Paper Type

Complete

Paper Number

PACIS2025-1614

Description

A growing research agenda in human-agent teaming (HAT), this field lacks an overview of how single humans or human teams collaborate with multiple agents. We addressed this gap by conducting a structured literature review, using an updated Input-Mediator-Output framework that examined the nexus between independent and dependent variables at a more detailed teamwork interprocess level, shedding light on the intricacies of team dynamics. This first-of-its-kind consolidation explores when human-multi-agent teaming (HMAT) excels or falters by comparing the impact of different team compositions on performance. Our findings showed that single-agent HAT experiences weaker team processes and mixed performance compared with human-only teams. HMAT demonstrates superior task allocation, workload management, and resilience under high-stakes conditions, though trust remains an area for improvement. By distilling contextual dimensions, we explained inconsistencies in findings. Based on these insights, we proposed future research avenues to advance human-agent collaboration.

Comments

HCI

Share

COinS
 
Jul 6th, 12:00 AM

A Systematic Literature Review on Human-Agent Teaming with Insights into Multi-Agent Interactions

A growing research agenda in human-agent teaming (HAT), this field lacks an overview of how single humans or human teams collaborate with multiple agents. We addressed this gap by conducting a structured literature review, using an updated Input-Mediator-Output framework that examined the nexus between independent and dependent variables at a more detailed teamwork interprocess level, shedding light on the intricacies of team dynamics. This first-of-its-kind consolidation explores when human-multi-agent teaming (HMAT) excels or falters by comparing the impact of different team compositions on performance. Our findings showed that single-agent HAT experiences weaker team processes and mixed performance compared with human-only teams. HMAT demonstrates superior task allocation, workload management, and resilience under high-stakes conditions, though trust remains an area for improvement. By distilling contextual dimensions, we explained inconsistencies in findings. Based on these insights, we proposed future research avenues to advance human-agent collaboration.