Edition #32: The Middle of the Curve

AI is quietly pulling our work, our attention, and even our bodies toward average. The most human thing we can do is refuse.

By Mykel Salomon · 2026-08-08

<h3>What the neuroscience of the brain and body reveals about staying yourself in the age of AI</h3><hr><h3>🔓 Opening Reflection</h3>Here's a scene that might sound familiar.It's a little after 5am. The birds are starting up outside. In the old days, and "the old days" here means about a year ago, this was the moment to close your eyes and drift back to sleep. Instead, your eyes snap open and your first fully formed thought is: I wonder what the AI did overnight.You gave it a prompt before bed so it could work while you slept. Now you can't wait to see. You don't make coffee. You don't stretch. You sit down and you're gone, ninety minutes with the machine before the first sip, one prompt pulling you into the next, a low electric hum of this is amazing running underneath the whole thing.That's not a cautionary tale about some hollow-eyed tech addict. That's the honest self-report of <a target="_blank" rel="noopener noreferrer nofollow" class="text-[#4FD1C5] underline hover:opacity-80 ember-view" href="https://www.linkedin.com/in/awsamuel/">Alexandra Samuel, Ph.D.</a> , a Harvard-trained social scientist, a writer for the Wall Street Journal and Harvard Business Review, one of the sharpest voices thinking about AI and work. She's not describing someone who lost control. She's describing herself, noticing — in her own body, at 5am — a pull.And when we sat down for this week's conversation, what struck me most wasn't the pull itself. It was the direction of it. Because AI, it turns out, is always pulling us the same way.Toward the middle.Toward the average output. The average attention span. The average, disembodied, heads-in-a-screen version of a human being. This edition is about that pull, the real neuroscience underneath it, and about the quiet, deliberate work of pulling the other way.Let's get past the noise and into the brain and body of it.<hr><h3>🔦 Signal of the Week</h3>Start with what AI does to your work, because the science here is clean and a little unnerving.Researchers at UCL and Exeter ran a careful experiment: some writers got story ideas from an AI, some didn't. The AI-assisted stories were rated as more creative, better written, more enjoyable, especially for the less creative writers. A clear individual win.But when the researchers looked at all the AI-assisted stories together, they found something else. The stories had become more similar to each other. Individually more creative; collectively more the same. A separate study found human-written essays contributed somewhere between two and eight times more diversity of ideas than AI-generated ones, and the gap got wider the more people used it. Tweaking the settings didn't fix it.Here's the part that should reframe how you think about these tools. The same research found no evidence that AI pushed the top of the range, it never out-created the most creative humans working alone. It lifts the floor. It doesn't raise the ceiling.Why? Because of what an AI fundamentally is. Strip away the magic and a large language model is a machine for predicting the most probable next word. It is, by design and by definition, an averaging engine. Its gift and its cost are the exact same feature: it gives you the middle of the curve, beautifully.Which leads to the insight Alexandra keeps returning to, and the one I want you to carry all week:If AI lives in the middle of the curve, then your entire human value is your distance from it.<hr><h3>🌍 In the Wild</h3>Three things AI pulls toward center, and how to pull back.<h3>1. It averages your work — so find where you're off-center</h3>We talk constantly about "human skills," and Alexandra made me realize we're usually being lazy about the phrase. We use "human skills" as a polite euphemism for people skills, being warm, relating to others. But that's a tiny slice of it. And it's a slice AI is already competent at: it'll put a kind interpersonal gloss on your email in half a second. In the middle of the curve, as always.Her reframe: stop asking "what are the human skills" and start mapping your full, eclectic picture of talent across every domain, then hunt for the places where you sit furthest from the middle of the curve. The odd angle. The unusual combination. The thing you do a little strangely. That off-centeredness isn't a flaw to sand down. In an age of averaging machines, it's the only thing that can't be replaced.This is where her work gets genuinely original. Her next book applies the insights of the neurodiversity movement, decades of research on autistic and ADHD minds, to everyone. The core lesson from that world: we get into trouble the moment we treat intelligence as one thing, as a single ladder everyone climbs. There is no one way to think. And every one of us has places where the standard environment simply doesn't fit our particular brain. Those points of friction? They're also the map to your edges.AI can help you find them. Just don't let it talk you back toward the middle.<h3>2. It pulls you out of your body, and your body is where the good stuff lives</h3>This was the most important thread in the whole conversation, and it's pure neuroscience.There's a sense most people can't name: interoception, your awareness of your own internal state, the signals rising up from inside your body. It runs largely through a brain region called the insula, which sits at the crossroads of bodily sensation, emotion, and decision-making. And it matters enormously. People with sharper interoceptive awareness report richer, more intense emotional lives. It's tied to well-being. It's literally what lets you notice you're having a heart attack, or notice, in a meeting, that the room has quietly turned against you.Now: an AI is disembodied. When you spend hours in what Alexandra calls "the virtual brain space of the mind meld," you go cerebral. You dissociate from the body. The internal signals get faint.And that's a creativity problem, not just a wellness one, because creativity is an embodied practice. This isn't woo. A landmark Stanford experiment found that people walking generated roughly twice as many creative ideas as people sitting, and it was the walking itself, not the scenery. The body isn't the container the creative mind rides around in. The body is part of the engine. Alexandra knits. She does floor exercises. She takes the dog to the woods. That's not downtime from the real work. On some level, that is the work.The risk was never that AI writes the poem. It's that you spend fifty hours a week in the mind meld and slowly lose contact with the body that was going to write something only you could.<h3>3. It hijacks your attention, one more prompt</h3>Alexandra is careful with the word "addiction," and I want to honor that, because she's right: there's a real difference between compulsion, overuse, and clinical addiction, and we cheapened all three by throwing "addiction" around too early with social media. So let's be precise.But then look at the mechanism. AI responses are non-deterministic, unpredictable. You never quite know what you'll get back. Neuroscientists have a name for that: reward uncertainty. And reward uncertainty is the single most potent driver of compulsive behavior we know of, it's the engine of the slot machine, the loot box, the pull of the lever you can't stop pulling. Unpredictable rewards spike dopamine harder than predictable ones. Researchers who recently audited the interfaces of eight major AI chatbots found these dopamine patterns baked right into the design, down to the word-by-word way the text streams in, keeping you leaning forward.So "one more prompt, one more prompt" isn't a character flaw. It's variable-ratio reinforcement, the most habit-forming schedule in all of behavioral science, running on you in real time. And it shows up in the body exactly the way Alexandra described her own experience: compulsive use is now being linked in the research to anxiety, burnout, and disrupted sleep — the midnight-to-5am loop, precisely.Here's her tell, and it's a good one. She said the thing that alarms her most isn't feeling bad. It's how ecstatic she feels doing sixteen things at once. It doesn't have to feel bad to be a problem. The high is the signal.We're roughly where we were with social media around 2012, the architecture locking in, the habits forming, the research lagging the adoption by years. The difference is that this time, we've already seen the movie. We know how it ends if we're passive.<hr><h3>💬 The Big Question</h3>I went into this conversation certain the key question was: What is it that we should refuse to outsource to AI?Alexandra gently took that apart. It's a moving target, she said, the lines she swore she'd never cross last year, she's already crossed. "Never say never." So the question isn't stable enough to build on.Here's the better one I landed on:Not "what will I refuse to give AI" , but "what parts of me do I want to keep strong enough that I still could?"Your off-center skills. Your contact with your own body. Your ability to sit in a room full of people and feel the current in it. Keep those alive, and outsourcing becomes a choice. Let them atrophy, and it becomes a default you never actually decided.And underneath it all, the simplest version:Am I using this to become more myself, or less?<hr><h3>🧠 A Small Exercise</h3>Alexandra offered three habits she's hard-won through her own overuse. I've paired each with the science of why it works. Try one this week.<ol><li>Set an intention before each session. Write it on a sticky note: I am here to finish this one thing. The AI will always open the door to the next thing, this is your prefrontal cortex holding the line against the dopamine pull.</li><li>Check your body. Keep a rough mood log or a timer. After an hour with AI, ask: what's happening in my chest, my shoulders, my jaw? You're rebuilding interoception, and remember, the high counts as a signal, not just the crash.</li><li>Pack one day with people, every week or two. Meeting to meeting, lunch to walk, no room to reach for the AI. It's how you remember, in your body, what human presence actually feels like.</li></ol>And the meta-question underneath all three: Where am I furthest from the middle of the curve, and am I using AI to sharpen that edge, or quietly sand it off?<hr><h3>🎙️ This Week on The Human Protocol Podcast</h3>A More Human Workplace: Neurodiversity, the Body, and What AI Can't AverageGuest: <a target="_blank" rel="noopener noreferrer nofollow" class="text-[#4FD1C5] underline hover:opacity-80 ember-view" href="https://www.linkedin.com/in/awsamuel/">Alexandra Samuel, Ph.D.</a> — Harvard-trained social scientist, WSJ and HBR contributor, author of Work Smarter with Social Media, co-author of Remote Inc., and founder of Social SignalAlexandra has been at the bleeding edge of every wave, early online communities, social media, remote work, and now AI, and she brings something rare to this one: the discipline of a social scientist and the honesty to narrate her own struggle with the tools in real time.In this episode, you'll learn:<ul><li>Why "human skills" isn't a synonym for people skills, and how to map where you sit furthest from the middle of the curve</li><li>What the neurodiversity movement teaches all of us about working with our own brains</li><li>Why AI pulls us out of our bodies, and why interoception and embodiment matter for creativity and well-being</li><li>The real difference between compulsion, overuse, and addiction, and why she suspects AI will prove physiologically hard to put down</li><li>Three concrete habits for an intentional, empowered relationship with AI</li><li>Why "you can only be who you are", and how to make AI serve that instead of eroding it</li></ul>🎧 Available now on <a target="_self" rel="noopener noreferrer nofollow" class="text-[#4FD1C5] underline hover:opacity-80 RVDFChJswVRnGfgfLGuwQIqBPRqoGHOWSrA " href="https://open.spotify.com/episode/5o4F03dyf8LseWY4HWHtX0?si=QFRqER-xRYOeSgXZVtnWng">Spotify</a>, <a target="_self" rel="noopener noreferrer nofollow" class="text-[#4FD1C5] underline hover:opacity-80 RVDFChJswVRnGfgfLGuwQIqBPRqoGHOWSrA " href="https://podcasts.apple.com/us/podcast/s2e31-ai-makes-everyone-average-a-harvard-scientist/id1843402023?i=1000780041591">Apple Podcasts</a>, and <a target="_self" rel="noopener noreferrer nofollow" class="text-[#4FD1C5] underline hover:opacity-80 RVDFChJswVRnGfgfLGuwQIqBPRqoGHOWSrA " href="https://youtu.be/pyGkJf20ICc?si=kaIOG2BUcnSzBXIc">YouTube</a>.<img class="max-w-full h-auto rounded-md my-4" src="https://media.licdn.com/dms/image/v2/D4E12AQETUPZwREs8LA/article-inline_image-shrink_1000_1488/B4EZ_eyNhJHsAI-/0/1786149132660?e=1787788800&v=beta&t=St53loGIGHlQVhFQcyssTx12x7W2_MBZjvPUAQq_PvQ" alt="Article content"><hr><h3>🧭 Thought to End the Week</h3>Near the end, Alexandra said something that I haven't stopped thinking about.You can only be who you are. AI won't change that. It won't make you into someone miraculously different. All it can do is offer you a path, one that either supports you in becoming the most you version of yourself, or one that constrains you and fills you with so much anxiety that you never find your own path at all.Which one you get isn't decided by the technology. It's decided by the intention you bring to it.So here's where I land this week. We are being handed the most powerful averaging machine ever built, at the exact moment it's easiest to drift toward the center of everything, the average output, the average attention span, the disembodied hum of a life lived in a screen.The most human thing you can do in that current is to swim the other way. Toward your off-center edges. Back into your own body. Toward the people who remind you what presence feels like. Toward the strange, specific, irreplaceable person only you can be.The machine will always offer you the middle of the curve.Your job, your beautifully human job, is to live out at the edges of it.The code may change. And it will.But our humanity is the constant.See you next week.— Mykel<hr><h3>📚 Sources</h3><ol><li>Doshi, A. R., & Hauser, O. P. (2024), "Generative AI enhances individual creativity but reduces the collective diversity of novel content," Science Advances 10(28) — AI-assisted stories rated more creative but more similar to one another; no evidence AI raises the top of the creative range. <a target="_blank" rel="noopener noreferrer nofollow" class="text-[#4FD1C5] underline hover:opacity-80" href="https://www.science.org/doi/10.1126/sciadv.adn5290">https://www.science.org/doi/10.1126/sciadv.adn5290</a></li><li>Moon, Green & Kushlev (2024–25), "Homogenizing effect of large language models on creative diversity," ScienceDirect — human-written essays contributed ~2–8x more collective diversity than GPT; gap widens at scale. <a target="_blank" rel="noopener noreferrer nofollow" class="text-[#4FD1C5] underline hover:opacity-80" href="https://www.sciencedirect.com/science/article/pii/S294988212500091X">https://www.sciencedirect.com/science/article/pii/S294988212500091X</a></li><li>"The Insular Cortex: An Interface Between Sensation, Emotion and Cognition," Neuroscience Bulletin (2024) — the insula as the hub integrating interoception, emotion, and decision-making; links to anxiety, depression, and addiction. <a target="_blank" rel="noopener noreferrer nofollow" class="text-[#4FD1C5] underline hover:opacity-80" href="https://pmc.ncbi.nlm.nih.gov/articles/PMC11607240/">https://pmc.ncbi.nlm.nih.gov/articles/PMC11607240/</a></li><li>Interoception and emotion, Social Cognitive and Affective Neuroscience (Critchley) — heightened interoceptive awareness is associated with more intense emotional experience and wellbeing. <a target="_blank" rel="noopener noreferrer nofollow" class="text-[#4FD1C5] underline hover:opacity-80" href="https://academic.oup.com/scan/article/8/3/231/1726170">https://academic.oup.com/scan/article/8/3/231/1726170</a></li><li>Oppezzo, M., & Schwartz, D. L. (2014), "Give Your Ideas Some Legs: The Positive Effect of Walking on Creative Thinking," Journal of Experimental Psychology: LMC — walking roughly doubled divergent creative ideation; the effect came from the movement itself. <a target="_blank" rel="noopener noreferrer nofollow" class="text-[#4FD1C5] underline hover:opacity-80" href="https://www.apa.org/pubs/journals/releases/xlm-a0036577.pdf">https://www.apa.org/pubs/journals/releases/xlm-a0036577.pdf</a></li><li>"The Dark Addiction Patterns of Current AI Chatbot Interfaces," CHI 2025 — dopamine-driven design patterns mapped across eight major chatbots. <a target="_blank" rel="noopener noreferrer nofollow" class="text-[#4FD1C5] underline hover:opacity-80" href="https://dl.acm.org/doi/10.1145/3706599.3720003">https://dl.acm.org/doi/10.1145/3706599.3720003</a></li><li>"What Research Says About AI Chatbots and Addiction," Tech Policy Press (2025) — non-deterministic AI responses create "reward uncertainty" that elevates dopamine like a slot machine. <a target="_blank" rel="noopener noreferrer nofollow" class="text-[#4FD1C5] underline hover:opacity-80" href="https://www.techpolicy.press/ai-chatbots-and-addiction-what-does-the-research-say/">https://www.techpolicy.press/ai-chatbots-and-addiction-what-does-the-research-say/</a></li><li>"Problematic use of generative AI chatbots," Frontiers in Public Health (2026) — the clinical distinction between compulsion, overuse, and addiction; withdrawal-like responses in heavy users. <a target="_blank" rel="noopener noreferrer nofollow" class="text-[#4FD1C5] underline hover:opacity-80" href="https://www.frontiersin.org/journals/public-health/articles/10.3389/fpubh.2026.1816917/full">https://www.frontiersin.org/journals/public-health/articles/10.3389/fpubh.2026.1816917/full</a></li><li>Duong et al. (2024), Acta Psychologica — compulsive ChatGPT usage linked to anxiety, burnout, and sleep disturbance. <a target="_blank" rel="noopener noreferrer nofollow" class="text-[#4FD1C5] underline hover:opacity-80" href="https://doi.org/10.1016/j.actpsy.2024.104622">https://doi.org/10.1016/j.actpsy.2024.104622</a></li></ol>