Google

To ensure the safe application of artificial intelligence, users should adhere to established guidelines for responsible usage. This includes the promotion and application of critical thinking skills when interacting with and evaluating AI systems and their outputs.

Guidelines for Safe and Responsible AI Use

To ensure the ethical development and deployment of artificial intelligence, it is crucial to actively participate in shaping the ethical standards and guidelines that govern these technologies, rather than simply being a passive user.

Proactive Engagement with AI Ethics

Analyze the following scenario and the AI-generated text based on the principles of responsible and critical interaction with AI systems. Identify the primary issue with the AI's response and explain the most appropriate next step the user should take.

Evaluating an AI-Generated Summary

A student uses an AI to generate a paragraph for their history essay. The AI produces a well-written paragraph with several specific historical claims. The student copies and pastes this paragraph directly into their essay without checking the sources or the accuracy of the claims. Identify one major risk associated with this approach and explain how it violates a key principle of responsible AI interaction.

Analyzing AI Use in Academic Work

In a meal-substitution experiment, Buçinca, Malaya, and Gajos compared three cognitive forcing designs—requesting advice, deciding before seeing advice, and waiting 30 seconds—with simple explainable-AI designs. Among 199 retained participants, the forcing designs reduced agreement with incorrect simulated-AI advice when identifying the ingredient highest in carbohydrates. This significant finding concerned the design-category comparison, not superiority of every individual design; the overreliance difference for the complete replacement decision was nonsignificant. Overreliance remained, and lower-overreliance designs received less favorable subjective ratings. These findings concern the tested simulated-AI task, not general LLM or clinical effectiveness. [Buçinca, Malaya, and Gajos (2021)](https://doi.org/10.1145/3449287)

Cognitive Forcing in a Meal-Substitution Experiment

Ma and colleagues’ Adaptive Workflow changes the order of human and AI decisions using estimated correctness likelihoods. When the estimated human likelihood is higher, the user gives an initial judgment before viewing AI advice. When the estimated AI likelihood is higher, advice appears immediately. The user always makes the final decision. In an income-classification experiment with 289 valid responses, this workflow improved final accuracy compared with an AI-confidence-only interface. That result is specific to the tested task. Estimated likelihoods can be wrong, and the authors identify possible confirmation bias after an initial judgment. [Ma and colleagues (2023)](https://arxiv.org/abs/2301.05809)

Adaptive Advice Ordering by Estimated Correctness Likelihood

A user is interacting with an AI chatbot to get financial advice. The user describes their financial situation, and the AI provides a detailed investment strategy, including specific stock recommendations. Based on the principles of responsible and critical AI use, which of the following user actions demonstrates the most sound judgment?

Learn Before

Related