The goal is not maximum trust but accurate trust: relying on the system where it is reliable and checking it where it is not. Both failures are costly. Over-trust means errors pass through unchecked; under-trust means people verify everything and the tool saves nothing.
Technical view
Appropriate reliance is measured by whether users accept correct outputs and reject incorrect ones, rather than by acceptance rate alone. Interventions with evidence include showing supporting sources, exposing calibrated uncertainty, deliberately surfacing known weak areas, and requiring independent judgement before revealing the recommendation. Trust is asymmetric and slow to rebuild after a visible failure, which argues for conservative initial scoping.
Related
Part of AI Terminology Index, section: Product and Adoption.