AI Role-Perspective Loop
Assign contrasting expert roles to stress-test and strengthen your thinking
- Difficulty
- Starter
- Time to result
- ~days to results
- Steps
- 5
- Confidence
- 99%
Hoffman's introductory AI technique uses the model's ability to adopt roles. Give it an argument and first ask it to act as a critic, producing the strongest case against your position. Then ask it to take your side and add a stronger supporting argument. Finally, select a domain perspective relevant to the issue, such as a historian of technology, and request that role's analysis and criticism. The loop broadens the angles considered and can reveal weaknesses, missing evidence, or better formulations. Hoffman frames this as a practical way to learn AI through matters connected to your expertise. The outputs remain generated analysis and require human judgment and external verification.
Origin
Extracted from The Diary of a CEO
Core principles
- 01AI can adopt a requested analytical role
- 02Opposing and supporting views reveal different weaknesses
- 03Domain lenses expose assumptions a generic answer may miss
- 04Useful AI capability is found through experimentation
How to run it
- 1
Supply the argument
Give the AI the actual claim, reasoning, and relevant context you want examined.
Pro tip Use a question from your real work rather than a novelty prompt.
Watch out Do not include confidential material without appropriate safeguards.
- 2
Assign the critic
Ask the model to take the role of a capable critic and argue against your position.
Pro tip Request the strongest objection rather than a list of superficial concerns.
- 3
Assign the advocate
Ask the model to support your position with an argument you have not already made.
Watch out Supporting output is not independent validation.
- 4
Change the expert lens
Select a relevant profession or discipline and ask how that expert would analyze and criticize the argument.
Pro tip Choose a role with a genuinely different method or evidence base.
- 5
Synthesize and verify
Update the argument using the strongest insights, then independently check factual claims and citations.
Pro tip Record what changed and why.
Watch out A fluent role-play can still be factually wrong.
In the wild
Hoffman proposes giving AI his argument that general-purpose technologies repeatedly provoke fears about social collapse. He would then ask how a historian of technology would analyze and criticize that claim, adding a disciplinary lens beyond a generic pro-and-con response.
→ The argument receives a targeted critical perspective that can guide further research and revision.
Common mistakes
Stopping after one answer
The method gains value by switching sides and roles rather than accepting the first plausible response.
Confusing fluency with truth
Role prompting can improve perspective without making generated claims reliable evidence.
Is it for you?
Best for
Writers, professionals, and founders who want a fast adversarial review of consequential thinking.
Not ideal for
Decisions where unverified model output would be treated as authoritative evidence.
From the transcript
“Take the role of critic. Argue against me.”
“Take my side. Give me another argument.”
From the episode
Reid Hoffman, LinkedIn Founder: It’s Time To Quit Your Job When You Feel This! Selling LinkedIn For $26 Billion Taught Me This About Life! Trump Is Going To Punish Me!