[BEAST in Deep Learning] Cognitive Model Priors for Predicting Human Decisions

Time: Thursday 23-Jul-2020 23:30 (This is a past event.)

Discussion Facilitator:

Artifacts

Motivation / Abstract
The work elaborates on previous attempts to outperform cognitive models developed by social scientists by ML algorithms. The authors use the BEAST model to create a synthetic dataset and internalize its weight in a neural network. Unlike previous attempts, the model can work only on raw data without exhaustive feature engineering and achieve unseen performance. 
Questions Discussed
- What is the BEAST model
- synthetic dataset for human choice decisions: advantages and disadvantages 
- application of cognitive model priors in ML products
- Graph NN for predicting human decisions
Stream Categories:
 ML in Economics