Do You Use AI to Think, or to Avoid Thinking?
The right amount of trust is not the maximum, and it is not zero
Long before anyone was talking to chatbots, researchers were studying how people rely on machines that are good but not perfect - autopilots, medical software, forecasting tools. They found two opposite failures, both expensive. Some people trust too much and stop checking. Others abandon a system that outperforms them the first time it gets something wrong, and go back to doing it worse by hand. Both are failures of calibration rather than of intelligence, and everything found about earlier tools applies directly to this one.
Before you start There are no right answers. Pick what is usually true for you, not what sounds best.
Answer about how you actually use these tools in an ordinary week - work, writing, decisions, looking things up. There is no score here that makes you a better or worse person.
What it measures
Three things that decide whether a good tool helps you. Reliance is how much you hand over. Verification is whether you check what comes back. Substitution is whether the tool is doing the thinking instead of carrying it - the difference between a calculator that saves you arithmetic and one that means you can no longer estimate whether an answer is plausible.
Based on: Trust in automation (Lee & See, 2004), algorithm aversion (Dietvorst et al., 2015) and cognitive offloading (Risko & Gilbert, 2016)
Possible results
Reliance
How much you hand over.
Verification
Whether you check what comes back.
Substitution
Whether it is carrying the thinking or replacing it.
Common questions
Is relying on AI making me worse at thinking?
Nobody can answer that yet with evidence, and the honest position is that it depends on order rather than on amount. Handing over work after you have formed a view is what people have always done with tools and there is no sign it costs anything. Handing over the forming of the view is different in kind, because the thing being offloaded is the practice, and practice is how a skill is kept. The useful check is not how much you use it but whether you could still do the thing without it.
Is it safer to just not trust it?
No, and this is the part people find counterintuitive. The automation research treats under-trust as a failure alongside over-trust, because refusing a tool that is better than you at something costs you real accuracy. The experiments on algorithm aversion found people abandoning a system after seeing it make a single mistake, even when they had also seen it beat the human alternative - which is not caution, it is a bad trade. The target is not high trust or low trust but trust that tracks how good the thing actually is at the specific task in front of you.
What does a good profile look like here?
High reliance, high verification, low substitution. That combination describes someone getting the whole benefit of a powerful tool while keeping both the judgement and the practice. The two profiles worth a second look are the opposite corners: high reliance with low verification, which is the one the research warns about, and very low reliance driven by a single bad experience, which quietly costs you something every week without ever announcing itself.
Sources
- Lee, J. D., & See, K. A. (2004). Trust in automation: Designing for appropriate reliance. Human Factors.
- Dietvorst, B. J., Simmons, J. P., & Massey, C. (2015). Algorithm aversion: People erroneously avoid algorithms after seeing them err. Journal of Experimental Psychology: General.
- Risko, E. F., & Gilbert, S. J. (2016). Cognitive offloading. Trends in Cognitive Sciences.
An honest note The research this draws on is real and well established, but none of it was done on the tools you are using now. Trust in automation, algorithm aversion and cognitive offloading were all studied with earlier systems - autopilots, forecasting software, notes and reminders - and the findings transfer by argument rather than by direct evidence. That argument is reasonable, because the mechanisms are about how people rely on things rather than about any particular machine, but it is an extrapolation and should be read as one. There is no validated scale for how people use these tools; this is a description built from three lines of work, not a measurement. Finally, the question in the title is deliberately provocative and the honest answer is that nobody knows yet what sustained use does to the skills involved. Anyone telling you confidently in either direction is ahead of the evidence.
Try one of these next
How Good Is Your Attentional Control?
Keeping your focus where you put it, and moving it when you need to
How Well Do You Get On With Your Body?
Positive body image is not the absence of complaints - it is something you can actually have
How Well Do You Listen to Your Body?
Noticing the signals from inside, and knowing what to do with them