Examining learning and influences on performance in visual search
Abstract
Soldiers perform a breadth of visual searchesÑscan a vehicleÕs path for potential threats, monitor radar screens, look for enemy combatants, inspect the structural integrity of military vehicles, conduct ID checks, and more. It is vital to understand what factors affect search performance and why some individuals succeed whereas others fail. The current project focuses on a particularly troublesome type of search: multiple-target search. Multiple-target searches are common in many settings, they are especially error prone, and missed targets can have life threatening implications. To better understand the nature of multiple-target search to inform both academic theory and real-world performance, we propose a multifaceted experimental approach. First, we will administer laboratorybased experiments in conjunction with a large set of individual differences measures. The goal is to reveal both the underlying mechanisms that cause multiple-target search errors and why some individuals are more capable than others. Second, we will make use of mobile app technology to examine the nature of multiple-target search from a Òbig dataÓ perspectiveÑwith billions of trials, we have the power to ask critically important questions that cannot be addressed in a laboratory setting. Finally, we will use modeling techniques and the Òbig dataÓ to understand how long-term learning occurs in visual searchÑit is critical to know how expertise evolves, and this dataset affords us a unique ability to examine it. Project plan and basic approach We will (1) administer laboratory-based tasks and assess individual differences, (2) analyze data from a mobile app, (3) combine these two approaches into one protocol and (4) model long-term learning. First, participants will complete computer-based behavioral tests and a large collection of surveys about their personality, clinical traits, and predilections. The goal is to address novel questions about the nature of multiple-target search errors and to understand variability between and within participants. Second, we have partnered with the makers of a mobile app that is a visual search game (Airport Scanner, Kedlin Co.). We can examine questions that typical laboratory-based studies do not have the power to examine by interrogating billions of trials of anonymous gameplay data. Third, we can combine these by having laboratory participants play the game; this allows us to link individual differences traits to the gameplay data for our recruited individuals. Finally, we will model trial-by-trial performance to examine learning. ProjectÕs significance for the Army Any task in which a soldier is responsible for conducting visual searches that may contain more than one target (e. g., looking for contraband, vehicle inspections) is open to dangerously high miss rates. There are two important ways to counteract this problem: (1) understand the nature of multiple-target search errors such that standard operating procedures can be appropriately modified and (2) find the right people for the job. This proposal will simultaneously address both of these solutions by combining controlled laboratory-based experiments, individual differences assessments, and the power of Òbig dataÓ from mobile app technology.
Document Details
- Document Type
- DoD Grant Award
- Publication Date
- Sep 11, 2018
- Source ID
- W911NF1610274
Entities
People
- Stephen R. Mitroff
Organizations
- Army Contracting Command
- George Washington University
- United States Army