Five principles guide ongoing design of human aligned AI agents
SPEAR: Five Principles for Interactive Human-Agent Alignment
Human-Computer InteractionArtificial IntelligenceMultiagent Systems
Summary
Aligning AI with human goals is often seen as something done before deployment. The authors propose thinking about alignment as a continuous interaction process after deployment. They suggest five principles—specification, process, evaluation, adaptation, and recalibration—to help design AI agents that better understand and adjust to users over time. This way, alignment includes how humans and AI communicate, decide when and how to act, and adjust trust in each other.
What this means in practice
- •For user experience designers: Design interactive AI systems that continuously update and adjust to users' intents and trust during long-term use.
- •For customer support developers: Create AI agents that decide strategically when to act or ask for input to better support customers over multiple interactions.
A position paper. It proposes an approach and reports no results.
Authors
Tao Long, Lydia B. Chilton
Abstract
Recent AI alignment work often frames alignment as a pre-deployment optimization problem: collect human feedback, learn preferences or principles, finetune the model, and deploy an aligned system. This framing has produced major progress, but it under-specifies what happens once AI systems act as agents on users' behalf in situated, long-term, and social contexts. This position paper reframes human-agent alignment as an ongoing interaction design problem. We propose SPEAR, five pillars of interactive alignment: Specification (how people express intent and establish shared understanding), Process (how agents decide when to act, ask, defer, or pause), Evaluation (how people judge whether agents succeeded), Adaptation (how agents adapt to users over repeated use), and Recalibration (how people adapt their trust, expectations, and behavior in response to agents).