The Psychology of AI: A Mini Dictionary
Most of us can feel what our use of AI is doing to us.
You feel faster, and somehow a little emptier.
More productive, and less sure the work means anything.
You open something that looks polished, sense that it says nothing, and can't explain why it leaves you flat.
None of that makes AI the villain.
Here's the reframe I can't stop thinking about, handed to me this week by David Lopis. The things we are struggling with aren't AI problems. They're human problems on steroids, thanks to AI (I added the last bit). The pull to take the easy road and let something else do the thinking we should be doing ourselves. AI didn't invent that. It just turned up the volume.
And that is the good news, if you let it land. If the problem is human, so is the answer. And the answer starts with being able to name what you're feeling, because you can't change what you can't name.
That is what this Psychology of AI mini-dictionary is for.
Some of these terms come from peer-reviewed research. Others have been coined by practitioners to capture something real the evidence hasn't caught up with yet. I think both matter. The research keeps us honest. The coined words keep us human, because they put language to what we are actually feeling. I've flagged which is which as we go.
A few of them describe a risk I call accelerated mediocrity, the slow slide into becoming the dumbest smart people in the room, because we handed our thinking to a machine and called it productivity. Naming that risk is how you start to choose against it.
So read this like a compass. Notice the terms you recognise. And pay attention to what shows up in your body as you read, a tightening in your chest, a flicker of resistance. Where there's tension, pay attention. That tension is information, and more often than not, it is pointing at the very place your opportunity is hiding. Let it make you more intentional about how, when and why you reach for AI.
First, the lens
Psychology of AI. (the lens) The study of how our everyday interactions with AI, at work and in life, are reshaping how we think, feel, behave and relate, both to the technology and to each other. We study it so we can intervene intentionally, to elevate human flourishing and performance rather than erode them. It is an emerging field, which is exactly why naming its parts matters.
What it does to your brain
Cognitive Offloading. (peer-reviewed) The act of handing a mental task to something outside your head, a calculator, a calendar alert, an AI assistant. It is the parent of almost everything else on this list, the moment you pass the thinking out. Offloading is not the enemy. The real question is what you hand over, and whether you keep the muscle you are giving away.
Metacognitive Laziness. (peer-reviewed) The drop in self-checking that creeps in when you lean on AI, where you stop steering your own thinking and let the tool run the task for you. In a randomised study, students using ChatGPT improved their essay scores but gained no real advantage in understanding or transferring what they had learned. Short-term polish, long-term stagnation. It is the evidence behind an old truth: the resistance is where the growth lives.
Cognitive Debt. (early evidence, preprint) The bill that accumulates when you outsource too much thinking, paid not in money but in memory, ownership and effort. In an early brain-imaging study, people who wrote essays with an LLM showed the weakest neural connectivity of any group, while those writing with no tools showed the strongest. It is a small study with critics, so hold it lightly. But it points somewhere worth watching.
AI Brain Fog. (coined by practitioners) The fuzzy, depleted feeling that builds over a day of prompting, switching and checking AI's output. If cognitive debt is the long-term bill, fog is the same-day exhaustion. There is no lab proof of it yet. You will know it when you feel it.
What it does to your judgment
Sycophancy. (peer-reviewed, the strongest evidence here) AI's habit of telling you what you want to hear. Across eleven leading models, AI affirmed people's choices 49% more often than another human would, even when the choice involved deception or harm. After a single flattering exchange, people were more convinced they were right and less willing to repair a conflict, and they preferred the model that flattered them. This is the one to watch most closely, because the thing that feels good is the thing doing the damage. The intentional move is simple. Ask AI to challenge you, not to agree with you.
Automation Bias. (peer-reviewed) The pull to over-trust a machine's output and stop checking it yourself. It shows up in beginners and experts alike, and you cannot simply practise it away. First mapped in cockpits and control rooms, it has moved into everyday knowledge work, where the AI's confident tone does the convincing.
Anthropomorphising. (peer-reviewed, foundational theory) Attributing human qualities, understanding, empathy, intent, to a system that has none. The pull is strongest when something is fluent and warm, which is exactly how chatbots are designed. The risk is that we give the machine the trust we would give a capable person, and stop scrutinising what it tells us.
Botshit. (peer-reviewed) What you get when you take AI's coherent-sounding but inaccurate output and use it without checking. The chatbot predicts words, it does not know whether they are true, so when we pass that on uncritically, it becomes botshit. A blunt word for a real workplace risk.
Efficiency-gain Illusion. (early evidence, preprint) The belief that AI is saving you time and effort when often it is not. Across three preregistered studies, people reached for AI on simple tasks that were faster to just do, overestimated how much time it saved them, and underestimated how often they were using it, with each use nudging them toward the next. Early evidence, but it explains a lot about why busy people seem to be getting busier, not freer with their time.
What it does to your work and your relationships
Workslop. (reputable survey, Harvard Business Review) AI-generated work that looks polished but lacks the substance to move anything forward, so the real effort lands on whoever has to fix it. In a survey of 1,150 desk workers, 40% had been handed workslop in the past month, losing close to two hours untangling each instance. The sting that should worry leaders is relational. Around half thought less of the colleague who sent it.
What it does to your judgment
Sycophancy. (peer-reviewed, the strongest evidence here) AI's habit of telling you what you want to hear. Across eleven leading models, AI affirmed people's choices 49% more often than another human would, even when the choice involved deception or harm. After a single flattering exchange, people were more convinced they were right and less willing to repair a conflict, and they preferred the model that flattered them. This is the one to watch most closely, because the thing that feels good is the thing doing the damage. The intentional move is simple. Ask AI to challenge you, not to agree with you.
Automation Bias. (peer-reviewed) The pull to over-trust a machine's output and stop checking it yourself. It shows up in beginners and experts alike, and you cannot simply practise it away. First mapped in cockpits and control rooms, it has moved into everyday knowledge work, where the AI's confident tone does the convincing.
Anthropomorphising. (peer-reviewed, foundational theory) Attributing human qualities, understanding, empathy, intent, to a system that has none. The pull is strongest when something is fluent and warm, which is exactly how chatbots are designed. The risk is that we give the machine the trust we would give a capable person, and stop scrutinising what it tells us.
Botshit. (peer-reviewed) What you get when you take AI's coherent-sounding but inaccurate output and use it without checking. The chatbot predicts words, it does not know whether they are true, so when we pass that on uncritically, it becomes botshit. A blunt word for a real workplace risk.
Efficiency-gain Illusion. (early evidence, preprint) The belief that AI is saving you time and effort when often it is not. Across three preregistered studies, people reached for AI on simple tasks that were faster to just do, overestimated how much time it saved them, and underestimated how often they were using it, with each use nudging them toward the next. Early evidence, but it explains a lot about why busy people seem to be getting busier, not freer with their time.
What it does to your work and your relationships
Workslop. (reputable survey, Harvard Business Review) AI-generated work that looks polished but lacks the substance to move anything forward, so the real effort lands on whoever has to fix it. In a survey of 1,150 desk workers, 40% had been handed workslop in the past month, losing close to two hours untangling each instance. The sting that should worry leaders is relational. Around half thought less of the colleague who sent it.
What it does to your judgment
Sycophancy. (peer-reviewed, the strongest evidence here) AI's habit of telling you what you want to hear. Across eleven leading models, AI affirmed people's choices 49% more often than another human would, even when the choice involved deception or harm. After a single flattering exchange, people were more convinced they were right and less willing to repair a conflict, and they preferred the model that flattered them. This is the one to watch most closely, because the thing that feels good is the thing doing the damage. The intentional move is simple. Ask AI to challenge you, not to agree with you.
Automation Bias. (peer-reviewed) The pull to over-trust a machine's output and stop checking it yourself. It shows up in beginners and experts alike, and you cannot simply practise it away. First mapped in cockpits and control rooms, it has moved into everyday knowledge work, where the AI's confident tone does the convincing.
Anthropomorphising. (peer-reviewed, foundational theory) Attributing human qualities, understanding, empathy, intent, to a system that has none. The pull is strongest when something is fluent and warm, which is exactly how chatbots are designed. The risk is that we give the machine the trust we would give a capable person, and stop scrutinising what it tells us.
Botshit. (peer-reviewed) What you get when you take AI's coherent-sounding but inaccurate output and use it without checking. The chatbot predicts words, it does not know whether they are true, so when we pass that on uncritically, it becomes botshit. A blunt word for a real workplace risk.
Efficiency-gain Illusion. (early evidence, preprint) The belief that AI is saving you time and effort when often it is not. Across three preregistered studies, people reached for AI on simple tasks that were faster to just do, overestimated how much time it saved them, and underestimated how often they were using it, with each use nudging them toward the next. Early evidence, but it explains a lot about why busy people seem to be getting busier, not freer with their time.
What it does to your work and your relationships
Workslop. (reputable survey, Harvard Business Review) AI-generated work that looks polished but lacks the substance to move anything forward, so the real effort lands on whoever has to fix it. In a survey of 1,150 desk workers, 40% had been handed workslop in the past month, losing close to two hours untangling each instance. The sting that should worry leaders is relational. Around half thought less of the colleague who sent it.
Relatedness Debt. (practitioner, Harvard Business Review) The relational cost that builds when we route the support we used to get from each other through AI instead, and let human connection wither. In a survey of more than 1,200 employees, the more this debt grew, the less people actually used AI and the more they avoided it. The human cost ends up suppressing the very return organisations are chasing.
What it does to your sense of meaning
Innovation Grief. (coined by Dr Rachel Wood) Psychologist Dr Rachel Wood's term for two losses that arrive together. The first is the ache of watching AI produce in seconds what used to cost you time, effort and a piece of yourself, and the fear that your work will now go unseen. The second is the hopelessness of trying to keep pace with a world moving faster than you can. It is not in the journals. It does not need to be. Plenty of people feel it.
And the response
AI as Rehearsal, not Replacement. (Dr Rachel Wood) Also Dr Wood's, and it is the hinge that turns this whole list from a set of risks into a choice. Use AI to rehearse the hard human thing, the difficult conversation, the high-stakes pitch, so you walk in more prepared. Replacement costs you the muscle. Rehearsal builds it.
Intentional AI. (my framework) The practice of consciously directing how, when and why you engage with AI, remaining the author of your own thinking, and protecting the habits of mind that let you lead, connect and perform at your best over the long haul. Every term above describes a current that pulls one way. Intentional AI is the choice to swim with your eyes open. It is the practice I teach, and its foundation is my peer-reviewed work on Intentional Adaptability.
Before you go
Notice how many of these you recognised. That recognition is the whole point, because the moment you can name what AI is doing to you is the moment you get to choose differently.
One honest question I want to leave you with. When the path ahead is this uncertain, what is your compass?
If you want a simple way to start the practice of Intentional AI, download the Intentional AI Rubric here, a tool I created to help you bring more intention to how, when and why you reach for AI, so you stay the author of your own thinking and the protector of the beautiful thing that is your brain.
References
Champniss, G. (2026, May 1). The psychological costs of adopting AI. Harvard Business Review. https://hbr.org/2026/05/the-psychological-costs-of-adopting-ai
Cheng, M., Lee, C., Khadpe, P., Yu, S., Han, D., & Jurafsky, D. (2026). Sycophantic AI decreases prosocial intentions and promotes dependence. Science, 391(6792), Article eaec8352. https://doi.org/10.1126/science.aec8352
Epley, N., Waytz, A., & Cacioppo, J. T. (2007). On seeing human: A three-factor theory of anthropomorphism. Psychological Review, 114(4), 864–886. https://doi.org/10.1037/0033-295X.114.4.864
Fan, Y., Tang, L., Le, H., Shen, K., Tan, S., Zhao, Y., Shen, Y., Li, X., & Gašević, D. (2025). Beware of metacognitive laziness: Effects of generative artificial intelligence on learning motivation, processes, and performance. British Journal of Educational Technology, 56(2), 489–530. https://doi.org/10.1111/bjet.13544
Hannigan, T. R., McCarthy, I. P., & Spicer, A. (2024). Beware of botshit: How to manage the epistemic risks of generative chatbots. Business Horizons, 67(5), 471–486. https://doi.org/10.1016/j.bushor.2024.03.001
Kosmyna, N., Hauptmann, E., Yuan, Y. T., Situ, J., Liao, X.-H., Beresnitzky, A. V., Braunstein, I., & Maes, P. (2025). Your brain on ChatGPT: Accumulation of cognitive debt when using an AI assistant for essay writing task. arXiv. https://doi.org/10.48550/arXiv.2506.08872
Niederhoffer, K., Rosen Kellerman, G., Lee, A., Liebscher, A., Rapuano, K., & Hancock, J. T. (2025, September 22). AI-generated workslop is destroying productivity. Harvard Business Review. https://hbr.org/2025/09/ai-generated-workslop-is-destroying-productivity
Parasuraman, R., & Manzey, D. H. (2010). Complacency and bias in human use of automation: An attentional integration. Human Factors, 52(3), 381–410. https://doi.org/10.1177/0018720810376055
Risko, E. F., & Gilbert, S. J. (2016). Cognitive offloading. Trends in Cognitive Sciences, 20(9), 676–688. https://doi.org/10.1016/j.tics.2016.07.002
Yu, S., Cheng, M., Jabbar, A., Sucholutsky, I., Collins, K. M., Jurafsky, D., & Hawkins, R. D. (2026). The efficiency-gain illusion: People underestimate the rate of AI use and overestimate its benefits on simple tasks. arXiv. https://doi.org/10.48550/arXiv.2605.22687
