Eight Guiding Questions

1. Who or what might be missing from our assumptions when AI is used?

Drafting a reply with AI.

Asking AI to make a passage clearer.

Generating images for presentations or documents.

2. Are we treating AI output as if it can fully understand someone's experience or situation?

Inferring how someone feels.

Asking AI to answer from a particular person's position.

Feeling tempted to show AI someone's words and have it infer their background or personality.

3. What sources, records, or people can be used to check this AI output?

Using AI to research a topic.

Asking AI to explain a policy, law, or rule.

Putting AI-generated numbers, quotations, examples, or sources into a document or post.

4. Has AI affirmation stopped us from questioning our assumptions?

Receiving reassurance from AI about something uncertain.

Having AI organize anger or anxiety into a persuasive argument.

Wanting to feel reassured too quickly by an AI-generated proposal.

5. In AI summaries, translations, and recommendations, whose voices are carried forward and whose are lost?

Summarizing a meeting with AI.

Choosing what to read through AI recommendations or search results.

Receiving someone's words through AI translation or captions.

6. Where can people notice, stop, and correct AI-assisted judgments and actions?

Sending or publishing AI-generated text.

Letting AI recommendations influence what to buy, where to go, or what to apply for.

Using a service whose guidance changes according to what people enter.

7. If AI appears effortless, what labor, energy, or resources make that ease possible?

Instant replies that make AI feel costless.

Fast text or image generation that hides the labor behind it.

Reduced work in one place that may create checking or explanation work elsewhere.

8. Are we still using the same assumptions after the AI or the situation has changed?

Noticing that a familiar AI system responds differently than before.

Reusing a prompt even though the person or situation has changed.

Relying on an AI-generated answer or summary that may be cited repeatedly later.