[Paper Review] Display of Information for Time-Critical Decision Making

Eric Horvitz, Matthew Barry|arXiv (Cornell University)|Feb 20, 2013

Human-Automation Interaction and SafetyPsychology13 references158 citations

TL;DR

This paper proposes decision-theoretic methods for optimizing information display in time-critical, high-stakes environments, using the Expected Value of Revealed Information (EVRI) and Expected Value of Displayed Information (EVDI) to dynamically control what information is shown. The approach enhances human-computer interfaces in monitoring systems by prioritizing information based on its expected impact on decision quality, demonstrated in a NASA Mission Control application.

ABSTRACT

We describe methods for managing the complexity of information displayed to people responsible for making high-stakes, time-critical decisions. The techniques provide tools for real-time control of the configuration and quantity of information displayed to a user, and a methodology for designing flexible human-computer interfaces for monitoring applications. After defining a prototypical set of display decision problems, we introduce the expected value of revealed information (EVRI) and the related measure of expected value of displayed information (EVDI). We describe how these measures can be used to enhance computer displays used for monitoring complex systems. We motivate the presentation by discussing our efforts to employ decision-theoretic control of displays for a time-critical monitoring application at the NASA Mission Control Center in Houston.

Motivation & Objective

• To address the challenge of information overload in time-critical decision environments where timely, accurate decisions are essential.
• To develop a principled method for dynamically controlling the configuration and quantity of information displayed to users.
• To design flexible human-computer interfaces that adaptively present information based on its expected value to the decision-maker.
• To apply decision-theoretic principles to monitor system states and prioritize information delivery in real time.
• To validate the approach in a real-world, high-stakes application—NASA Mission Control Center.

Proposed method

• Introduces the Expected Value of Revealed Information (EVRI) as a metric to quantify the expected benefit of revealing additional information.
• Defines the Expected Value of Displayed Information (EVDI) as a measure of the expected utility of information currently shown to the user.
• Uses EVRI and EVDI to guide real-time decisions on which information to display, update, or suppress.
• Applies Bayesian reasoning and probabilistic models to estimate the impact of information on decision outcomes.
• Designs adaptive user interfaces that reconfigure based on EVRI/EVDI scores to optimize decision support.
• Employs a decision-theoretic framework to balance information richness against cognitive load and time pressure.

Experimental results

Research questions

• RQ1 How can information display be dynamically optimized to improve decision quality under time pressure?
• RQ2 What metrics can quantify the value of information in real-time decision contexts?
• RQ3 How can the trade-off between information completeness and cognitive overload be managed?
• RQ4 In what ways can decision-theoretic models guide the configuration of human-computer interfaces for monitoring tasks?
• RQ5 Can EVRI and EVDI effectively prioritize information in high-stakes operational environments?

Key findings

• The EVRI and EVDI metrics successfully quantify the expected utility of information in time-critical decision scenarios.
• The use of these metrics enables dynamic, adaptive control of information presentation, reducing cognitive load without sacrificing decision accuracy.
• The framework was validated in a real-world application at the NASA Mission Control Center, demonstrating improved decision support.
• The method enables prioritization of information based on its potential to alter decision outcomes, enhancing situational awareness.
• The approach supports real-time reconfiguration of displays to reflect changing priorities and system states.
• The results show that decision-theoretic control of information display leads to more effective and efficient human-in-the-loop monitoring.