Deciding when to escape a mine emergency: Modeling accumulation of evidence about emergencies through Instance-based Learning

Abstract

A decision to stop work and escape a mine in a mine emergency depends on the accumulation of evidence about the emergency over time. This evidence comes as situation attributes (e.g., smoke, fire, noise, etc.) from the work environment. In this paper, we investigate how a decision of when to stop work and escape a mine is influenced by the sequence in which evidence accumulates about an emergency. We model stop work decisions of miners from a cognitive modeling perspective, and create a model based on Instance-based Learning Theory (IBLT). Three risky mine-emergency situations are created where attributes communicating the risks are presented in increasing or decreasing sequences with the same total risk, and in a flat sequence with an intermediate total risk. An IBL model is calibrated to human data in the flat sequence and later generalized to the increasing and decreasing sequences. The model is evaluated based upon a situation-awareness measure of timeliness. We expected that on account of recency, human stop work decisions will be earlier in the increasing sequence compared to the decreasing sequence. Results reveal that both human's and model's stop work decisions are earlier in the increasing sequence compared to the decreasing sequence. We discuss future directions in this research. Language: en