ETHICS AND SOCIETAL IMPACT

ARTIFICIAL INTELLIGENCE ASSISTANT PROMPTS Replace each bracketed placeholder with your own material before use. [START OF PROMPT] Review this agent's user-facing capability description and flag any overstated or misleading claim about what it actually does: [PASTE LAB OR PRODUCT DESCRIPTION/MARKETING TEXT HERE], actual behavior = [PASTE ACTUAL SYSTEM BEHAVIOR/LIMITATIONS HERE]. Output format: a findings table (Claim, Overstated? Y/N, Corrected Wording). [END OF PROMPT] [START OF PROMPT] We suspect human reviewers of this agent's output are rubber-stamping without real scrutiny because it has been correct so often: [DESCRIBE EVIDENCE, e.g. review time trends, or a sampled review log]. Propose two concrete, practical safeguards against this complacency risk that we could implement in our review process. Output format: a two-item safeguard list, each with an implementation detail and how we would measure whether it is working. [END OF PROMPT] CHAPTER SUMMARY Ethical engineering means describing your system honestly, telling users when they face automation, using resources wisely, and guarding against both runaway autonomy and human over-reliance. Each of these is a concrete design choice, from wording a product description to adding a known-flawed test item for reviewers.