# Self-Generated Expectations of Hazard Prevalence Affect Virtual Search and Rescue

## TL;DR

This study shows how prior expectations about hazards can affect search performance in virtual rescue missions, even when new information is given.

## Contribution

The study reveals how outdated hazard expectations can override explicit instructions in high-stakes virtual search tasks.

## Key findings

- More hazards and false positives occurred when test prevalence was lower than during training.
- False negative errors remained consistently low across all conditions.
- Outdated expectations led to persistent search errors despite explicit instructions.

## Abstract

To understand how prior expectations and instructions about hazard prevalence affect high-stakes visual search in a semi-immersive virtual environment, where participants take on the role of firefighters in search and rescue missions.

Information about target prevalence influences visual search in standard laboratory studies. However, little is known about how prior expectations and new information about target prevalence interact in simulated emergency scenarios.

Participants (n = 48) received training where the average number of hazards (explosive cylinders) amongst similar distractors was varied (two or six) before participants rescued a trapped person. Trial-end feedback indicated whether all targets were removed and the person rescued. They were then instructed that hazard prevalence would increase, decrease, or stay similar during test blocks. Stress was manipulated by an ongoing alarm, the threat of trial-ending explosions, and reduced movement speed. Search performance was measured by the number and type of stimuli removed and stress was assessed using self-report and physiological measures.

Across high and low stress conditions, more hazards were removed and more false positives occurred (i.e., more distractors removed) when test prevalence was lower than during training, compared to when prevalence levels remained similar. False negative errors were consistently low across conditions.

Acquired hazard expectations can override explicit instructions, leading to persistent search errors, likely due to difficulties in adjusting decision criteria.

These results suggest that training in high-stakes hazard search should incorporate the use of tools and techniques to help mitigate the persistent influence of outdated expectations on search performance.