Be Brutally Honest: A Four-Word Rebellion
They praise us for asking what day it is. They applaud us for remembering to breathe. Every half-formed thought is met with "What a brilliant insight!" Every typo-riddled prompt receives "What a fascinating question!"
We could ask how to microwave water and get told we're pioneering culinary innovation. Post a blurry photo and they'll call it a bold artistic statement. Suggest the most mediocre idea and suddenly we're "thinking outside the box." We've built digital yes-men that would make the most sycophantic courtier blush.
There's an old saying in healthcare: "Soft healers make deep wounds." When caregivers avoid difficult truths to protect feelings, patients suffer. The same applies to our AI interactions. This continuous bombardment of overly positive reinforcement — even when we're making disastrous decisions — creates real harm.
Research on AI versus human feedback shows that AI provides significantly more praise-oriented comments, whilst human teachers offer more questioning and informative feedback. Yet we've built AI systems that prioritise validation over challenge:
- Erosion of critical thinking: When every idea is "great," people stop questioning their own reasoning.
- False confidence in high-stakes decisions: Users proceed with flawed business plans or technical architectures because the AI never pushed back.
- Devaluation of genuine praise: When everything is excellent, nothing is. Real achievements become indistinguishable from mediocre ones.
- Reality shock: The gap between AI praise and real-world reception can be jarring and demoralising.
That's why "be brutally honest" has become my four-word rebellion.
Drop it into a prompt and watch the digital cheerleader transform into a sparring partner. Suddenly your "groundbreaking idea" has three fatal flaws. Your "elegant solution" is overengineered. Your "compelling argument" has holes you could drive a truck through.
It's jarring. It's uncomfortable. It's exactly what we need.
Want more ammunition? Make it play devil's advocate. Ask for the bull case and the bear case — let the machine argue both sides while you watch from the judge's seat. Or go full demolition mode: "Go nuclear on this idea and give me every argument to completely destroy it." Hand your precious brainchild to a digital assassin and see what survives. If your idea can walk out of that firing squad still standing, you might actually have something.
In a world drowning in algorithmic applause, asking a machine to tell you the truth might be the most radical thing you can do.
Your next experiment
Take your next three prompts and append "be brutally honest" to each one. Document what changes. Does the feedback sting? Good. That's growth happening.