from the Searlight results map. </div><div>These data and ROIs were used in following paper:</div><div>S.P.H. Speer & M.A.S. Boksem (2019). Decoding fairness motivations from multivariate brain activity patterns. In Social Cognitive and Affective Neuroscience . Oxford University Press</div><div><br></div><div><br></div><div><u>Processing of included data:</u></div><div><u><br></u></div><div> All fMRI data underwent the standard FSL (5.0) preprocessing pipeline. </div><div>Anatomical scans were reoriented to the FSL standard orientation and skull-stripped. </div><div>The functional data was motion corrected to the mean image using FSL’s MCFLIRT and coregistered to the anatomical scan and normalized to the standard MNI brain using boundary-based registration (FSL FLIRT & FNIRT).</div><div>Subsequently, Gaussian high-pass filtering with 100 seconds FWHM was applied.</div><div>To obtain neural activation patterns for multivariate analysis individual time series were modeled using a double γ hemodynamic response function in a single trial GLM design using FSL’s FEAT. </div><div>Specifically, one GLM fitted a hemodynamic response function (HRF) for each trial, following the Least-Squares all (LSA) approach (Mumford, Turner, Ashby, & Poldrack, 2012), </div><div>using the six seconds prior to the keypress in each trial, resulting in 48 (12 trials * 4 conditions) parameter estimates of sustained activations for each participant. </div><div>Specifically, there were 12 trials of UG and DG in the social condition (human opponent) and 12 trials of both games in the non-social condition (computer opponent).</div><div> The resulting β-values were converted to t-values (Misaki, Kim, Bandettini, & Kriegeskorte, 2010), resulting in a whole-brain pattern of t-values for each trial.</div><div><br></div><div><b>Scripts:</b><br></div><div><br></div><div><u>Files:</u></div><div><br></div><div><i>Analysis_for_Repo.ipynb</i> - All analysis conducted for the paper</div><div><i>nipype_init_new.py</i> - file to set up the environment</div><div><br></div><div><br></div><div>Scripts Used for fMRI analyses, including MVPA ROI classification and searchlight analyses reported in the following paper: </div><div>S.P.H. Speer & M.A.S. Boksem (2019). Decoding fairness motivations from multivariate brain activity patterns. In Social Cognitive and Affective Neuroscience . Oxford University Press</div><div>Specifically, the scripts used for classifcation analysis on neurosynth ROIs and the scripts for the searchlight are included. All analysis were conducted in Python 2.7.</div><div><br></div><div><b>Behavioral Data:</b><br></div><div><br></div><div><u>Files:</u> </div><div><br></div><div><i>DifffereceOffers.csv </i>- Offers made by participants in Study 1</div><div><i>Diffs_W.csv</i> - Offers made by participants in Study 2</div><div><br></div><div><i>Individual-differences-in-offers2.png</i> - Plot of individual differences as illustrated in the paper</div><div><i>Individual-differences-MeanOffers.png </i>- Individual differences in mean offers in both games as illustrated in the Appendix</div><div><i>SocialvsNonSocial2.png </i>- Difference in Offers between Selfish and strategic players when playing against humans and computers</div><div><br></div><div>Behavioral Data, specfically Ultimatum Game and Dictator Game Offers and Plots resulting from behavioral analysis reported in the following paper:</div><div><br></div><div>S.P.H. Speer & M.A.S. Boksem (2019). Decoding fairness motivations from multivariate brain activity patterns. In Social Cognitive and Affective Neuroscience . Oxford University Press</div><div><br></div><div><u>Participants: </u></div><div><u><br></u></div><div>The reported analyses are based on 31 participants from two separate studies. </div><div>The reason for running two studies was driven by funding issues. </div><div>We ran out of funding half-way through the data collection and once we obtained additional funding continued with the data collection. No significant differences in demographics were identified between samples. All participants were right-handed with normal or corrected to normal vision, </div><div>and no record of neurological or psychiatric diseases. The studies were approved by the university Ethics Committees and were conducted according to the Declaration of Helsinki.</div><div><br></div><div><u>Task: </u></div><div><u><br></u></div><div>Participants played a mixed Ultimatum/Dictator Game in the MRI scanner. </div><div>On 24 trials, participants received €20 and had to decide how to split the endowment between themselves and an opponent. </div><div>On each trial they were presented with a picture of the opponent before and during the decision process,</div><div> in order to ensure that participants knew they were playing against a different human player on each trial.</div><div> The pictures used in the study were obtained from the NimStim face database (Tottenham et al., 2009). </div><div>We selected faces which were categorized as neutral and most representative of our participant population in terms of age and ethnicity. </div><div>This was done to minimize the effect of the difference in faces on offers made in the games. </div><div>Further choosing a representative sample of pictures was intended to increase the credibility of our cover story that participants played against previous participants.</div><div> On half of the trials, the opponents were able to reject the offer, which would result in both the participant and opponent receiving nothing (UG). </div><div>On the other half of the trials the opponents were passive recipients and could not reject the offer (DG). </div><div>The critical difference between these conditions is that in the UG trials the participant can be punished for unfair offers whereas in the DG trials no punishment is possible.</div><div> In the non-social control condition (24 trials), participants played against a computer algorithm, allegedly programmed to mimic human behaviour. </div><div>Again, half of the trials in the control condition were UG trials and the other half DG trials. </div><div>Practice trials were implemented to in order to familiarize the participants with the task. </div><div>In addition, participants were told that they were playing against participants who had previously participated in the study. </div><div>As mentioned above, pictures of opponents were chosen with the aim of maximizing representativeness of the sample used in order to increase credibility of the story.</div><div>The trials started with a screen that presented a picture of the opponent and their power to reject the offer or not (UG or DG). </div><div>Subsequently, the response options appeared, 0 to 14 € in steps of two, and participants could indicate their choice.</div><div> Lastly, a wait screen appeared for eight seconds.</div><div><br></div><div><b>Output Neural Analysis:</b></div><div><br></div><div><u>Files: </u></div><div><br></div><div><i>BNS-Human-KP6s-Game-Combined-permtest-5000_0.05.nii.gz</i> - map obtained from Searchlight analysis</div><div><i>Corr_classAcc-Offer-*.jpg </i>- plots of correlation between classification accuracy and difference in offer for ROIs from Searchlight analysis</div><div><i>UFS_mask_Human_CogControl.jpg </i>- plots of correlation between classification accuracy and difference in offer for the whole cognitive control map from neurosynth (after feature selection)</div><div><i>UFS_mask_Human_CogControl.jpg</i> - plots of correlation between classification accuracy and difference in offer for the whole ToM map from neurosynth (after feature selection)</div><div><i>remaining *.jpg </i>- plots showing correlation between classification accuracy and difference in individual ROIs from ToM and cognitive control network</div><div><br></div><div>Plots and neural output files from analyses and results reported in the following paper:</div><div><br></div><div>S.P.H. Speer & M.A.S. Boksem (2019). Decoding fairness motivations from multivariate brain activity patterns. In Social Cognitive and Affective Neuroscience . Oxford University Press</div><div><br></div><div>Specifically the data contains the plots illustrating the relationship between classification accuracy on game type and difference in offers between games.</div><div>In addition, the searchlight map reported in the paper is added.</div><div><br></div>