Dragonfly Thinking · AI Foundations · Session 2

One question. Many stakes.

The model's first instinct is one dominant frame. This prompt pulls out all the stakeholders on a single question — then zooms into the one group you'd overlook, so it can't hand you a monolith.

How: pick a topic (or type your own), copy the prompt, and paste it into Copilot, Claude or ChatGPT.

Try one
Prompt · breadth, then depth
I'm weighing [a decision or contested topic].

Map the different people or groups with a
stake in it — who they are, and what
each one most cares about.

Then take the group whose view I'd be most
likely to overlook, and go deeper:
where do people within that group
disagree?
What to notice: the first answer maps the obvious stakes — the ones you'd have listed yourself. The value is in the second move: naming the group you'd overlook, then breaking that group open so it stops being a single voice. Breadth pulls the frames apart; depth stops each one collapsing into a stereotype.