Should I Sample or Should I Go? An approximately optimal model for deciding when to stop sampling information

People are often faced with decisions for which they need to sample noisy information from the environment. Sequential sampling models provide valuable insight into how people navigate such decisions, but the actual sampling process usually remains a black box. We propose a computationally light linear model that can elucidate what factors people use during this sampling process, and whether they are optimal in doing so. We simulated agents using our model on expanded judgement tasks with different error cost (Study 1) and sampling cost (Study 2) scenarios to determine the optimal strategies in each condition. We then tested human participants in these scenarios to see if they behave optimally and if our model could capture their sampling decisions. We found that our model fit human data well and that people could shift their sampling strategy in an optimal direction when the cost of making an error changed. When sampling cost was manipulated, however, we observed a non-optimal shift in sampling strategy. This study contributes novel insights into the effects of symmetrically manipulated cost, as well as the optimality and use of dynamic decision boundaries.