A thought experiment about what happens when an allergy jumps the species barrier.
In the woods of New England, a small tick is rewriting an old ecological story. When the lone star tick bites a person, it can leave behind an allergy to mammalian meat — a condition called alpha-gal syndrome. The mechanism is strange and specific: a sugar found in non-primate mammal tissue, transferred through tick saliva, that the immune system learns to attack. A person who develops it can no longer safely eat venison, pork, or beef.
Here is the thought experiment this essay rests on. What if alpha-gal — or something like it — could jump the species barrier into the wild carnivores that hunt deer? Coyotes, bobcats, fishers, and bears are all parasitized by the same lone star ticks that are now spreading north into New England.1 The conventional answer is that these species produce alpha-gal themselves and so are tolerant to it: tick bites can’t induce an allergy to a molecule the immune system already counts as self. That answer is the textbook story, but it isn’t airtight — we’ll come back to why in the closing section. For now, imagine that a tolerance-breaking variant emerges. A coyote bitten by enough ticks loses the ability to digest deer. The deer survives the encounter. The coyote starves.
You might expect this to be good news for the deer. This essay is about why it usually isn’t — and how to think clearly about the difference between “fewer animals dying violently” and “less animal suffering.”
To get there we’ll build up a small mathematical model in pieces. Each section adds one idea. You can pull the sliders and the math will pull right along with you. At the end there’s a master simulation that puts the whole picture on one chart, and lets you set your own ethical weights to ask: did removing the predator help, or did it just move the suffering somewhere harder to see?
Every chart in this essay uses the same color for the same idea. The actors:
Start with a single deer population and nothing to stop it. The simplest model says: more deer make more deer, so the rate of change is proportional to how many you already have.
That little equation says “the speed at which 🦌 deer appear is some growth rate α times the current number of deer.” Pull the slider for α below. Higher growth rate, faster the curve climbs.
Deer eat plants. Plants are finite. When deer density gets high, food runs short, fawns starve, fewer fawns are born. The cleanest way to add this to our model is the logistic term — a brake that gets stronger as the population approaches a ceiling 🌿 K.
Now the curve S-shapes its way to a stable level. Try lowering K and watch the ceiling drop with it.
Now add the predators that hunt deer. They eat well when deer are plentiful, and starve when deer are scarce. The classical way to write this, due to Lotka and Volterra in the 1920s, looks like this:
The four Greek letters do simple work. α is how fast deer reproduce. β is how good the predators are at catching deer. δ is how much of each deer becomes more predator (calories converted to new coyote pups, new bobcat kits, surviving another winter). γ is how fast predators die off when deer get scarce.
Notice we’ve quietly dropped the 🌿 K term from the prey equation for now. The classical Lotka-Volterra model assumes predation alone keeps deer in check — the land’s ceiling never comes into play. The result is the perfect oscillation that has fascinated ecologists for a century: predators rise, prey falls, predators starve, prey recovers, repeat. Forever. Try the sliders.
Here is where the model gets interesting. As the tick bloom rolls through, a critical mass of coyotes, bobcats, and fishers acquire the alpha-gal sensitization from accumulated bites. The biology isn’t instant — alpha-gal disables the predator’s ability to digest its main prey, but doesn’t kill it directly. So we track a third quantity, E(t), the predator effectiveness:
Read the first line as: the bloom z(t) eats away at the predator’s usable hunting capacity, and once gone it doesn’t come back. The second line is the predator equation from before — except predation rate now scales with E. Once E approaches zero, the predator can’t feed effectively and slowly starves out via background mortality γ. No instant disappearance — a gradual collapse over a few years.
The bloom also hurts the deer directly. Real lone star tick infestations cause anemia, secondary infections, and stress in deer too — so the model includes a temporary toll −τ·z(t)·x on the prey while the bloom is at its peak.3 The deer dip but recover; the predators fade and don’t. In this section, assume nothing else changes — no density-dependent disease yet. Deer eventually drift up toward K and settle.
Drag the orange 🦠 marker along the time axis to slide the bloom around, or widen the bloom to make the irruption more drawn-out.
If this were the whole story, alpha-gal would be a gift to the deer. It isn’t. The next section is why.
Two things missing from the naive picture. First, the carrying capacity K in the logistic equation describes space — how many deer fit — but a healthier carrying capacity, call it Ksafe, is much lower. Above it, animals get crowded enough that disease spreads and parasites flourish.
Second, the mortality from crowding is sharply nonlinear. A herd at twice its healthy density doesn’t have twice the disease — it has many times more, because contact rates rise faster than population.
Where the new mortality terms kick in nonlinearly when the herd is over its healthy density:
The result, in the model, is not a brief catastrophe but a slow one. The herd grows past Ksafe and settles at a new, higher level — one where disease and starvation are now constant pressures, not occasional events. Total mortality is comparable to before. Its character is not. Move the orange marker and watch the disease and starvation lines stay elevated forever after.
(In real ecosystems, this elevated equilibrium often does punctuate with crashes — overwintering die-offs, disease outbreaks, density-dependent waves. A toy model with two state variables can’t capture those delays. The story it can tell, and tells clearly, is the one above: removing a predator doesn’t remove the pressure on its prey.)
So far we’ve been counting animals. But we usually care about animals because we care about what they experience. A deer killed in under a second by a hunter’s clean rifle shot has had a different end than a deer chased through deep snow by coyotes and slowly torn apart, or one that spent its last winter coughing up parasites. The math doesn’t make these distinctions for us — we have to put them in by hand.
One way is a weighted suffering metric. The components are different kinds of dying; the weights are what each kind costs:
Each component is something the model can compute from the populations. The weights w are where ethics enters. w₁ rates a wolf attack, w₅ rates a clean rifle shot — the same dead deer, very different last minutes. Different preset combinations encode different worldviews. The "Hunter Compensation" preset, for example, says hunting deaths are much less bad than predation deaths and asks whether ramping hunting up after the predator collapse can keep total suffering down.
The readout under the chart shows mean suffering per year in each era — not an integral, since the integral grows arbitrarily with how long you run the simulation. The per-year rate is the honest comparison: was a typical year before the tick better or worse than a typical year after it settled?
Here’s every line and every dial in one place. Drag the 🦠 marker to move the tick event. Toggle any line on or off. Pick an ethical preset or build your own. The readout shows mean suffering rates per year — independent of how long you let the simulation run.
Three sliders appear here for the first time. 🦠🐺 alpha-gal μ sets how thoroughly the bloom destroys predator hunting effectiveness — at zero the predators carry on unaffected. 🦠🦌 deer toll τ controls how badly the bloom hurts deer directly through tick parasitism. 🏹 hunting rate h introduces human harvest as a clean, fast alternative to predation; the suffering weight w₅ sets how bad you think a hunter’s shot is compared to a coyote’s teeth.
The question this essay was built to ask: can you find a setting where removing the predator reduces total suffering per year? Try.
This is a toy model, and the scenario it depicts is speculative. But the standard reason to dismiss it outright — that non-primate mammals produce alpha-gal as a self-molecule and so can’t be sensitized to it — is shakier than it first appears.
It’s mostly right: tolerance to self does mean that, in the typical case, tick bites can’t induce an allergy to a molecule the immune system already counts as its own. But that tolerance isn’t indestructible. The same tick saliva that sensitizes humans is loaded with immunomodulators, and there’s now evidence in dogs — published in 2019 — that tick bites do induce anti-alpha-gal antibodies in animals that should, in theory, be tolerant.2 Whether that breakthrough antibody response is enough to cause a functional meat allergy in a wild canid is unknown. It’s also genuinely possible that breaking tolerance to a self-molecule is worse than developing a foreign-antigen allergy — that’s the territory of autoimmune disease, not just food intolerance. The thought experiment in this essay doesn’t depend on which version is right. It only needs the predator to vanish, by some plausible mechanism.
And the model itself is a stand-in for any pressure that takes an apex predator out of a system: a novel pathogen, a habitat collapse, a hunting policy change, a sudden behavioral shift. The math doesn’t care about the cause; it cares about the absence.
Real ecosystems also have more than two species, more than one limiting resource, weather, geography, and the unpredictable. Real predator collapses don’t happen all at once. Deer aren’t the only thing that fluctuates.
But toy models do something the full picture can’t: they let you see the shape of an argument clearly. The shape here is that removing a predator does not remove the pressure on its prey. It just changes how that pressure gets paid. The currency shifts from fast deaths to slow ones, from predation to disease, from an autumn ambush to a winter’s thaw.
Whether that’s better or worse depends on weights you assign — and the model lets you assign them honestly, with the math fully visible.
The playground asks whether you can find a setting where losing the predator lowers suffering per year. You can — and tracing why is more revealing than the answer.
Sweep the parameters under each ethical preset and four of the five can be pushed below a 1.0 after/before ratio. The exception is Ecological, which rates predation as natural and barely costly (w₁ = 0.3) while weighting disease and starvation heavily; for that worldview, removing the predator only ever makes things slightly worse, no matter how you tune the other dials.
The reason the others succeed is worth sitting with. Take Anti-Hunting weights, which treat a predation death as the worst kind (w₁ = 2.5). Before the tick, predation contributes the lion’s share of suffering — coyotes running deer down. After the tick, predation is zero, replaced by density-dependent disease and starvation. Roughly the same number of deer die either way. What changes is the label on each death, and this worldview considers a slow sick death much less bad than a violent fast one. Swap the labels and the weighted total falls. The win is real, but it was baked into the weights from the start.
The honest test is Utilitarian — every weight equal to 1, no thumb on the scale. Even there you can nudge the ratio just under 1.0, but only by stacking conditions: a late tick event, heavy compensating hunting, a high deer-toll τ that makes the pre-tick era worse to begin with. It is a fragile, marginal win. With neutral weights and ordinary parameters, removing the predator roughly breaks even or loses.
So the model doesn’t settle the question. It shows that the question is an ethical one wearing a quantitative disguise. If you believe a fast violent death is far worse than a slow sick one, predator loss can look like mercy. If you believe the reverse, it looks like a mistake. The differential equations are identical in both cases. The conclusion turns entirely on the numbers you, the reader, decided to put in.