Improving measurements of similarity judgments with machine-learning algorithms

Intertemporal choices involve assessing options with different reward amounts available at time delays. The similarity approach to intertemporal choice focuses on judging how similar and delays are. Yet we do not fully understand the cognitive process of these judgments are made. Here, use machine-learning algorithms predict (1) investigate which best judgments, (2) assess predictors most useful in predicting participants’ (3) determine minimum number required accurately future judgments. We applied eight for amount delay made by participants two data sets. found that neural network, random forest, support vector machine generated highest out-of-sample accuracy. Though networks machines offer little clarity terms a possible making forest generate decision trees can mimic computations human judgment making. also numerical difference between values or was important predictor replicating previous work. Finally, performing such as make highly accurate predictions relatively small sample sizes (~ 15), will help minimize numbers extrapolate new value pairs. In summary, provide both theoretical improvements our understanding involved well practical designing better ways collecting data.