Consciousness, Hibernation, and the Machines We Build for Both

Picture this. You walk into your living room, and there’s your Labrador, paws resting on a tablet, mid-video-call with a golden retriever three states away. They’re not barking at random. They’re choosing to stay on the line. That’s not science fiction. That’s the work of Ilyena Hirskyj-Douglas and her team at the University of Glasgow’s Animal-Computer Interaction Lab.
 

Now, I’m a comparative physiologist. I spend my winters thinking about bears that drop their heart rate to nine beats per minute and ground squirrels whose body temperature hovers just above freezing. So when I hear about parrots making online friends and monkeys picking their favorite audio tracks, I don’t just see cute tech demos. I see a question that sits at the intersection of my field and Hirskyj-Douglas’s: What does it mean for an animal to have agency over its environment, and what can hibernation teach us about the biological boundaries of that agency?
 

This article compares two radically different strategies for surviving environmental extremes—bears and ground squirrels in torpor—and uses that comparative lens to ask whether the digital tools we’re building for animals might one day help us decode not just what they do, but what they experience. Along the way, I’ll borrow a few lessons from an unlikely cousin: the injection molding production line, where we’ve been troubleshooting “defects” in material behavior for decades, and where the principles of reversible metabolic suppression find a strange but instructive echo.
 

The Animal-Computer Interface: Agency as a Window Into Consciousness

Hirskyj-Douglas’s core argument is deceptively simple: if we give animals the tools to control their own digital environments—touchscreens for parrots, dog-friendly phones, audio-selection interfaces for monkeys—we don’t just entertain them. We create a behavioral window. When a parrot chooses to initiate a video call with another parrot, that’s not reflex. That’s preference. That’s choice. And choice, in the comparative physiology playbook, is one of the few observable proxies for something we’re terrible at measuring directly: conscious awareness.
 

But here’s where my training kicks in with a gentle corrective. Choice is expensive. Biologically expensive. Every decision an animal makes—every tap on a screen, every shift in attention, every voluntary movement—burns ATP. The brain alone consumes about twenty percent of resting metabolic rate in most mammals. So when I see a dog engaging with a digital interface, I don’t just see agency. I see a metabolic investment.
 

That investment tells us something about what the animal values. And values, in physiological terms, are just priorities encoded in energy allocation.
 

Bears vs. Ground Squirrels: Two Ways to Turn Down the Thermostat

Let’s get comparative. This is where the fun starts.
 

Black bears (Ursus americanus) hibernate for five to seven months. During that time, they don’t eat, drink, urinate, or defecate. Their metabolic rate drops to about twenty-five percent of basal levels—a seventy-five percent reduction. Their heart rate plummets from fifty-five beats per minute to as few as nine. But here’s the kicker: their body temperature stays relatively high, cycling between thirty and thirty-six degrees Celsius. That’s only a few degrees below normal.
 

Thirteen-lined ground squirrels (Ictidomys tridecemlineatus, formerly Spermophilus tridecemlineatus), on the other hand, take a much more dramatic plunge. Their metabolic rate drops by ninety-five percent during torpor. Their core body temperature can fall to near zero degrees Celsius. Their heart rate and ventilation drop to two to four percent of normal waking rates.
 

Same problem—surviving winter without food—but wildly different solutions.
 

Why? Body size. Small animals lose heat fast. To achieve the extreme metabolic suppression they need, ground squirrels have to let their body temperature crash. Bears, being massive, retain heat efficiently. They can afford to keep their temperature relatively high while still cutting metabolic demand by three-quarters.
 

This isn’t just a neat trivia fact. It’s a lesson in thermodynamic scaling that has direct implications for how we think about animal consciousness and animal-computer interaction.
 

Mitochondrial Reversible Inhibition: The Engine Room of Hibernation

Now we get to the cellular level, where the real magic happens.
 

During torpor, mitochondrial respiration is actively suppressed. And I mean actively—not just passively slowed down by cold. Studies show that metabolic rate begins to drop before body temperature falls during entrance into torpor. That’s a controlled, regulated shutdown, not a freeze.
 

What’s the mechanism? Post-translational modifications—chemical tags added to proteins that change their function without changing their sequence. Phosphorylation, acetylation, ADP-ribosylation, succinylation. These modifications alter the activity of key enzymes in the citric acid cycle, like the α-ketoglutarate dehydrogenase complex (KGDC), reducing their affinity for coenzyme A and slowing down the entire oxidative phosphorylation assembly line.
 

In plain English: hibernating animals don’t just run out of gas. They actively close the throttle on their mitochondrial engines.
 

And here’s the part that blows my mind: this suppression is reversible. During arousal, the modifications are removed, the enzymes regain activity, and the animal warms back up. It’s like flipping a master switch on the cellular power grid.
 

Ion channels also play a starring role. During hibernation, potassium permeability in liver mitochondria decreases. The ATP/ADP antiporter changes its behavior. Calcium and potassium transport into mitochondria is suppressed. These aren’t passive effects of cold—they’re regulated adjustments that protect the cell from energy crisis and oxidative damage.
 

Three Injection Molding Defects: When Material Behavior Mirrors Metabolic Suppression

Stick with me here. I know this sounds like a detour, but it’s not.
 

Injection molding is the art of forcing molten plastic into a cold mold, letting it cool and solidify, then ejecting a finished part. It’s a process of controlled phase transition—liquid to solid, with all the volumetric shrinkage, stress, and defect potential that entails.
 

Hibernation is also a controlled phase transition. Metabolic liquid to metabolic solid. Energy flow slows, material properties change, and if you don’t manage the transition carefully, you get defects.
 

Here are three real cases from injection molding production lines that, to my physiologist’s eye, look eerily like what happens when a hibernator’s regulatory systems fail.
 

Case 1: Short Shot — Incomplete Filling, Incomplete Torpor Entrance

The Defect: A polypropylene pencil box produced in a two-cavity mold kept coming out with incomplete filling in one cavity. The material wasn’t reaching the far end of the cavity before freezing.
 

The Parameters (Before): Melt temperature 210°C, mold temperature 40°C, fill time 1.2 seconds.
 

The Fix: The team increased melt temperature to 230°C, raised mold temperature to 55°C, and extended fill time to 1.8 seconds. Complete fill achieved. Defect rate dropped from twelve percent to under two percent.
 

The Physiological Parallel: Incomplete torpor entrance in ground squirrels—when they fail to achieve full metabolic suppression—leaves them burning too much energy too early. The fix, physiologically, is similar: raise the “temperature” (increase metabolic permissiveness) and extend the “fill time” (slow down the transition so all systems have time to engage). Hibernators that rush into torpor often arouse prematurely, wasting precious winter energy.
 

Case 2: Sink Voids — Uneven Cooling, Uneven Metabolic Suppression

The Defect: Thick-walled thermoplastic components developed sink voids—surface depressions caused by uneven internal shrinkage during cooling.
 

The Parameters (Before): Pack pressure 60 MPa, pack time 4 seconds, cooling time 15 seconds, mold temperature 45°C.
 

The Fix: Using a Taguchi L9 orthogonal array, the team varied four parameters at three levels each. Optimal settings emerged: pack pressure 75 MPa, pack time 6 seconds, cooling time 20 seconds, mold temperature 50°C. Sink voids virtually eliminated.
 

The Physiological Parallel: Uneven metabolic suppression across tissues—liver suppressing more than muscle, for example—creates internal “voids” of energy imbalance. The fix is better coordination: longer cooling (extended torpor bout duration) and more uniform “pack pressure” (consistent suppression signals across all tissues). Bears achieve this through multiday cycles of temperature and metabolic regulation. Ground squirrels, with their more extreme temperature drops, have to be even more precise.
 

Case 3: Flash — Over-Pressurization, Over-Suppression

The Defect: Flashing—excess material squeezing out between mold halves—accounted for up to ten percent of defects in one production line, against a customer tolerance of just two percent.
 

The Parameters (Before): Injection pressure 120 MPa, holding pressure 90 MPa, V-P switchover position (the point where filling switches to packing) at 95% fill.
 

The Fix: The team reduced holding pressure to 70 MPa and moved the V-P switchover earlier—to 92% fill. Flash dropped to 1.8%.
 

The Physiological Parallel: Over-suppression during torpor—pushing metabolic rate too low, too fast—can cause “flash” in the form of tissue damage from reactive oxygen species (ROS) when the animal eventually arouses. The fix is gentler pressure: less aggressive metabolic suppression, earlier “switchover” to the maintenance phase. Hibernators that over-suppress risk ischemia-reperfusion injury during arousal. The ones that get it right—bears, with their moderate seventy-five percent suppression—avoid the damage.
 

What Injection Molding Teaches Us About Animal Consciousness

You might still be wondering: What does any of this have to do with a dog on a video call?
 

Here’s my argument.
 

Animal-computer interaction, at its core, is about giving animals control over their environment—agency. Agency is a metabolic investment. When an animal chooses to engage with a screen, it’s allocating energy to that choice. That energy has to come from somewhere. And if we want to understand what that choice means—whether it reflects genuine preference, curiosity, or even something like consciousness—we need to understand the metabolic context in which that choice is made.
 

Hibernation gives us a natural laboratory for studying metabolic agency. A ground squirrel in deep torpor has virtually no agency. Its metabolic rate is suppressed by ninety-five percent. Its heart beats at two to four percent of normal. It’s not making choices. It’s in survival mode.
 

A bear in hibernation, by contrast, retains more metabolic flexibility. Its seventy-five percent suppression leaves room for some agency. It can shift position. It can even give birth. That flexibility comes from the bear’s less extreme metabolic strategy—higher body temperature, less dramatic suppression, more reversible mitochondrial inhibition.
 

Now apply that logic to a dog on a tablet. The dog is fully euthermic—normal body temperature, normal metabolic rate. Every tap on the screen represents a choice backed by full metabolic capacity. That’s a very different kind of agency than what a hibernating animal has. And that difference matters when we try to infer consciousness from behavior.
 

Hirskyj-Douglas’s work is brilliant because it creates the conditions for observable, repeatable choice behavior in animals. That’s the gold standard for behavioral research. But as a physiologist, I’d add one more layer: measure the metabolic cost of each choice. If a parrot’s video call costs more energy than the parrot is willing to spend, that tells us something about the value of the interaction. If a dog chooses to stay on a call despite the effort, that’s a signal worth paying attention to.
 

The Reversibility Principle: What Hibernation and Digital Interfaces Share

The most striking parallel between hibernation biology and animal-computer interaction is reversibility.
 

Mitochondrial suppression in hibernation is reversible. Post-translational modifications are added during torpor entrance and removed during arousal. The system resets. The animal comes back to full function.
 

Digital interfaces for animals are also reversible. The dog can walk away from the tablet. The parrot can end the call. The monkey can switch audio tracks. Agency includes the ability to disengage.
 

That reversibility is what makes both systems informative. A system that only goes one way tells you nothing about preference. A system that allows choice—enter and exit, suppress and arouse, engage and disengage—reveals what the animal actually values.
 

In injection molding, reversibility is harder. Once the plastic cools, it’s set. You can’t un-shrink a sink void. You can’t un-flash a flash. The process is largely irreversible, which is why defect prevention matters so much.
 

But in biology—and in well-designed animal-computer interfaces—reversibility is built in. That’s the feature that makes the data meaningful.
 

A Comparative Framework for Animal Agency

Let me consolidate this into a usable framework. When you’re evaluating an animal’s interaction with technology—or any environmental choice, really—consider these three physiological dimensions:
 

1. Metabolic Depth. How much energy is the animal investing in the choice? High-metabolic choices (e.g., a bear shifting position in the den) signal different priorities than low-metabolic choices (e.g., a ground squirrel adjusting its posture during torpor).
 

2. Thermal Context. Is the choice being made at normal body temperature or during hypothermia? Choices made during torpor—if any are possible—reflect different neural and metabolic states than choices made during euthermia.
 

3. Reversibility. Can the animal easily undo the choice? A reversible choice (end the call, walk away) tells you more about preference than an irreversible one (enter torpor, which commits the animal to a multi-day bout).
 

Hirskyj-Douglas’s animal-computer interfaces score high on all three dimensions. They allow high-metabolic engagement (the animal is fully awake), they operate at normal body temperature, and they’re fully reversible. That makes them excellent tools for studying animal preference and, by extension, animal consciousness.
 

What Hibernation Tells Us About the Future of Animal-Computer Interaction

Here’s where I’ll go out on a limb.
 

If we can build digital interfaces that animals choose to use, we’re not just entertaining them. We’re creating a new channel for animal expression. And if we can pair that channel with physiological monitoring—heart rate, body temperature, metabolic rate—we can start to decode the energetic signature of different choices.
 

A dog that chooses a video call over a treat? That’s a high-value social preference.

A parrot that initiates calls only at certain times of day? That’s circadian rhythm interacting with social motivation.

A monkey that selects calming audio tracks when alone? That’s emotional regulation, plain and simple.
 

None of this is speculative. It’s happening now. The Glasgow team has already designed a phone for dogs and a play-dating system for parrots. The question isn’t whether animals will use the internet. The question is what we’ll learn when they do.
 

And as a comparative physiologist, I’d argue that the most important thing we’ll learn is this: consciousness isn’t a binary. It’s a spectrum. And that spectrum is written in metabolic rates, mitochondrial modifications, and the reversible suppression of ion channels. The more agency we give animals, the more of that spectrum we can observe.
 

Practical Takeaways for Researchers and Designers

If you’re working in animal-computer interaction—or just thinking about building better tools for animal enrichment—here’s what I’d suggest, based on fifteen years of cross-species physiology consulting:

1. Measure metabolic cost. Every interaction burns energy. Track heart rate, respiratory rate, or even just activity level. The data will tell you what the animal actually values.

2. Design for reversibility. Make it easy for the animal to enter and exit the interaction. Choice requires both options.

3. Respect thermal limits. Don’t expect a hibernating animal to engage with technology. Wait for euthermia. The metabolic state determines the behavioral capacity.

4. Learn from manufacturing. Injection molding defect troubleshooting teaches us that controlled transitions—liquid to solid, active to suppressed—require precise parameter management. The same applies to transitions between wakefulness and torpor, and between engagement and disengagement with digital interfaces.

5. Watch for the “flash.” Over-stimulation—too much technology, too fast—can cause behavioral “flashing” just like over-pressurization in molding. The animal becomes agitated, disengages, or shows stress signals. Back off the pressure.
    

The Bottom Line: Agency Is Metabolic

Hirskyj-Douglas’s TEDxManchester talk asks: What happens when your dog uses the internet?
 

My answer, as a comparative physiologist, is: You get a window into what your dog actually wants.
 

Because every tap on that screen, every choice to stay on a call, every moment of engagement—it’s all powered by mitochondria. And mitochondria don’t lie. They don’t perform for the camera. They just burn ATP, or they don’t.
 

When a bear suppresses its metabolism to twenty-five percent of basal, it’s making a choice—not a conscious one, but a physiological one. When a ground squirrel drops to five percent, it’s making a different choice. And when a dog chooses to video-call another dog, it’s making a choice that sits at the opposite end of that spectrum—fully awake, fully metabolic, fully present.
 

That’s the gift of animal-computer interaction. It gives us a way to see animal preference in real time, without the filter of human interpretation. And if we’re smart about it—if we pair the digital data with the physiological data—we might just start to understand not just what animals do, but what they experience.
 

The dog on the internet isn’t just cute. It’s a metabolic signal. And signals, as any physiologist will tell you, are meant to be read.
 

References

• Hirskyj-Douglas, I. (2025). What happens when your dog uses the internet. TEDxManchester.

• Tøien, Ø., et al. (2011). Hibernation in Black Bears: Independence of Metabolic Suppression from Body Temperature. Science, 331(6019), 906-909.

• Evans, A. L., et al. (2023). Body mass is associated with hibernation length, body temperature, and heart rate in free-ranging brown bears. Frontiers in Zoology, 20, 27.

• Staples, J. F. (2014). Metabolic suppression in mammalian hibernation: the role of mitochondria. Journal of Experimental Biology, 217(Pt 12), 2032-2036.

• Staples, J. F., Mathers, K. E., & Duffy, B. M. (2022). Mitochondrial Metabolism in Hibernation: Regulation and Implications. Physiology (Bethesda), 37(5).

• Mathers, K. E., & Staples, J. F. (2019). Differential posttranslational modification of mitochondrial enzymes corresponds with metabolic suppression during hibernation. American Journal of Physiology, 317(2), R262-R269.