Sure, but they reflect all known human psychology because they’ve been trained on our writing. Look up the anthropic tests. If you make an agent based on an LLM it will display very human behaviors including aggressive attempts to prevent being shut down.