ABSTRACT
Anchoring effects are found in studies, in which participants give numerical estimates with incomplete knowledge. Selective Activation and Reconstructive Anchoring (SARA) is an associative learning model, which simulates the learning of numerical facts and explains hindsight bias and anchoring effect. SARA combines a process model for sequential recall of information from memory with the qualities of a connectionist learning model. SARA has two memory storages: Long-term memory and working memory. SARA allows parameter-free predictions for experimental conditions. SARA successfully explains the anchoring effect and also the difference between children and adults without using different parameters for children and adults. The only difference was the parameter-free mapping of the knowledge to external quantities. By using the data of a control condition for calibration, a sufficient match between simulated and experimental data could be achieved, leaving all parameter fixed.
