Keywords
Performance Assessment, Information Theory
Abstract
Human performance in many visual experiments is assessed using probabilities of specific responses to uncertain visual events. Using signal detection theory with specific distributional assumptions, we frequently represent the combination of correct detections with correct negative responses to estimate with measures of discriminability, d', and a response criterion. A shortcoming of this approach involves limitations in generalizing the performance measures across different tasks, especially in experiments with dual or multiple tasks that require comparisons of the performance of simultaneous tasks. Our research aims to develop a framework enabling us to estimate the cognitive resources allocated to different tasks.
We propose an approach to performance representation using an information-theoretic framework where the participants' performance can be expressed in terms of input and output entropy computed from response probabilities. In this presentation, we will first review the connection between signal detection theory and an information-theoretic representation of performance expressed in terms of channel capacity allocated to different tasks. The information-theoretic framework enables the characterization of cognitive load by the total channel capacity allocated by the participant to each concurrent task. To illustrate the approach, we will present a new analysis of multiple object tracking task using an information-theoretic framework and its use to define cognitive load.
Start Date
15-5-2024 9:30 AM
End Date
15-5-2024 10:30 AM
Information Theory Representation of Perceptual Processes
Human performance in many visual experiments is assessed using probabilities of specific responses to uncertain visual events. Using signal detection theory with specific distributional assumptions, we frequently represent the combination of correct detections with correct negative responses to estimate with measures of discriminability, d', and a response criterion. A shortcoming of this approach involves limitations in generalizing the performance measures across different tasks, especially in experiments with dual or multiple tasks that require comparisons of the performance of simultaneous tasks. Our research aims to develop a framework enabling us to estimate the cognitive resources allocated to different tasks.
We propose an approach to performance representation using an information-theoretic framework where the participants' performance can be expressed in terms of input and output entropy computed from response probabilities. In this presentation, we will first review the connection between signal detection theory and an information-theoretic representation of performance expressed in terms of channel capacity allocated to different tasks. The information-theoretic framework enables the characterization of cognitive load by the total channel capacity allocated by the participant to each concurrent task. To illustrate the approach, we will present a new analysis of multiple object tracking task using an information-theoretic framework and its use to define cognitive load.