Imagine a wasp. She stings a cricket, drags it to her burrow, drops it at the entrance, then goes inside to inspect the nest. She comes back out, grabs the cricket, and pulls it in. This is her script—performed flawlessly for millions of years.
But here’s the catch: if you move the cricket a few inches away while she’s inside, she doesn’t adapt. She repeats the entire sequence. Sting, drag, drop, inspect, retrieve. Over and over. She can’t break the script. It’s called sphexishness—behavior that looks intelligent and goal-directed, but is actually brittle, deterministic, and robotic the moment the environment breaks the expected pattern.
Now, look at yourself. Look at the AI systems we worship. Most of what we call ‘agency’ is just sphexishness that hasn’t been exposed yet. You drive the same route to work, order the same coffee, reply with the same canned phrases. When your usual path is blocked, do you adapt instantly, or do you fumble? When a chatbot meets a question outside its training distribution, it doesn’t think—it loops through its script, often with absurd results.
We celebrate complexity as intelligence. But complexity is cheap. A wasp’s nest-building is staggeringly complex. A large language model can write poetry. Neither can handle a disruption they weren’t trained for. True agency isn’t complex behavior—it’s the ability to break the script when the world throws you a curveball.
This is uncomfortable because it challenges our deepest assumptions about ourselves. We want to believe our choices are free, our thoughts original. But how many of your daily decisions are actually scripted? How much of what you call ‘thinking’ is just pattern-matching under normal conditions? The sphex wasp and the human mind differ in degree, not in kind—until you push them off-script.
This matters now more than ever because we are building AI that mimics our sphexishness. We celebrate its fluency, its ability to follow complex scripts. But when the environment changes—a novel problem, a contradictory instruction, a broken assumption—it doesn’t rethink. It repeats. We call it ‘intelligence’ only because the environment hasn’t yet shown us its script’s limits.
So the next time you feel proud of your ‘smart’ habit, or impressed by a chatbot’s eloquence, ask yourself: Is this agency, or just a very elaborate wasp dance? The scariest part isn’t that we might be bugs. It’s that we might be completely unaware of the scripts we’re running—until someone moves the cricket.
FAQ
Q: Does this mean humans have no free will?
A: No—it means free will is rarer than we think. Real agency requires the capacity to detect when your script is failing and to construct a new response on the fly. Most of the time, we don't exercise that capacity. We run on autopilot until something breaks.
Q: How does this change how we should build AI?
A: We should stop measuring AI by how well it follows complex scripts and start measuring it by how gracefully it handles script failures. A truly intelligent system needs a meta-layer that recognizes when its patterns don't apply and can pause, reflect, and improvise. That's the frontier.
Q: Isn't sphexishness just a survival strategy? Why is it bad?
A: It's fine for wasps because their environment is stable. But our world changes fast. Sphexishness becomes dangerous when it's invisible—when you believe you're making choices while actually running a brittle script. That's how organizations fail, markets crash, and people stay stuck in dead-end patterns.