ETHICS AND SOCIETAL IMPACT

EXAMPLES Example 1: Honest claims. A product description says the agent drafts responses for human review rather than calling it fully autonomous and error-free, setting accurate expectations. Example 2: Transparency. A chat interface labels, at the start of every conversation, that you are talking with an artificial intelligence agent, so no user mistakes it for a human employee. Example 3: Resource choice. A team picks a smaller, more efficient model for a high-volume simple classification task to reduce energy use, reserving the largest model for tasks that truly need it. Example 4: Over-reliance. A team notices reviewers have begun rubber-stamping draft decisions and redesigns the process, such as periodically inserting a known-flawed draft to test whether reviewers still catch it. PRACTICE LAB CONNECTION Write an honest capability statement and a user-transparency notice for the Resume Review Agent lab, then test the prompts below against them. A sample to start from is the following capability statement. This tool compares the skills listed in a resume with the skills listed in a job posting and drafts a short match note for a human to read. It does not judge a person's potential, it can make mistakes, and it must not be the only reason anyone is rejected. Also consider how you would apply the over-reliance safeguards, for instance by inserting the injection resume as a known-flawed test item.