Cognitive Enablers: Towards Interfaces that Empower Users to Notice and Contest Hidden AI Defaults of their Own Volition

LLM assistants routinely resolve what humans leave unspecified, invisibly embedding normative assumptions, framings, and values into fluent responses. Consequently, users with limited time and attention default to accepting outputs with underlying commitments they would have otherwise challenged. Existing interventions combat this overreliance through friction, in essence, coercing engagement at the expense of user satisfaction. To combat this tradeoff, we instead focus on lowering the cost of contestation through cognitive enablers: oversight interfaces that empower users to exercise judgment of their own volition. We introduce Priori, an interface that surfaces a model's high-level choices alongside actionable alternatives in a dedicated sidebar. In our user study (N=160), participants felt a greater sense of autonomy and agency in how they interacted with the AI, while reporting no additional mental effort and a strong preference for the interface over base chat. These findings suggest meaningful oversight can emerge from accessible agency rather than engineered friction.