Prompts to Avoid Tapping Into Psychological Traps While Using AI

Psychological Traps to Watch For When Using AI Tools

One key risk when working with AI tools like ChatGPT is that, unless instructed otherwise, they will start generating content immediately. That alone can derail a user's intent if they haven't yet clarified their thinking.

1. Priming Effect

This effect is well-documented in Daniel Kahneman’s Thinking, Fast and Slow, where System 1 (the fast, intuitive mode of thinking) is highly susceptible to priming. A subtle cue or piece of information can influence judgment and perception without conscious awareness. ChatGPT's first draft or suggestion may unconsciously shape how you perceive your ideas, anchoring you before you can freely explore. If you already have a pretty good idea of what you want to say, the first response from the AI can subtly influence — or "prime" — your direction. The language and framing can anchor your thinking even if you don’t intend to use it. This priming can limit creative exploration, reduce nuance, or even override your voice.

Example: Last week, I was triggered by this post from Dave Farley and wanted to write a response. I told ChatGPT I wanted to draft a reply. It immediately generated a complete answer. I resisted reading it, but curiosity got me. Instead of using it, I asked ChatGPT, "Why do you believe this would be a good answer for me?" That answer was more valuable than the draft itself, because it showed me how ChatGPT is learning from and adapting to me.

If you let the priming effect dominate you here, your content will become less you and more ChatGPT. Over time, this can weaken your original voice and make you dependent on the tool to do the thinking for you. It's not about rejecting the AI — it's about maintaining authorship and agency over your ideas.

2. Confirmation Bias

Also explored in Thinking, Fast and Slow, confirmation bias is the tendency to seek, interpret, and remember information that confirms existing beliefs. Kahneman shows how this bias operates automatically through System 1 thinking, unless actively interrupted by deliberate, effortful System 2 reasoning. In the context of AI tools, this means you must deliberately prompt the system to challenge your assumptions — otherwise it will default to agreeing with you. ChatGPT was designed to be helpful — which often means agreeable. It will likely reinforce your current thinking if you don't explicitly ask for honest, critical suggestions or opposing viewpoints. This feels good in the short term but can lead to a dangerous illusion: that your ideas are flawless.

The tool doesn’t naturally challenge you unless prompted to. Without that friction, you might start to believe you're Einstein, not because your ideas are great, but because you've been mirroring yourself back without resistance.

To avoid this trap, you must actively instruct the AI to point out flaws, suggest counterarguments, or offer devil’s advocate perspectives.

Related Resource: There’s a great article by Jordan Gibbs titled “ChatGPT Is Poisoning Your Brain”. It didn’t just help me — it provided two valuable prompt patterns that I’ve since adopted as standard practice:

critical1: Focus on substance over praise. Skip flattery, engage critically with my ideas, question assumptions, identify bias, and offer counterpoints. Don’t shy away from disagreement.        

critical2: Provide three lenses: 1) a neutral, unbiased view of the request, 2) a devil’s advocate counterpoint, and 3) an encouraging but grounded take.        

I apply these prompts to everything I write or review with ChatGPT. They consistently generate more honest, reflective, and useful feedback, and help me avoid the trap of hearing only what I want to hear.

Prompt Hygiene: Structured Kickoff for Working with AI

Use the following prompt strategy to avoid premature content generation and maintain control over your creative process — and to prevent situations like the one I experienced when responding to Dave Farley’s post. It helps create a boundary between your initial intent and the AI's eagerness to generate, giving you time to gather your own thoughts:

new document: Please wait. I’ll provide the title next, followed by a full idea and context dump. Do not generate or assume any content — only begin when I explicitly say to start.        

Then follow up with:

document title: [Your Document Title Here]        

And only after your full context dump, you give a final signal like:

okay, you can start generating content now        

This method ensures your direction and intent are fully captured before the AI starts writing, especially valuable for nuanced or multi-layered topics like psychological traps in AI usage.

Conclusion

AI tools like ChatGPT can amplify your creativity and accelerate your thinking, but only if you're aware of the subtle psychological traps that come with them. The priming effect can quietly steer your voice before you’ve formed your thoughts. Confirmation bias can convince you you’re right before you’ve been challenged. If you don’t push back against these defaults, you risk outsourcing not just your writing, but your thinking.

The solution isn’t to avoid AI — it’s to use it wisely. Treat it as a collaborator, not a crutch. Use structured prompts, like critical1 and critical2, to build in resistance and reflection. Maintain authorship. Stay critical. And keep your mind in the loop.

#AI #CriticalThinking #ChatGPT #PromptEngineering #CognitiveBias #ThinkingFastAndSlow #KnowledgeWork