Presentation
From Automation to Adaptation: A Systematic Review of Adaptive Human-Machine Systems
DescriptionDespite improving efficiency, safety, and reliability in industrial settings, automated systems tend to make humans over-reliant on them, diminishing situational awareness and leading to human-out-of-the-loop. To address this, adaptive systems have been developed to modify behavior in response to changes in users or environmental states, keeping humans in the loop while preserving automation benefits. Reaching these systems' full potential depends on two essential tasks: 1) accurately identifying the human state as the adaptation trigger and 2) initiating appropriate and timely adaptations once triggered. Following the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) method across nine databases, this systematic review examines existing research on adaptive systems, focusing on how they detect and respond to human states. After the full-text screening, 35 papers met the inclusion criteria, and the information was organized around four core components: human state assessment, human state modeling, adaptive intervention, and human-machine interaction performance. Findings suggest that affective human state extraction must be approached through multimodal sensing. Additionally, there is little evidence of how these systems function in real-world settings or whether they measurably improve human-machine interaction performance. Finally, avoiding task interruptions such as visual content and the learning process between human-robot collaboration should be considered.
Event Type
Lecture
TimeWednesday, October 15th1:30pm - 1:50pm CDT
LocationGrand B
Human AI Robot Teaming (AI)