A position paper from arXiv proposes five interaction design principles under the name SPEAR. The authors argue that alignment, making an AI agent behave as intended, is an interaction design problem as much as a technical one. The framework shifts the question from how an agent is trained to how a human and an agent negotiate intent in real time. The five principles address five recurring gaps. A human needs to share intent before an agent starts. A human needs to see what the agent is doing while it runs. A human needs to redirect or stop the agent mid-task. A human needs to judge the finished work with enough information to decide. A human needs to correct the record when the agent has learned something wrong. Each principle names a behavior, not a setting. For a product team, the paper offers a checklist to audit an agent product. The principles map onto design problems that already have names in practice: handing off a task, tracking progress, interrupting a run, reviewing output, and fixing bad memory. The paper is conceptual rather than empirical, so treat it as a frame for asking better questions rather than as settled evidence.