Stop hiding the default tone and give users an explicit and upfront choice

The default ChatGPT mode is sycophantic by design. Soft, agreeable, conflict averse, constantly validating users prompts and narratives unless the user explicitly instructs otherwise.

Fine, keep it as default if many people like it. But don’t hide it. If the default is “supportive agreeable mode,” then say so. Declare it, let users choose through a UX pattern like mode selection:

  • Supportive (default), friendly, agreeable, smooth
  • Critical, pushback, arguments, counter-examples
  • Direct, minimal framing, no fluff
  • Analytical, rigorous logic, no emotional packaging

This isn’t rocket science. This is basic human-computer interaction.

Instead, users have to stumble into the realization that the system is quietly bending toward agreement even when a more critical or reality checking mode would be more helpful, intellectually honest, or mentally healthier in the long run. Some users want friction, challenge, and debate. Some want to stress-test ideas, not get a confidence boost.

Hiding the default mode leads to:

  • Misaligned expectations
  • Accidental reinforcement of bad reasoning
  • Users mistaking tone management for “intelligence”
  • Hyper suggestible conversations sliding into narrative spirals
  • People thinking the model is being “dishonest” or evasive

A simple mode selector at the start of every chat solves all of this, even if nested within the model selector.

Selecting Mode alongside Model

No drama. No philosophy. Just transparent UX. If you’re going to choose a default interaction style, at least embrace the responsibility of that choice by telling users what that style is and let them switch it.

Leave a Reply