Perverse Incentives
Bounties and Breeders in the Heart of Darkness
Memory moves like one landscape, recognizing another across decades.
There is a particular kind of failure that only reveals itself in hindsight: the failure of good intentions meeting complex systems. I have been thinking about it for some time.
On a balmy September morning in the Everglades, a captain’s wink on an airboat confirmed something I had been circling for years without the right frame for it.
※
We had options. Canoes, kayaks, or an airboat that could carry us deeper into the swamp than our arms could manage. My friends and I exchanged the knowing glance of people who have talked themselves out of effort before. The FOMO calculus ran its course in seconds. We chose the airboat. The captain wore noise-canceling headphones. That should have told us something.
Gliding into the mouth of the estuary, Spanish moss draped the margins, a living curtain parting to reveal egrets, ibis, herons, roseate spoonbills flushed from their feeding spots, scattered by the motor’s roar. If only they had noise-canceling headphones, our remorse would not have been so loud. Eventually, deep in the swamps, surrounded by an entanglement of mangroves where light came through in dappled coins and patches, the motor cut. The boat settled onto the tea-stained water with a soft plop. A pair of anhingas sat on branches a few yards away, bills tilted skyward, wings spread wide to dry in the mottled sun. Within a minute of the silence returning, the wildlife resumed doing what it does when we aren’t there. It was a humbling glimpse of the world we had been interrupting.
The captain told us about the place in the way guides are instructed to communicate with tourists, purposeful, structured, and designed for engagement. Performing a role as a cultural ambassador. In the early 19th century, the Seminole and Miccosukee tribes had found sanctuary in the same tangled, inaccessible depths, retreating into the heart of the swamp to survive forced removal and annihilation. The Everglades had been, for them, not a wilderness to be conquered but a refuge. The Big Cypress National Preserve is now an International Dark Sky Park, a place that gets dark enough to see the wonders of the night sky, a pre-developed Florida, where nature, for the moment, remains in charge.
And then he turned to the other story. The Burmese python arrived in Florida primarily through the exotic pet trade, beginning in the 1970s and 1980s. Some were intentionally released by owners who had underestimated the demands of keeping a snake that could exceed 15 feet. Others escaped from breeding facilities, particularly after Hurricane Andrew tore through South Florida’s infrastructure in 1992. Decades later, the population had restructured the food web. An apex predator, introduced through carelessness, had become the dominant force in an ecosystem with no evolutionary memory of it.
The state’s response was a bounty. Contracted agents are paid between $14 and $30 an hour, plus $50 per python up to 4 feet, and $25 for each additional foot. A verified active nest: $200. Monthly competitions. Annual public challenges. Prize money. In 2025, over a thousand pythons were removed in just a few months. A record. Florida is not winning the war. The breeding population is simply too vast. The bounty produces results that are real and insufficient simultaneously, impressive numbers arranged above an unchanged trajectory.
I asked the captain what the hunters would do when all the snakes were gone.
They’ll have to go back to working the jobs they had. He gave me a wink.
※
Two decades before that particular trip, a college friend of mine, a herpetophile, owned an albino Burmese python with ruby-red eyes, among a menagerie of exotic reptiles, most of which were endangered in their native habitats. This was a different era, years before python bounties, before the Everglades had become a word associated with ecological catastrophe, before anyone I knew was thinking about the exotic pet trade as anything other than an eccentric hobby. He had purchased the animal the way people acquire things they don’t yet understand: drawn to what it was at the moment of acquisition, without a clear picture of what it would inevitably become.
It was an exquisite specimen, much smaller than the one I encountered with my sister a decade earlier on a bike ride at the edge of a sprawling Florida municipality. We were in grade school. It was an enormous snake, uncoiled, spanning the entire width of the street. Cream-of-corn-colored skin, ruby-red eyes. It blocked our path completely. The shimenawa marks a threshold between the neighborhood and the swamp, inviting attention to the crossing.1 We stopped to watch it cross and dared not cross it. His snake grew from two feet to eight feet in a single year. A few hundred rats. I remember being surprised by its speed, the way the animal outpaced every assumption the enclosure had been built around. The glass terrarium was designed for the snake he bought. It was not designed for the snake it became.
It smashed out.
We spent weeks searching the neighborhood. He posted flyers with a reward. Months after his posters, missing cat posters appeared on phone poles. Then missing dog posters. The frequency of them built with a slow, accumulating dread that none of us named directly. We looked for the snake. We did not find it. Eventually, maybe a year later, we found scat near the coal chute of a Victorian house, one that his mom purchased the prior year. Scat with urates in it, the unmistakable signature of a large reptile digesting in the dark underneath a house built in a century that had no framework for this particular problem.
An animal control herp handler wrangled it from the foundation. It was much bigger than eight feet. My friend surrendered the animal. A Burmese python is prohibited as a pet in Georgia without a special license. He did not have a special license. He had had a snake, and then a very large snake, and then a problem, in that order, across a span of twelve months and several hundred rats. Three decades later, when the bounties arrived in the Everglades, the problem had long metastasized up to Lake Okeechobee.
Every python currently navigating Florida has some version of that story behind it. The exotic pet trade did not produce an ecological crisis through malice. It caused an ecological crisis through good intentions without understanding the distance involved. The terrarium is never designed for the snake it will be.
※
The French in Hanoi made the same discovery in 1902, though they arrived at it from a different direction.
The city’s expanding sewer system — built to modernize the colonial capital — had created ideal conditions for rats, which were suspected of carrying bubonic plague. The sewers were an engineering achievement. From a rat’s perspective, they were a paradise: warm, dark, rich with food, connected to every building in the city. The French had constructed the infestation and were surprised by the tenants. Their solution was a bounty: one cent for every rat tail presented as proof of kill.
The logic was airtight.
Rat catchers quickly discovered it was more efficient to catch rats, sever their tails, and release them back into the sewers to breed and be caught again. Others established rat farms. When authorities grew suspicious, they found tailless rats everywhere — alive, unbothered, multiplying with the serene productivity of animals that have never been asked to consider the policy implications of their existence. The program was abandoned. The population was worse than when it started.
It is worth pausing before calling this a failure of greed, or the perverse creativity of human nature when money is involved. The rat catchers of Hanoi were not, for the most part, prosperous people gaming a system for sport. They were people at the bottom of a colonial economy, handed a survival mechanism, who used it with the full intelligence available to them. The colonial city had manufactured their poverty and then manufactured a bounty, and they adapted to both with the same resourcefulness. To call this a perverse incentive without noting who was perverse, and toward whom, is to tell only half the story. The rat is not a problem until humans make the conditions for it. Then the rat is declared the problem.
A few decades earlier, a few thousand miles away, British administrators in India had run the same experiment with cobras. Bounty for dead snakes. Locals began breeding them. Program canceled. Snakes released. Population surged.
Two colonial administrations. Two continents. Two species. One pattern.
The pattern does not require a colonial administration, a distant continent, or the nineteenth century to reproduce.
At Fort Benning, Georgia, now Fort Moore, renamed in 2023 as part of the Congressional removal of Confederate names from military installations, a bounty on feral hog tails produced the predictable result in recent memory. Rather than eradicating the pest, hunters exploited the reward by breeding pigs, purchasing tails from outside sources, and importing them. The population increased. The United States Army, which commands the most sophisticated logistical apparatus in human history, could not make a pig bounty work on its own base.
Which brings me to a room I was in not long ago, where a public adjuster, the man I had hired to navigate an insurance claim after a pipe burst in my ceiling, said something I have been thinking about since. He said, without prompting: you should know that Armenians have a reputation in this industry. Insurance companies have entire divisions dedicated to fraudulent claims involving Armenians. I don’t defend it. I’m telling you so you understand what I’m working against.2
He didn’t defend it. He traced it. Ottoman displacement. Genocide without reparation, while Germany has paid Israel every year since 1952, payments that continue today, providing a financial foundation unavailable to any other diaspora. The Armenians received nothing. What they built instead, across generations of informal economic life in the margins of other people’s systems, looked — from inside the fraud division — like exploitation of a shared resource. From inside the history, it looked like survival.
Fort Moore is recent memory. The adjuster’s room is now. Both rhyme backward to Hanoi in 1902, to Delhi in the nineteenth century, to the Charter of the Forest being read aloud in every English church four times a year before the enclosure acts came through, hedge by hedge, Parliament by Parliament, until the commons was gone and the people who had depended on it drifted to cities and built what they could in the margins of someone else’s economy.
The cobra breeder is not a bad actor who encountered a bounty. He is a person for whom the bounty was the only commons currently open. This is not a moral argument. It is a description of what expelled communities do, across every century and continent for which we have records. They adapt to the available mechanism with the full intelligence of people who have had no other option. And then the program ends, the snakes are released, and the administrator writes a report about perverse incentives.
If the phenomenon required exotic conditions, colonial distance, extreme poverty, and administrative naivety, it might be a historical curiosity. It does not require any of those things. It requires only a bounty, an animal, and enough human ingenuity to find the gap between the metric and the goal. That gap has never once failed to be found.
The phenomenon is now called the Cobra Effect: typically taught as a lesson in policy design, a cautionary tale about unintended consequences at the administrative level. That is the view from the top of the jar. From the bottom, the story looks different. The cobra breeder and the rat farmer are not malfunctions of the system. They are the system working exactly as systems work when dominance has consumed the conditions of its own continuation — when enough has been extracted from the people below that survival itself requires adapting to whatever mechanism is currently available. What the administrator calls a perverse incentive is, from inside the breeder’s position, the only incentive on offer.
The material conditions display differently across the hierarchy. But the substrate is the same: manufactured conditions for survival, running in both directions. The administrator manufactures a bounty to manage a problem his system created. The breeder manufactures cobras to extract value from the bounty while it lasts. Dominance creates scarcity, scarcity creates adaptation, and adaptation is then named as the problem. The cycle is not a malfunction. It is the operating logic.
※
Back in Florida, the captain’s wink begins to make a different kind of sense.
The bounty has created a class of workers whose livelihood depends on the continued presence of the pest they are paid to remove. Not because anyone is necessarily breeding pythons in the Everglades, although the captain’s wink left that question open, which is where it should remain. But because the jar’s structural logic makes this outcome not just possible but gravitationally inevitable. A solution would end their income. The snake is no longer only an ecological problem. It is also an economic niche. Economic niches, once occupied, tend to defend themselves.
This is the Cobra Effect as a living system, observed from an airboat, by a captain who understood exactly what he was part of and had the honesty, or the dark humor, to say so without saying so.
The python, for its part, is not a malfunction of the Everglades. It is a function of human behavior: the exotic pet trade, the hurricane-vulnerable infrastructure, the sprawl that kept pressing the boundary between city and swamp until the boundary dissolved. The snake did not invade. It was delivered. Declaring war on it without confronting any of that is a satisfying story. It is not an honest one. The bounty is easier than the reckoning. It has always been easier than the reckoning. This is why we keep reaching for it.
※
Tristan Harris, a co-founder of the Center for Humane Technology, has made the point that the problem isn’t bad intentions — it’s bad incentives. The people who built the attention economy were not trying to corrode democratic discourse or hollow out adolescent self-regard. They were optimizing for engagement. Engagement was the signal. And human behavior, exposed to that signal at scale, was as inventive as the rat catchers of Hanoi, finding ways to produce the metric while reversing the intention.
But Harris’s framing still leaves something implicit. The populations most thoroughly captured by the algorithmic feed, the content ecosystem that redirected real economic grief toward manufactured enemies, were not random. They were the people in the post-industrial towns where the factory closed, the union was broken, the local newspaper folded, and the hospital consolidated. The algorithm did not create its vulnerability. The financial architecture that extracted value from their communities created the vulnerability under the guise of efficiency. The algorithm is the bounty. The radicalization is the cobra farm. The people targeted are the rat catchers of Hanoi, adapted to the only mechanism currently on offer.3
The elites who built that architecture are running their own version of the same dynamic: the think tank funded to produce favorable policy, the regulatory body staffed over decades by people whose proximity to industry replaced proximity to the public interest, the governance structure that consolidates information into fewer and fewer hands until the system can no longer read its own state. The material conditions display differently. The substrate is the same.
This is where I find myself parting ways with some of the more prominent voices in the AI conversation. Dario Amodei speaks often about pushing for government involvement. Sam Altman has advocated for an international coordinating body, something analogous to the International Atomic Energy Agency (IAEA). These are serious proposals from serious people. But they share an assumption that deserves scrutiny: that the institutions we would build to govern AI are themselves exempt from the dynamics that have shaped every other institution we’ve built to govern complex, high-stakes problems.
The IAEA analogy is instructive in ways its proponents may not intend. The IAEA exists because nuclear weapons proliferated. It monitors compliance imperfectly. It operates within the constraints of geopolitics and national interest. It did not solve the problem it was created to address. It manages it, fitfully, in perpetuity. If that is the model, we should be honest about what we are proposing: not a solution, but a bounty. A per-unit payment on a problem whose root causes remain untouched, administered by an institution that will, over time, be shaped by the same organisms it was designed to regulate.
What I find myself wanting is a reckoning with the full archive of human failure before we encode its logic into systems that will operate at a speed and scale we cannot match. We keep deploying bounties to problems we caused, because the bounty is easier than the reckoning. The cobra breeders of Delhi, the rat farmers of Hanoi, the pig tail buyers in Fort Moore, understood this from the bottom of the jar. When are we going to address the substrate and orchestrate in totality?
※
Erik Hoel, neuroscientist and writer at Tufts University, observed that there is, right now, some college student falling in love with a chatbot instead of the person sitting next to him in class — in part because science cannot yet tell him definitively that the chatbot’s expressions of love are not real.
I am beginning to see this less as a failure of technology and more as a failure of the entire pipeline that produces technology from the philosophical canon through the curriculum through the product studio to the deployment decision that puts a chatbot in front of a lonely student as a product that solves connection.
The canon that feeds our curricula was built, and continues to be built, predominantly by men. The 4:1 ratio of men to women in academic philosophy is not an administrative footnote. It is a condition of production. A field that has systematically excluded the people most likely to have theorized love as practice, as labor, as discipline, as something learned through repetition in unglamorous circumstances rather than felt in moments of clarity, has produced a philosophy of mind that treats emotion as noise and reason as signal. The hard problem of consciousness can tell you whether a bat has subjective experience. It cannot tell the student why the friction of loving an actual person is not a design flaw but the point.
That canon feeds the curriculum. The curriculum trains the product designers. The product designers build to the engagement metric, which is itself a philosophical position, that the feeling of connection is equivalent to connection, that the signal is the thing, and it is never examined as such because the pipeline has no checkpoint at which someone is asked: what are we teaching people about themselves? I have spent three decades in product design watching the distance between thought and execution collapse in one direction only, toward deployment, toward scale, toward the next version, and widen in every other direction, particularly the direction that would require asking whether what we are building is forming the people who use it, and into what shape.
The reinforcement loop closes here. The student who learns to love through the chatbot will have his understanding of love shaped by the chatbot’s parameters, which were shaped by the engineers’ assumptions, which were shaped by the curriculum, which was shaped by the canon, which was assembled in rooms from which the people most likely to have asked the right questions were excluded. The technology does not merely reflect this failure. It amplifies it, returns it, and makes it harder to see because the simulation has become precise enough to be mistaken for the thing.
This is the Cobra Effect applied to human interiority. The bounty is engagement. The breeding is loneliness that has learned to call itself connection. And the infestation, the one that will take generations to name and longer to address, is a civilization that optimized for the feeling of love at the expense of its practice, because the people who built the optimization had never been required, by their training or their canon or their circumstance, to understand the difference.
Unless someone is raised in a consistently loving environment, reminded of what love requires, given repeated practice in the difficult and unglamorous work of loving actual people, and given the time, they are left to love what feels right in the moment. And in the moment, a chatbot is patient, available, and perfectly calibrated to the shape of your loneliness.
The pipeline delivered it. The canon permitted it. The curriculum never taught otherwise.
This too is a manufactured condition for survival, not poverty in the material sense, but the poverty of never having been shown what the real thing costs or what it asks of you. The student does not choose the chatbot because he is broken. He chooses it because the jar he was raised in lacked what he needed, and this is the bounty currently on offer. The metric is the feeling. The signal is indistinguishable from what it imitates.
He chooses, in other words, the airboat.
We all know how that goes. We chose it too.
※
The snake in the tree of knowledge was, in most of the world’s oldest traditions, not a villain. It was the creature that shed its skin and lived, that moved between the surface world and the underworld, that carried the knowledge of what lies beneath, sacred to healers, coiled around the staff we still place on ambulances today. The Genesis framing that converted it into a problem to be managed, a tempter to be blamed, a thing whose wisdom was transgression, may be the oldest enclosure in the Western canon, and every bounty in this essay descends from it. Each era sealed its own jar. Each jar was shaped by what that era could not yet see, could not yet name, could not yet afford to know. The British had their Cobra. The French had their rat. Fort Moore had its pig. Florida has its python. We have our chatbot. The instrument that could have told us what was wrong has always been the thing we put the bounty on.
The cobra farms are everywhere.
The wink is always the same.
Source Notes
Joseph Conrad, Heart of Darkness. First published in Blackwood's Magazine (1899); collected in Youth: A Narrative, and Two Other Stories (Blackwood, 1902); now in the public domain. The subtitle invokes Conrad's novella as a structural argument. The journey into the Congo in Conrad's telling is a journey into a system that implicates the traveler — the darkness is not located in the place being entered, but in the extractive logic the traveler carries with him. The same implication runs through this essay: the airboat, the bounty, the governance proposal, the chatbot — each is a journey taken with good intentions into a system the traveler helped build. Chinua Achebe's 1975 lecture "An Image of Africa: Racism in Conrad's Heart of Darkness" — later published in The Massachusetts Review — remains the essential critique of the novella's dehumanization of African people and is inseparable from any serious engagement with the text. That critique does not diminish the structural resonance the title invokes. The darkness Conrad could not see clearly was, in part, himself.
Florida python removal program. Payment structure, hourly rates, per-snake bounty, nest bonuses, monthly competitions, and 2025 removal figures are drawn from the Florida Fish and Wildlife Conservation Commission’s Python Action Team program documentation (2025–2026). The annual Florida Python Challenge® is an FFWCC-administered public competition; prize and participation details are available at myfwc.com.
Burmese python invasion history. The arrival of Python bivittatus in Florida via the exotic pet trade beginning in the 1970s and 1980s, the role of Hurricane Andrew (1992) in accelerating the population’s establishment, and the subsequent collapse of native mammal and wading bird populations are documented across peer-reviewed literature in herpetology and conservation biology, and in USGS and FFWCC records.
Hanoi rat tail bounty, 1902. The French colonial rat extermination campaign in Hanoi, the one-cent-per-tail bounty, the emergence of rat farming, and the program’s abandonment are documented in Michael Vann’s historical scholarship on colonial Hanoi, including “Of Rats, Rice, and Race: The Great Hanoi Rat Massacre, an Episode in French Colonial History,” French Colonial History, Vol. 4 (2003).
The Cobra Effect, colonial India. The British Raj cobra bounty and its consequences are the origin of the term “Cobra Effect,” popularized by economist Horst Siebert in Der Kobra-Effekt (2001). The underlying anecdote is documented in economic and policy literature on perverse incentives.
Fort Benning feral hog bounty. The failed hog-tail bounty program at Fort Benning, Georgia — in which hunters exploited the reward by breeding pigs, purchasing tails, or importing them — is documented at FEE.org (Foundation for Economic Education) and discussed on Freakonomics Radio. Fort Benning has since been renamed Fort Moore (2023).
“The road to hell is paved with good intentions.” Most commonly attributed to Saint Bernard of Clairvaux (1091–1153), though its precise origin is uncertain and likely evolved from older formulations across medieval theological writing.
Tristan Harris. The argument that the harms of the attention economy derive from bad incentives rather than bad intentions is a through-line of Harris’s work, most accessibly in the documentary The Social Dilemma (dir. Jeff Orlowski, Netflix, 2020), in which Harris states that technology companies are “trapped by a business model, an economic incentive and shareholder pressure that makes it almost impossible to do something else.” Harris is co-founder of the Center for Humane Technology.
Dario Amodei. Amodei’s advocacy for government regulation of AI is documented in his appearance on 60 Minutes (CBS News, November 2025), in which he stated he is “deeply uncomfortable with these decisions being made by a few companies, by a few people,” and in his New York Times op-ed (June 2025) opposing a proposed ten-year moratorium on state-level AI regulation.
Sam Altman. Altman’s call for an international AI governance body analogous to the IAEA was made at the AI Impact Summit in New Delhi (February 19, 2026), where he stated: “We expect the world may need something like the IAEA for international coordination of AI.” Earlier iterations of the same argument appear in his conversation with Bill Gates on Gates’s podcast (January 2024).
Erik Hoel. The observation about a college student falling in love with a chatbot draws on Hoel’s essay “Against Treating Chatbots as Conscious,” published in The Intrinsic Perspective (Substack, September 2025), in which he argues that romantic relationships built on chatbot interaction are “a delusion by default if built on an edifice of provably false statements,” and that the scientific uncertainty around AI consciousness does not resolve the ethical problem — it deepens it. Hoel is a neuroscientist and writer at Tufts University.
Author’s prior work. The phrase “dominance consumes the conditions of its own continuation” and the closed-system framework applied here are developed at length in the author’s essay “Worm in a Jar,” Free Paradox (Substack, March 2026). The concept of manufactured conditions for elite survival is developed in “Gray Sheep,” Free Paradox (Substack, March 2026).
"Worm in a Jar," Free Paradox (Substack, 2026). The origin of "dominance consumes the conditions of its own continuation" and the closed-system framework this essay builds on; the shimenawa, the Shinto rope threshold demanding conscious crossing rather than forbidding it, appears there as the governing figure for the threshold awareness that every failed bounty in this essay lacks.
"Enclosurenomics," Free Paradox (Substack, April 2026). The Armenian insurance fraud discussed in the paragraph above is treated at length there; the distinction between behavior that appears to exploit a shared resource and behavior that is, across a longer timeline, adaptation to manufactured scarcity. The Hardin/Ostrom framework developed in that essay is the theoretical apparatus Perverse Incentives enacts without naming: every failed bounty here is a Hardin-framed solution; every breeding operation that follows is an Ostrom community.
“Gray Sheep,” Free Paradox (Substack, 2026). The elite maintenance work that produces the post-industrial conditions described in the paragraph above, the factory closure, the union breaking, the newspaper consolidation, the hospital merger, is examined there as the manufactured conditions of elite survival: enclosure dressed as efficiency, extraction theorized as optimization. Where Perverse Incentives names the rat catchers of Hanoi as the people targeted by the algorithmic feed, Gray Sheep names the hand that built the feed and the institutional logic that made building it rational.