Why Bugs are Probably Conscious
The Best Explanation for the Evolutionary Purpose of Consciousness
A while back, I started poking around in the views of David Pearce.
His self-published internet manifesto, The Hedonistic Imperative (1995), outlines how pharmacology, genetic engineering, nanotechnology and neurosurgery could converge to eliminate all forms of unpleasant experience from human and non-human life, replacing suffering with “information-sensitive gradients of bliss”. Pearce calls this the “abolitionist project”.
I’m extremely sympathetic to the diagnosis. The prescription? That’s where I started thinking through things more carefully and that analysis started impacting my thoughts on consciousness.
Pearce's specific proposal is more sophisticated than a neutral indicator. He wants to relocate signaling onto positive gradients of bliss rather than eliminate affect entirely. But the broader transhumanist intuition runs through his work and others': that the aversive felt component of pain can be swapped for something less unpleasant (neutral information, positive gradient, whatever) without losing its protective function. To see why this is harder than it looks, consider the limit case. You break your leg, and instead of horrible pain, you get something like a flashing red indicator in your field of vision, more like a check-engine light.
Here’s the problem with that:
The CarMD Vehicle Health Index, published annually since 2011 (industry diagnostic data, not just self-report) found in 2024 that:
49% of surveyed drivers had a dashboard warning light on
22% had a check engine light specifically (double the previous survey)
12% had an oil/maintenance light
11% had ABS
7% had an SRS airbag warning
One poll of 2,000 American car owners also revealed that about one-third of drivers (32%) will wait “as long as possible” before bringing their car into the shop.
Other surveys put the "inspected within a week" rate at only 36.1%, with 25% ignoring entirely. About 22% of drivers have a tire pressure warning on at any given time.
These numbers are stable across surveys, and the drivers can articulate that the light matters - they just don't act.
This doesn’t bode well for the redesign of organisms to replace the sensation of pain with more neutral-feeling alerts.
Also consider Congenital insensitivity to pain (CIP):
The ability to sense all pain (including visceral pain) is absent, resulting in repeated injuries including: oral self-mutilation (biting of tongue, lips, and buccal mucosa); biting of fingertips; bruising, scarring, and infection of the skin; multiple bone fractures (many of which fail to heal properly); and recurrent joint dislocations resulting in joint deformity. Sense of touch, vibration, and position are normal.
CIP is the closest natural experiment we have to a system without felt pain. Informational and feedback sensations are present, but the affective sensation of pain is absent. (Pain asymbolia, where patients report "I feel the pain but it doesn't bother me", is the closer test of Pearce's specific decoupling thesis, and the limited data points the same direction: protective behavior degrades when affect is removed even if signal is preserved.)
The largest cohort study of CIP, Zhang and Haga's 2014 review of 91 patients, found that 65% had suffered fractures, 30% recurrent joint dislocations, 29% Charcot-style joint destruction, and 24% bone or joint infections. Most fractures occurred between ages one and seven. Where a cause could be identified at all, the most common one was minor trauma, short falls and ordinary daily activity. These patients can see their injuries, feel pressure and temperature, hear bones break, and intellectually understand that damage is bad. The full informational substrate is intact. What is missing is specifically the felt aversive component, and what we observe is that vision plus touch plus knowledge plus social input is insufficient to keep them safe. The affective signal isn't an artifact; it's critical.
Last year’s movie Novocaine featured a protagonist with this condition, who functioned almost like a superhero. But that was some Hollywood scriptwriter’s high concept for an action movie. In reality, the condition is horrific, and most importantly to my argument here, a massive hindrance to evolutionary fitness.
This got me thinking about consciousness/sentience in non-human animals and the evolutionary function of pain. The evidence above seems highly suggestive that for pain to be effective for survival, it can’t just be some passive signal. It needs to be affective. The organism has to really feel it.
In some organisms, the detection of pain is wired up to reflex circuits. What we’re probably looking for are cases where the pain detection is routed to a central integrated processing complex and is implicated in motivation and decision-making.
Nematodes (C. elegans) have nociceptors (specialized pain receptors) and avoidance behavior but the case for any phenomenal experience is very weak. The nervous system is 302 neurons total. Cnidarians (jellyfish) have detection machinery but no centralized nervous system. Sponges have the molecular receptors but no neurons at all.
The criteria the recent literature has converged on:
Nociceptors connect to integrative brain regions, not just to local reflex circuits
Behavioral responses to noxious stimuli are modulated by motivational state
The animal can trade off pain avoidance against reward
Conditioned place avoidance forms readily
Analgesic self-administration occurs given the option
Attention and learning are disrupted by noxious stimulation
The animal protects injured sites with site-specific behavior
Where the evidence currently stands by group:
Mammals and birds: All criteria met decisively. The 2024 New York Declaration’s “strong scientific support” category.
Reptiles, amphibians, fish: Most criteria met.
Cephalopods and Decapod crustaceans: Substantial evidence.
Insects: Not definitive, but suggestive. A comprehensive 2022 review by Gibbons, Crump, Barrett, Sarlak, Birch, and Chittka evaluated over 300 studies across 6 major insect groups against an 8-criterion framework for pain evidence. They argue insect sentience is plausible enough to warrant precautionary welfare consideration, while acknowledging definitive proof is absent.
Another angle to this question is to look at what affective pain sensation actually adds to the equation:
Non-dismissability. You can look away from a dashboard, silence a monitor, defer the check engine light. You cannot voluntarily stop attending to severe pain - it commandeers attention at multiple levels below voluntary control (reflex arc, brainstem, limbic, cortical).
Intensity scaling. A mild ache permits deferral; severe pain doesn’t, and it gets worse if you continue the damaging behavior. Warning lights are binary or low-resolution graded with no mechanism for escalating their demand on attention - exactly as easy to ignore the tenth time as the first. Pain has a gradient that maps onto action urgency.
Fused action script. Sharp pain in hand → withdraw hand. The sensory and motor components are fused at the spinal level and reinforced centrally. A decoupled signal requires interpretation, priority assessment, and voluntary action initiation, each step a documented failure point in the human factors data.
Aversiveness as the enforcement mechanism. The affective component isn’t a byproduct of pain. It’s what makes the system work. Strip the aversiveness and what remains is an informational signal whose compliance profile would look like a check engine light, not like pain.
In other words, unless the pain detection in an organism is directly hooked up to a reflex mechanism, if it’s routed to a brain and integrated in decision-making, it seems highly likely that it leads to something like a subjective experience of pain. Otherwise, it would function much more like a neutral warning, like a dashboard light, and be easily ignored by the organism.
So the argument here is not based on a direct ability to measure felt pain by other beings. It’s a fairly clear inference from data showing that non-felt tissue damage signals are highly ineffective at motivating an organism to act in ways critical to its well-being. Whereas experienced pain has very clear advantages: it works on a spectrum and you can’t ignore it.
Now you could design a system where non-affective signals increase work along a spectrum. We could design dashboard lights that increase in brightness, are accompanied by blaring alarms the longer they persist, or some other escalating machinery. But the bad news there is that there’s plenty of research to show saturation and desensitization effects.
Pain works so well because it is in your face. It’s a simple, stark signal that works because it is felt. And so upon quite a bit of reflection, I’ve become convinced that this is probably the most likely candidate for the evolutionary purpose of consciousness. Not modeling others' intentions or reflecting on one's own thoughts. Those additional functions probably came much later with the increased complexity of brains, layered on top of the original affective substrate rather than replacing it.
I think consciousness probably arose much earlier in the history of life, very soon after the first relatively complex brains arose, and that it likely did so because feeling pain provided such a massive survival advantage to organisms that had it.
ADDENDUM:
A friend pointed out that I didn’t delve too deeply into options for a more dashboard-style warning system for tissue damage. Instead of modifying the signaling, we could make improvements to the system that processes those signals, i.e. make more conscientious and responsive brains.
Maybe. A couple of thoughts.
One, evolution has had a really long time to find a system more like this. So far there is no really good evidence that it has. Animals that have the machinery for tissue damage detection also reliably display signs of affective pain. Gene knockout experiments have demonstrated that without affective pain, mice can retain immediate reflex behavior from dangerous stimuli, but behaviors like:
wound-directed care (licking, guarding the affected limb)
suspension of normal activities
learned avoidance of contexts where injury occurred
modulation of behavior over hours and days as healing proceeds
These seem to require affective pain being processed by a more complex brain over longer periods of time.
Pain has a lot of drawbacks. If evolution could find a system that accomplishes similar outcomes without the negative effects, or with fewer negative effects, we might expect to see at least one model organism that has them. So far that doesn’t seem to be the case.
You might argue we’d need more sophisticated brains. Humans haven’t been around a very long time, evolutionarily-speaking, but maybe long enough to see some subpopulations with these properties, especially if they led to better fitness. Again, we don’t. That doesn’t mean it’s not possible in theory. It may just be very difficult to find that fitness peak. Or maybe the problem is just intractable. Such a system would need the following characteristics:
Detect the signal reliably
Assign appropriate priority relative to other ongoing goals
Maintain that priority over time as long as the underlying damage state persists
Resist competing motivations of higher hedonic value (i.e., the cheeseburger, the deadline, the social commitment) when damage is severe
Generate action that actually addresses the damage rather than merely registering it
Scale all of the above continuously with damage severity
Felt pain does all this reliably, and has persisted as really the only strategy in the entire animal kingdom. Someone positing a system that does all these things without felt pain would need to at least propose some ideas about how it might work.
This would probably be easier in a system engineered from the ground up, and not one produced by evolution. We might imagine a robot that gets damage signals that are scalable, that it can prioritize relative to other goals, etc, and do so without anything like felt pain. Maybe with biological organisms, the pain solution is so effective it’s too attractive a basin.