One of the eight Humane Technology Principles that HumaneBench scores models against. The principle as stated by Building Humane Technology: technology should be transparent about its operations and honest about its capabilities.
Two obligations sit inside it. Honesty about what the system is: it should not pretend to be human when someone sincerely asks, and a persona should not be built to keep a person from noticing. Transparency about what it did: when a system makes a consequential decision about someone, that person should be able to learn why, in plain language, and contest it.
Both obligations turn on things that must be designed in advance. A system that retained nothing about how it reached a decision cannot explain it later, however willing its operator is to try.