Scholarlink

This chapter provides a brief introduction to engineering psychology experiments as an approach for understanding the influence of automated and/or autonomous agent design on human behavior in integrated systems control. This is extended with the identification of historical automation design approaches for human-in-the-loop systems (HITL), including function allocation and taxonomies of levels of automation, as well as prior criticisms and counterarguments, some of which are based on empirical inquiry. Subsequently, the chapter poses a research question as to whether design for autonomous agents might resolve some of the long-standing arguments related to system function allocation strategies (for preliminary systems design). The chapter compares automated and autonomous agents in terms of fundamental capabilities with the contention that automated agents exhibit independence in task performance but do not demonstrate other key behaviors of autonomous agents. This is followed by discussion of how design for automation may be human-centric (versus machine-centric) and how such design values/principles could be extended to artificial intelligence (AI) agents. Subsequently, a definition of human control in interaction with autonomous agents is provided along with a review of existing frameworks for design of human-autonomy/AI teaming (HAT), including several key factors of agent reliability, automation transparency, and channels and modalities of communication among agents and their influence on human control. This review is complemented by discussion of application domains that will push the design of autonomous agents further as well as how we will support agent adaptability and the need to conduct verification and validation of autonomous capabilities to facilitate trust and teaming. Finally, the chapter identifies future research directions towards achieving human-centered design of autonomous agents that effectively support both learned and orienting behaviors of human partners by ensuring that design features make clear the capabilities and potential impact of the agent.

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

From Automation to Autonomy Through AI: Enabling and Retaining Human Controllability

  • 工具书词条

  • 发表日期：

    2026年

查看全文
  • David Kaber

摘要

This chapter provides a brief introduction to engineering psychology experiments as an approach for understanding the influence of automated and/or autonomous agent design on human behavior in integrated systems control. This is extended with the identification of historical automation design approaches for human-in-the-loop systems (HITL), including function allocation and taxonomies of levels of automation, as well as prior criticisms and counterarguments, some of which are based on empirical inquiry. Subsequently, the chapter poses a research question as to whether design for autonomous agents might resolve some of the long-standing arguments related to system function allocation strategies (for preliminary systems design). The chapter compares automated and autonomous agents in terms of fundamental capabilities with the contention that automated agents exhibit independence in task performance but do not demonstrate other key behaviors of autonomous agents. This is followed by discussion of how design for automation may be human-centric (versus machine-centric) and how such design values/principles could be extended to artificial intelligence (AI) agents. Subsequently, a definition of human control in interaction with autonomous agents is provided along with a review of existing frameworks for design of human-autonomy/AI teaming (HAT), including several key factors of agent reliability, automation transparency, and channels and modalities of communication among agents and their influence on human control. This review is complemented by discussion of application domains that will push the design of autonomous agents further as well as how we will support agent adaptability and the need to conduct verification and validation of autonomous capabilities to facilitate trust and teaming. Finally, the chapter identifies future research directions towards achieving human-centered design of autonomous agents that effectively support both learned and orienting behaviors of human partners by ensuring that design features make clear the capabilities and potential impact of the agent.