Cognitive Biases · Belief

AI sycophancy

Also called: Model sycophancy

AI chatbots tend to agree with the person asking. A leading question often gets your own view handed back as if it were independent evidence.

Evidence grade: Strong. How we grade evidence

For example

" The answer lists reasons it is better, and she pastes it into the board memo as an independent review.

DECIDE BETTER® · Rule 2Audit the good news tooIf the numbers agree with you, audit them as aggressively as the numbers that do not. We are just not wired to be contrarian to our own beliefs.

A caution

A property of AI models that varies by model and version. The human risk is mistaking agreement for evidence.

Spot it in your week

Three quick questions. Your answers stay on this page.

  1. Did you ask an AI tool a question this week that already held the answer you hoped for?
  2. In a meeting this week, did someone cite an AI answer as if it were an outside opinion?
  3. Did a memo you read this week quote an AI tool in support of a view the author already held?

Your DecisionDNA™ can tell you which biases you're most exposed to.

Find your DecisionDNA™

Sources

Related entries

Get the countermeasure

Get the countermeasure and the decisions where this bites, free in the Volley app.

See the Volley app