This repository provides a selection of CCOBRA models for reasoning with misinformation. The models use a cache file and can thus be evaluated quickly, once a parameter setting has been pre-trained. For pre-training (optimizing parameters), an additional script is provided, as well as for transforming the originaly experimental data (source: https://osf.io/tuw89/ [st1, st2, pretest], https://osf.io/dg85h/ [st3]) into CCOBRA-readable format. Further, optimization hyperparameters for bounded basinhopping are consistent and managed in class "optPars".
- Classical Reasoning -- People who think analytically, classify news items more accurately.
- Motivated Reasoning -- People who think analytically, classify information as correct that is favorable with respect to their own political stance.
- Fast-And-Frugal Tree: Max -- Decision Tree strategy that implements the Take-The-Best heuristic.
- Fast-And-Frugal Tree: ZigZag (Z+) -- Decision Tree strategy that implements the Take-The-Best heuristic and alternates exit directions on every cue.
- Recognition Heuristic -- News items with perceived familiarity over a certain threshold are accepted.
- Recognition Heuristic (linear) -- News items with high perceived familiarity are accepted more often.
- Classical Reasoning & Reaction Time -- People who give slow responses, classify news items as incorrect more often.
- Linear Combination: Sentiment Analysis -- Acceptance probability can be determined by sentiment analysis of a news item headline.
Further:
- Hybrid model over all above models: --- Selects best predicting model per participant.
ccobra, pandas, numpy, random, math, scipy, empath, os, csv