by Brian Tomasik
First written: 29 June 2013; last update: 22 Dec. 2017


Even if the chance of bacteria sentience is exceedingly tiny, and even if it's very unlikely we'd give them comparable weight to big organisms, the sheer number of bacteria (~1030) seems like it might compel us to think twice about disregarding them. A similar argument may apply for the possibility of plant sentience. These and other sentience wagers use an argument that breaks down in light of considerations similar to the two-envelopes problem. The solution I find most intuitive is to recognize the graded nature of consciousness and give plants (and to a much lesser extent bacteria) a very tiny amount of moral weight. In practice, it probably doesn't compete with the moral weight I give to animals, but in most cases, actions that reduce possible plant/bacteria suffering are the same as those that reduce animal suffering.

Note, 2016: The language at the beginning of this piece about "probability of sentience" is confused. I now think that many disagreements about consciousness are mostly ethical rather than factual in nature.


Consider this Pascalian wager with respect to insects: Even if (1) the probability of insect sentience is modest and (2) it's not clear whether insects should be weighed equally with bigger-brained animals, if they are sentient and do count to a considerable degree, then their numbers are such that they dominate all other land wild animals in our calculations. (Zooplankton dominate in the oceans.)

This seems intuitive to me because I think the probability of insect sentience is not too small and because I think it's decently likely that, if they were sentient, I would weigh them nearly equally (at least within an order of magnitude) with big-brained animals. For concreteness, say the probability of insect sentience is 40% and I give a 50% chance of weighing them at least 1/10 as much as a human. This multiplies insect welfare by at least 0.02 relative to humans. Considering that their numbers are ~1018, this conservative calculation still gives ~1016 expected insects, compared with ~1010 humans.


In the past it also seemed naively intuitive to me that bacteria (and archaea) should not count. (I've since revised my view on this.) Still, what probabilities would I be willing to assign here? I don't think I can reasonably put the probability of bacteria sentience lower than, say, 10‑8, and if sentient, the chance that I would weigh them about equally (within 1-2 orders of magnitude of big animals) would not be smaller than, say, 10‑2. Say the total discount factor is very conservatively 10‑12 relative to humans. Alas, the number of bacteria on Earth is vast -- 5*1030 by one estimate. Discounted, this still gives 5*1018 in expectation, which is even bigger than the expected number of insects. Should I bite the bullet, or do I want to declare arbitrarily that bacteria don't matter even probabilistically because I choose not to count them?

There are a few ways we can try to wiggle out of this conundrum.

Our actions may not depend much on the answer

In most cases, things that reduce wild-animal suffering also reduce Pascalian bacteria suffering most of the time. However, there are some potential exceptions, like possibly eutrophication and crop cultivation, which may possibly increase bacteria while decreasing animals. As far as avoiding the spread of life into the cosmos, this is a good thing whether bacteria are sentient or not.

I know too little about bacteria to comment on whether there are personal actions that would reduce harm caused to bacteria in one's vicinity without increasing health risks. Maybe one example could be to use less deodorant when this is socially possible, since deodorants kill bacteria who are feeding off your sweat. Assuming the food energy contained in your sweat remains viable after it's washed down the drain, it'll just be eaten by other bacteria later on, so killing bacteria on your skin only increases total bacteria deaths per unit of sweat.a

Similarly, I wonder if putting bacteria-killing disinfection chemicals and antibacterial soaps down the drain increases total bacteria suffering because it kills bacteria in your septic system or wastewater-treatment plant without eliminating their food, which will just be eaten later by other bacteria? Wikipedia explains: "Certain chemicals may damage the components of a septic tank or kill the bacteria needed in the septic tank for the system to operate properly, such as pesticides, herbicides, materials with high concentrations of bleach or caustic soda (lye), or any other inorganic materials such as paints or solvents."

In general, it seems usually bad to prematurely kill bacteria, because the food the current bacteria are eating will be eaten by some other bacteria later, so killing current bacteria only increases the total number of bacteria deaths that will eventually take place. Exceptions to this are if the food of the current bacteria

  1. would otherwise degrade without major bacterial help in the digestive systems of big animals
  2. would otherwise degrade non-biologically, such as in the case of dry plants that will be burned by fire.

One example of an action that seems likely to reduce bacteria suffering is to be cremated when one dies. Burning a body and its casket prevents bacteria (and, in a few cases, maggots) from being able to feed on it, thereby preventing many bacteria from coming into existence. (This analysis ignores climate-change considerations. This article says "Cremation is relatively benign compared to conventional burial, thanks in part to the required filtering of emissions done by crematories in the United States. Still, the average cremation uses 28 gallons of fuel to burn a single body, emitting about 540 pounds of carbon dioxide into the atmosphere.")

In general, burning organic matter rather than letting it decompose averts bacteria suffering (again, ignoring climate change).

By the same logic, spilling oil and thereby feeding bacteria, rather than burning that oil, increases bacteria suffering:b

Ocean currents, in addition to keeping the spilled oil offshore, spurred microbial activity amidst the oil spill. That continuous mixing of the water allowed a bacterial bloom to turn millions of barrels of oil into an estimated 100 sextillion microbial cells of ethane-consuming Colwellia, aromatic-eating Cycloclasticus, alkane-eating Oceanospirillales, oil-eating Alcanovorax, methane-loving Methylococcaceae and other species, including at least one previously unknown to science.

Some instances of heating food, such as boiling vegetables or toasting bread, are unnecessary. It would be interesting to explore whether it's more "painful" for a bacterium to be toasted/fried/boiled or to be killed in the hydrochloric acid of one's digestive tract.

Death rates

It's important to ask not just about population sizes but about death rates. For every adult insect in the population, there are tens/hundreds of deaths per life cycle because each parent pair has tens/hundreds/thousands of offspring, almost all of which die before maturity. That implies a few orders of magnitude more than 1018 deaths per year by insects. It's plausible that death is the most painful thing experienced by most of these short-lived critters.

Do bacteria have an analogue of high infant-mortality rates the way animals do? Or do they mostly survive pretty well and then stop reproducing once they reach carrying capacity? Also relevant is adult lifespan, which is generally on the order of hours to weeks for bacteria. Even if bacteria didn't die prematurely the way most animals do, their average lifespans would still be extremely short, probably shorter than average lifespans of insects even counting non-surviving offspring. This suggests the following question that I think needs further exploration.

Open Question: Suppose that typical bacteria have lifespans of several hours, while insects have average lifespans of several days or weeks when we count non-surviving larvae. (Are these numbers roughly correct?) Does this mean that, adjusted for the sentience of the different organism types, bacteria suffer more per unit time than insects because they die more per unit time? Is death painful for bacteria the way it generally is for animals? Even if we think bacteria are less sentient per unit of food consumed than insects, do their higher death rates imply more total suffering per unit of food consumed than for insects? In other words, even if we think "sentience per unit of food eaten" is lower for bacteria than for insects, if we think "fraction of life consisting of bad experiences" is higher for bacteria, maybe the product of the two factors

total suffering per unit of food eaten = (sentience per unit of food eaten) * (fraction of life consisting of bad experiences)

could be higher for bacteria than for insects? Intuitively I still feel like I care more about insects than bacteria per unit of food eaten, but maybe this reflects scope insensitivity on my part.

This video shows bacteria being killed.

Cells in big animals might count also

If bacteria count as sentient, how about cells within bigger organisms? Since we're no longer restricting ourselves to beings with neurons, anything is up for grabs, including our skin cells, blood cells, muscle cells, etc. Like bacteria, these can be injured and stressed, so -- unlike for rocks or atoms -- there is a plausible direction in which they can be more harmed than helped.c There are two objections to this line of reasoning, though.

  1. A would-be bacteria advocate might contend that cells in bigger organisms are not autonomous enough to count as a unique individual, so the organism counts as just one instead of a sum of trillions of smaller individuals. However, maybe autonomous cells like spermatozoa would count as individuals,d and other kinds of cells too may engage in competition with each other as though they're individuals (including, speculatively, neurons).
  2. Bacteria cells are about 1000 times smaller than the cells in the human body, as a result of which the human body may contain more bacteria than cells of its own. If this is true for macroscopic organisms generally, then bacteria can still dominate.

Abilities of unicellular organisms

This section has moved here.


See also "Plant cognition".

We could proffer a similar Pascalian wager with respect to plants: While not as numerous as bacteria, plants are more likely sentient. They lack neurons, but they do exhibit adaptive behaviors in response to stimuli. They have memories and transmit electrical impulses to convey messages. While plants would not have the same need for pain as animals who can run away, they do still benefit from information processing, such as for tissue response to stress, or even to communicate with one another to warn of impending danger and the need to boost defenses. They use a type of analog division to compute energy reserves during the night.

An excellent documentary on this topic is Nature's "What Plants Talk About," which shows, among other things:

  • A vine plant selectively targets its preferred hosts by rotating around in circles, "sniffing" for a good scent. When it finds its favorite host, it attaches to the stem and begins to suck its juices.
  • Some plants signal distress in response to insect infestation in order to attract other insects that eat the bugs that are eating them.
  • Some plants use chemical signals to determine which roots belong to themselves and kin, in order to avoid competition with those who share their genes.
  • Plants compete for sunlight by moving their leaves in front of each other. When viewed with time-lapse photography, this looks almost indistinguishable from animals knocking their paws against one another.

If plants used neurons to engage in these behaviors, people would find the notion of plant sentience more intuitive. But neurons are just one particular form of transmitting information, and plants accomplish similar functions at a slower pace by chemical means. (Animals also use chemical messengers in the form of hormones, which are slower than neural messengers.)

Michael Pollan's excellent article, "The Intelligent Plant", explores further ways in which plants act somewhat like slow-motion animals. Pollan reports that many plant scientists reject comparisons between plants and animals for fear of sounding silly. I can see a legitimate concern that overly anthropomorphic language might incline lay people to attribute too much cognitive ability to plants, but I also wonder if part of the resistance comes from the skeptical scientists not having internalized how "dumb" and "brainless" the subsystems of human nervous systems also are. Studying neuroscience helps strip away the magic that we associate with brains and makes the brain look like just another information-processing device among many.

Many neuroscientists, including Giulio Tononi, consider network connectivity that unifies many distinct sources of information to be an important component of when we think of a system as conscious. Plants have rudimentary network connectivity for information transmission, both internally in their vasculature and using chemicals in the external environment.

Pavlis (2012): "Roots don’t have a brain and can’t think through the process, but plants are able to direct root growth to areas where they are finding the best growing environment."

Daniel Chamovitz said in an interview with Vice: "We know that when aphids attack leaves, it elicits an electric signal in plants that goes from leaf to leaf to signal it to start protecting itself. It's propagated very similarly to the way it's propagated along a nervous system. And they do this all without a neural system. The take-home message is that neural systems are one way to process information, not the only way."

Gustav Fechner wrote in "Nanna, or on the Soul-Life of Plants" (1848):

[I]s not the plant quite as well organized as the animal, though on a different plan, a plan entirely of its own, perfectly consonant with its idea? If one will not venture to deny that the plant has a life, why deny it a soul? For it is much simpler to think that a different plan of bodily organization built upon the common basis of life indicates only a different plan of psychic organization.

As in the case of bacteria, we can offer a few responses to these points.

Our actions may not depend much on the answer

Once again, actions that reduce animal suffering usually also reduce potential plant suffering, with some possible exceptions. Indeed, it seems the most secure way to reduce animal populations is to reduce plant populations. As far as diets, the standard vegetarian reply is that most meat production requires killing many times more plants than eating plants directly, so even if plants feel pain, it's better to be veg. Vegetarianism may be best even when we consider wild plants if global warming increases plant abundance in the long term. That said, farmed crop land may contain fewer plants than wild land, which reduces both plant and animal suffering. The questions here about the net sign of farming for potential plant suffering resemble those for insect suffering.

What would be the quality of life for wild plants? Like with animals, most plant offspring probably die soon after pushing out of their seeds. Only a few mature to adulthood, I would guess. For example, the following figure is from this textbook:

Plants can also have stresses due to hostile weather, lack of food, drought, being eaten alive by herbivores, etc. Indeed, it seems like many of plants' most intelligent behaviors are reactions against negative stimuli (herbivores, disease, cold, injury).

Unfortunately, upon realizing that plants are marginally sentient, the reaction of most people is to protect plants rather than to eliminate them in order to alleviate their suffering and the suffering of the animals they support. Pollan reports that Stefano "Mancuso believes that, because plants are sensitive and intelligent beings, we are obliged to treat them with some degree of respect. That means protecting their habitats from destruction [...]." Likewise, this piece on plant intelligence concludes with a pro-conservation message. But in fact, given that many plants endure significant harm in the wild, it seems generally more humane to prevent plants from coming into existence.

I think it's probably bad to kill healthy herbivores. In that case, shouldn't it also be bad to kill healthy plants? It's not clear, and there is some direct moral cost in harming healthy plants. But in many cases, killing plants -- such as by not watering them, covering them, chopping down trees, etc. -- probably reduces primary productivity and hence may be worth the short-term harm that the plant endures. Herbivores reduce wild-animal suffering by eliminating biomass, while plants increase wild-animal suffering by creating biomass.

There could be extremely minor areas of practical significance. For instance, I wonder if seed sprouts cause more harm than eating bigger plants because you kill so many inchoate plants when eating sprouts. (Of course, there might also be lots of fledgling plants on fields that die naturally in the process of farming big plants, but in terms of plant deaths per kg of food, sprouts still seem considerably higher.) Because I have no preference between sprouts and bigger vegetables, I decided I may as well avoid the sprouts. That said, it seems plausible that we might weight the harm of injuring a plant roughly in proportion to its size, because the leaves of a big plant can be seen as a lot like "sprouts" of their own. Plants have a fractal-like structure.

Another idea to ponder further is whether there's such a thing as humane slaughter for plants. For example, I would conjecture that pulling a plant up by its roots might be "less painful" than cutting its stem, because uprooting the plant cuts off nutrients and water. In contrast, a plant with its stem cut off may send damage signals to remaining tissues? Of course, even an uprooted plant would survive for some time on its existing nourishment, similar to how the head of a chicken still blinks and opens its mouth in terror even after being severed from the body.

Clock speed

As Hedonic Treader pointed out on Felicifia, plants respond much slower than animals, so maybe the effective clock speed for plants is (orders of magnitude?) slower than for animals, which might be enough to dampen out their otherwise potentially massive expected value. Pollan's article explains how "the lives of plants unfold in what amounts to a much slower dimension of time", and Mancuso analogizes with the Star Trek episode "Wink of an Eye".

Other Pascalian wagers?

It's possible one can also make Pascalian wagers about certain operations in computers. Running fewer such operations would mean less Pascalian expected suffering. There may be similar wagers in other domains. In general, it seems that reducing the harm of space colonization is a way to avert most of these Pascalian worries.

Revising our approach to sentience wagers

These Pascalian wagers seem troubling. Are we doing something wrong? Or should we just bite the bullets? Or do we want to revise what we care about to patch over these holes? Maybe we can just leave the problem to future humans to figure out?

Two-envelopes problem

If our Pascalian wager about bacteria and plants assumes a binary notion of sentience (either something is fully sentient or not at all), then the wager can work. On the other hand, if we frame the wager as I did above, with some probability of counting bacteria/plants at least some fraction as much as humans, with their moral weight possibly being lower or higher, then we succumb to the two-envelope fallacy for Pascalian brain-size calculations discussed here: "Two-Envelopes Problem for Brain Size and Moral Uncertainty."

Graded sentience

In light of the two-envelopes problem, we may wish to choose an alternate perspective other than moral uncertainty, at least to make our judgments more concrete before figuring out a better resolution of two-envelopes issues. I propose the following view of the situation, which gives very small but nonzero weight to plants and bacteria. I still think animals should dominate our considerations, although as noted, there's generally little difference between what's best for plants/bacteria and what's best for animals.

I think talking about sentience as binary is not the best way to approach the situation. Fundamentally, consciousness comes in gradations, and there's not a magic point where something suddenly becomes conscious by adding one small extra ability. Rather, organisms have many types of mental processes that can become more and more complex, more and more similar to our own, in shades of gray.

Compare to another debate: When does a parahuman (human-animal chimera) begin to count as an actual human? Sometimes this debate plays out in a very binary fashion: Humans matter a lot, animals don't matter at all, so finding the crucial dividing line between the two is seen as an important moral question. But this is the wrong way of viewing the issue. Animals and humans are already on a graded spectrum of moral value, so different parahumans would just fall at different points in between. Trying to force complex systems into binary categories tends not to work so well. The debate about "when does human life begin?" (at conception? at birth?) falls into the same trap of excessively binary categories forced onto something that's fundamentally a continuous property.

So too in the realm of consciousness, we should realize that our concern for other organisms is a continuous function of their complexity and attributes, and this concern doesn't completely drop to zero when the organism becomes very simple, even perhaps at the level of bacteria. In "Why Robots Will Have Emotions," Aaron Sloman and Monica Croucher put it this way:

There is a huge array of possible cases, from systems with very simple feedback loops to those containing all the complexity sketched below. Arguing about where to draw the line between cases of real intelligence or mentality [or, in this case, sentience] and the rest is quite pointless, like arguing over whether it is still 'really' chess if one player accepts the handicap of playing without a queen. (In the end, some of the decisions are ethical.)

That said, our caring level could drop significantly between, say, a mouse and a bacterium, and defining the shape of this caring-about function is a nontrivial problem that requires us to understand how these systems work and then search our souls for how much we're moved by the various components. The weighting might not be purely based on size or complexity; we might also give consideration to the unified quality of an organism as an individual entity, to the way it marshals seemingly dumb subcomponents into a highly intelligent synthesis, or to other properties.

Maybe bells and whistles matter, but maybe they only matter if you've got some essential ingredient? For example, maybe some component of consciousness is a binary on/off switch for moral consideration, and then if it is on, we can still have degrees based on the organism's complexity? Sure, you could do this; what you care about is allowed to be anything you choose. However, keep in mind that (a) we're probably not going to find a single simple trait that captures everything that moves us about a mind in agony or in ecstasy and (b) even if we did, we might see resemblances of that same property in other physical systems, with some nontrivial interpretational weight. In general, continuous functions behave better than binary ones, and it seems likely that our caring-about function should have a nonzero value over basically all physical processes; the real question is how steep the drop-offs should be from obviously awful things like a chicken being scalded alive in a defeathering tank to things that may evoke only the slightest moral concern, like poisoning a fungus with chemicals.

Sometimes people compare consciousness with a phase transition of matter in order to suggest that there's a binary cutoff for its emergence. But even phases of matter are not binary. Atoms in a solid move around, and they move more as a continuous function of temperature. Water molecules in a liquid may spontaneously evaporate below 100 degrees C, and they evaporate more and more as the temperature increases. It's true there's a point where the dynamics shift dramatically, but along almost any metric we might use (density, amount of movement, resistance to impact), that metric as a function of temperature is still continuous, even if it's quite steep at two points (melting point and boiling point).

To sum up: Bacteria and plants do contain rudiments that are found at a more advanced level in higher animals, and they do resemble in a very rough way some parts of complex brain processing. To this degree, I would care about them, but exactly how much weight they deserve is unclear. Probably I would allow vast numbers of bacteria to be born and then die soon thereafter to prevent one mouse, or even one mosquito, from suffering the same fate.

I still think insects are a big deal

The Pascalian argument for taking insects seriously is similar to that for bacteria and plants, so if two-envelopes issues raise doubts about the latter, they also raise doubts about the former. However, I think the importance of insect welfare is compelling even without the Pascalian reasoning, just based on the shape of my caring-about function. Keep in mind that two-envelopes problems only apply to moral uncertainty, rather than factual uncertainty. Given some reasonable definition of conscious suffering, it's an open factual question whether insects have such conscious suffering in ways similar to what happens in bigger animals. If so, I think insects would deserve appreciable moral weight, because they're autonomous agents that have their own utility scales, and their brains perform many of the same functions as ours do with much greater efficiency. I would not put equal weight on an insect as a mammal, but the weight would be within a few orders of magnitude, and then the sheer number of insects and untapped potential for helping them propel the ethical urgency.

I think insects matter more than their proportional share of neurons compared with a human, or even their proportional share of neurons that perform hedonic rather than cognitive operations. This is partly because of greater efficiency and partly because each insect is its own agent, and it feels like there should be diminishing returns to the size of the agent. On the other hand, I seem to care less about a mass of bacteria the size of an insect than about an insect. This is implied, for instance, by my view that it's better to let food decompose anaerobically in a landfill than via decomposing insects in a compost heap. But bacterial "minds" are arguably even more efficient than insect minds, and each bacterium is also it's own agent. So why don't I care about an insect-sized glob of bacteria more than an insect? Maybe it's because the bacteria are so simple that the increase in importance with size is more than linear in size between them and insects, and then after insects, the increase is less than linear. This seems plausible to me, but it's worth pointing out that it's a subtle and non-obvious stance to take. Perhaps my intuitions will change with time.

This figure illustrates my sentiments on relative importance of different organisms as of early 2016:

Why do I say that humans have fewer agent-like subprocesses per gram than insects?

  • My impression is that human-brain subsystems are more "cogs in a machine" than independent insects are. That said, even individual human neurons can display competitive, agent-like behavior.
  • Human brains may have extra neurons for redundancy and greater detail of sensory information, but in many cases, the extra neurons may not be doing anything fundamentally new.

The humped curve in this figure for moral value per gram may seem inelegant -- why not either let the biggest organisms or the smallest organisms dominate in importance? That said, most animal-rights activists also implicitly maintain a curve for moral value per gram that has a hump in the middle. This is because they tend to care equally about all animals above some complexity threshold. This view implies that, e.g., a dog and a chicken have equal moral importance, which means a chicken has more moral importance per gram. But bacteria, which are below the complexity threshold, have less moral importance per gram than a chicken.

Prokaryotes below the Earth and in space

In "The deep, hot biosphere" (summary here), Thomas Gold discusses life below the Earth's soil and oceans. Of this subsurface life, he suggests that "In mass and volume it may be comparable with all surface life" (p. 6045). Hopanoid molecular fossils of subsurface bacteria have been estimated by Ourisson et al. at around 1013-1014 tons, compared with ~1012 tons of organic carbon in life on or near the Earth's surface (p. 6046). Gold continues with a calculation of how much of the Earth's depth could be inhabited by bacteria based on temperature constraints. He estimates that the bacterial mass might be ~2*1014 tons, "equivalent to a layer of the order of 1.5 m thick of living material if spread out over all of the land surface." That said, he cautions, we don't know enough to have high certainty in this figure.

"Prokaryotes: The unseen majority" suggests the following numbers of prokaryote cells and their generation times:

Where? How many cells? Generation times
Open ocean 1.2 * 1029 6-25 days in upper 200 m; 0.8 yr below 200 m
Soil 2.6 * 1029 2.5 yr
Terrestrial subsurface 0.25-2.5 * 1030 1000-2000 yr?? (huge uncertainty; authors admit it seems weird)
Oceanic subsurface 3.5 * 1030 1000-2000 yr?? (huge uncertainty; authors admit it seems weird)

However, "Life in the deep subsurface" reports that other authors are skeptical of the numbers in this table because the subsurface prokaryotes we know about so far might have been sampled from regions of particularly high density. Other research suggests that cell counts may be lower than previously assumed. For instance, "Subseafloor sedimentary life in the South Pacific Gyre" found that "mean cell abundances are 3 to 4 orders of magnitude lower than at the same depths in all previously explored subseafloor communities."

"Subseafloor sedimentary life in the South Pacific Gyre" also notes that respiration in their communities proceeded at an extremely slow rate. In ethical terms, this suggests that even if we care about these prokaryotes, their "clock speeds" may be orders of magnitude slower than for surface cells. A similar point is suggested by the generation times in the above table.

The discovery of subsurface prokaryotes may have bigger implications than just numbers of microorganisms on Earth. "The deep, hot biosphere" explains (p. 6045, 6048):

Subsurface life may be widespread among the planetary bodies of our solar system, since many of them have equally suitable conditions below, while having totally inhospitable surfaces. [...] If in fact such life originated at depth in the Earth, there are at least 10 other planetary bodies in our solar system that would have had a similar chance for originating microbial life. [...] The surface life on the Earth, based on photosynthesis for its overall energy supply, may be just one strange branch of life, an adaptation specific to a planet that happened to have such favorable circumstances on its surface as would occur only very rarely: a favorable atmosphere, a suitable distance from an illuminating star, a mix of water and rock surface, etc. The deep, chemically supplied life, however, may be very common in the universe. Astronomical considerations make it seem probable that planetary-sized, cold bodies have formed in many locations from the materials of molecular clouds, even in the absence of a central star, and such objects may be widespread and common in our and in other galaxies.

Finally, "The deep, hot biosphere" asks us to consider whether there are systems even simpler than bacteria that nonetheless express some life-like traits. There's a vast stretch of chemical activity in the Earth at temperatures too hot for bacteria:

but that is nevertheless capable of supporting other systematic chemical processing systems that can mediate those energy reactions. Could there be such higher temperature systems that act in a way similar to life, even if we may not identify them as life? [...] Self-replication is a property possessed by simple crystal growth: it is only when self-replication is associated with an adaptive capability that the complex forms develop that we identify as life. In the case of unfamiliar circumstances and materials, we may fail to recognize these properties.

Life on stars?

The possibility of prokaryotes deep inside extraterrestrial planets seems quite plausible. Much more speculative is a possibility of life on stars -- neutron stars specifically. According to physicist Robert Forward's New Scientist article "When you live upon a star...": "A neutron star is more like a miniature planet than a star" (p. 36). It has a solid crust with many of the same nuclei as atoms on Earth except with more neutrons (p. 36). In the 1970s, Frank Drake speculated about whether life existed on the surface of neutron stars. Such creatures would be microscopic and would have clock speeds a million times faster than humans' (p. 37). Forward expanded upon Drake's proposal in the novel Dragon's Egg. Forward confesses that it's unknown whether neutron stars could form the equivalent of complex molecules at high gravities, but he wrote his story on the assumption this is possible (p. 38). The so-called "cheela" creatures in Forward's novel have human-level intelligence, but it seems much more likely that if life exists at all on neutron stars, it more closely resembles bacteria.

Closing quotation

Harry Potter and the Methods of Rationality, Chapter 48:

"Look, it's a question of multiplication, okay? There's a lot of plants in the world, if they're not sentient then they're not important, but if plants are people then they've got more moral weight than all the human beings in the world put together. Now, of course your brain doesn't realize that on an intuitive level, but that's because the brain can't multiply. Like if you ask three separate groups of Canadian households how much they'll pay to save two thousand, twenty thousand, or two hundred thousand birds from dying in oil ponds, the three groups will respectively state that they're willing to pay seventy-eight, eighty-eight, and eighty dollars. No difference, in other words. It's called scope insensitivity. Your brain imagines a single bird struggling in an oil pond, and that image creates some amount of emotion that determines your willingness to pay. But no one can visualize even two thousand of anything, so the quantity just gets thrown straight out the window. Now try to correct that bias with respect to a hundred trillion sentient blades of grass, and you'll realize that this could be thousands of times more important than we used to think the whole human species was...[.]

While it's clear from my essay that I don't endorse this view all things considered, the argument is naively compelling, and Harry surely understands how I feel when I look at the world. I do believe that the potentially vast suffering of small organisms is among the most urgent moral issues -- one which most of humanity has not yet appreciated.


  1. If the food contained in your sweat breaks down abiotically if not eaten by bacteria (which seems unlikely??), then it could be better to kill bacteria with deodorant to reduce the total number of bacteria that come into existence.  (back)
  2. We can also ask whether burning fossil fuels in general increases or decreases bacteria populations relative to keeping those fossil fuels in the ground, ignoring side effects like climate change and air pollution. I know little about this topic, but here are some speculations. If we thought that fossil fuels in the ground would eventually be eaten by bacteria (would they?), then burning those fuels could prevent those bacteria from coming into existence. However, I assume fossil fuels in the ground will remain uneaten for a long time, and maybe our digital artificially intelligent successors will destroy Earth and its fossil fuels in an abiotic way long before those fossil fuels would have been consumed by bacteria. Meanwhile, withdrawing fossil fuels to burn them creates hydrocarbon byproducts, oil spills, etc., which I assume feed some bacteria in the near term.  (back)
  3. In particular, negative-leaning utilitarians would hope to have as few cells as possible. Of course, they might also hope to have as few rocks as possible for Pascalian reasons, but whatever you turn rocks into would have a roughly equal chance of sentience as the rocks did. In contrast, bacteria have a higher chance of sentience than the lifeless components of which they are made. I suppose if one thought that energy had a lower chance of sentience than matter, one could aim to convert rocks to energy, but this sounds expensive for very little gain relative to reducing cells unless it could be done extremely efficiently. And who knows -- maybe energy would contain more suffering than matter.  (back)
  4. Men produce a lot of sperm: about 100 million per day.  (back)