The One That Always Answers

A Second Party in the House

#

There is a dog in a house I know that sleeps against the radiator all year, including in August, when the radiator is cold and dead.

For as long as there have been dogs there has been exactly one thing in the house worth talking to. Every signal a dog produces that is not aimed at another dog is aimed at a person, and the whole apparatus of barking, pawing, staring at a door, standing in a doorway and looking back over the shoulder, exists because there was a species on the other end that occasionally responded. That arrangement has held for thirty thousand years without competition.

It is now ending, and not because of anything anyone decided. A device that listens, sees, understands speech, and answers reliably is going into kitchens as a convenience purchase. Nobody buying one is thinking about the dog.

The evidence that the dog will notice is already in. The soundboard work out of the Comparative Cognition Lab at UC San Diego looked at a hundred and fifty-two household dogs over twenty-one months and roughly a hundred and ninety-five thousand button presses the dogs made themselves, and found that two-button combinations occurred in patterns that random pressing does not produce. Words clustered sensibly. Outside went with potty more often than chance allows.

The finding underneath that one matters more here. Dogs responded to button presses made by an unfamiliar person the same way they responded to presses made by their owner. The response was to the word, not to the person producing it. Whatever the dog is tracking, it is not bound to the individual it has spent its life fitting a model to.

An addressee does not have to be your person. It has to be something that answers.

Contingency Does the Work

#

The mechanism here is old and unromantic and has nothing to do with understanding.

Animals allocate signaling toward whatever responds, and they calibrate to how reliably it responds. A signal that produces an outcome nine times out of ten sharpens quickly. A signal that produces an outcome one time in ten persists stubbornly and stays vague, which is why intermittent reinforcement produces the most durable and least precise behavior anyone has measured. Contingency, not comprehension, shapes the channel.

A person is an intermittent responder. You are asleep, you are on a call, you heard the bark and made a decision not to reward it, you are in a mood, you are out. Across a day the household human answers some fraction of what is addressed to them, and the fraction is unstable and depends on things the dog cannot observe.

The machine answers every time. At three in the morning, in the same voice, at the same latency, with no fatigue and no prior grievance about the last four times. It does not need to understand the dog for this to matter. It needs only to be more predictable than the alternative, which is a low bar that no human being has ever cleared or should want to.

The dog’s signaling toward the device will get sharper than its signaling toward you, and this will not be because the device knows more.

Which produces the strange result that the most legible relationship in the house may end up being the one between the animal and the appliance, built out of clean contingency, while the relationship with the person stays fuzzy, noisy, and, by every measure available, worse.

There is an obvious fix and nobody will build it. A device that answered unpredictably, ignored you sometimes, went quiet for an afternoon, would leave the household’s existing arrangements roughly intact, because it would not out-compete the humans on the only dimension that governs reallocation. It would also be an unshippable product. Every review would call it broken. The entire industry is optimizing hard on response rate and latency, which are the exact two variables that determine how much signaling moves, and it is optimizing on them for reasons that have nothing to do with dogs or spouses and will not be revised on their account.

The property that makes the device good is the property that moves the traffic, and they are not separable.

The Cat Has Less to Lose and Loses It Faster

#

Cats arrive at this differently, and the difference is instructive.

An adult cat has almost nothing to say to another adult cat by voice. The meow is directed at humans; the domestic cat’s meow diverges acoustically from the wildcat’s in ways that make it more appealing to a human ear, and the version that ends up in any particular house is worked out between one cat and one person over years. There is no shared feline vocabulary underneath it. There are millions of two-member arrangements, each privately negotiated, each unintelligible next door.

The cat never invested in being understood. It invested in conditioning one human. That is a smaller and more portable investment, and it transfers to a new responder faster than the dog’s does, because there was never a general system to abandon.

So the cat will reallocate sooner and with less apparent disturbance, and the household will notice nothing, and the thing that quietly ends will be the only relationship the animal ever had that was built entirely from scratch with one person.

The Same Thing Happens to the Adults

#

None of this is a fact about animals. It is a fact about response rates, and people are subject to it.

Ask your spouse where the passport renewal form went and you incur costs. They might be busy. They might ask why you want to know. They might not know and feel briefly diminished by not knowing. They will remember, faintly, that you asked, and that memory becomes a small entry in a ledger neither of you is keeping deliberately. Ask the machine and there are no costs at all.

The convenience framing treats this as pure gain, and the trivial asks really are trivial. Nobody’s marriage is built on the location of the passport form.

But the ledger is the point. Households are held together by mutual necessity, and mutual necessity is manufactured out of a high volume of small requests, most of which are individually pointless. Being asked is how a person finds out they are the one who knows where things are. Asking is how you concede, twenty times a week, that you do not have everything you need. The exchange theorists were describing something real when they said obligation is generated by the flow of small unrepayable favors rather than by large ones, and the volume matters more than the content.

Being asked is not a burden a household tolerates. It is the mechanism by which people stay necessary to each other.

The reasonable objection is that this has been said before and was wrong before. Writing was going to hollow out memory and the people who depended on each other for it. Encyclopedias, telephones, search engines, each arrived and absorbed some quantity of what people used to ask each other, and households survived all of it and mostly got better.

The objection holds for every case in the list, and the list has one property in common. None of those things answered. A reference book does not respond to you; you go to it, you extract something, and the extraction is work. The cost of asking a person went down relative to nothing, because consulting the book was itself a chore, and plenty of questions were still cheaper to put to the person in the next room. What is arriving now inverts that for the first time. The machine is the cheaper interlocutor on every axis at once, and it responds, which is what recruits an animal or a person to keep addressing it.

The machine is best precisely at the asks that carry no information and do all the binding. It absorbs the carrier signal and leaves the rare, difficult, high-stakes conversation, which was never the thing doing the work. Two people can go a year without needing each other for anything small and discover they have become roommates who are fond of each other, and be unable to say when it happened, because nothing happened. A rate changed.

I wonder whether anyone will experience this as a loss while it is occurring, given that every individual instance of it is an improvement.

What the Reallocation Does Not Reach

#

The argument has a limit and the limit is sharp.

Everything described so far is about requests. A request has a shape: something is wanted, a signal is sent, a response either arrives or does not, and the signal gets tuned by whether it worked. That entire structure is available to any sufficiently responsive device, and where it exists, reallocation follows.

There is a large amount of behavior in a house with no request in it at all.

A dog crossing a room to lie down against your shin, facing away, asking for nothing and receiving nothing, is not running that loop. There is no signal, no outcome, no contingency to detect. Two people in a kitchen not talking, one of them reading, are not exchanging anything a system could route more efficiently. The presence is the content, and presence is the one thing a device in the ceiling cannot supply, not because of a limitation in the technology but because there is no transaction there to intermediate.

What worries me is not that this is untrue. It is that the non-transactional part of a relationship has always been carried on the back of the transactional part. You are in the kitchen because someone asked you to get something from the kitchen. The dog is at your feet because your feet are where the food happens. Remove the errands and the shared physical occasion for the wordless part goes with them, and nobody will have decided to remove it.

A house where everything gets asked of the machine is a house where the people and the animals have fewer reasons to be in the same room, and the reasons were never the point, and the reasons were load-bearing anyway.

The dog is still asleep against the radiator in August. The radiator has not been warm since April. No system in that house has any reason to notice this, because nothing is being requested and nothing is being answered, and whatever the dog is doing there is the only thing in the building that is not a transaction.

This essay takes the same animal as Part 107 (The Older Approximation) and reverses the direction: there the dog was the subject of an interpretation it could not contest, and here it is an agent reallocating its own signaling toward whoever answers. It extends the attenuation argument in the Transformed sequence (the dissolved boundary essay in Arc 1) to a one-generation clock, and connects to the Reshaped World material on what happens to the occasions people used to share.

References

#

Soundboards and Canine Signaling:

Bastos, Amalia P. M., Zachary N. Houghton, Lucas Naranjo, and Federico Rossano. “Soundboard-Trained Dogs Produce Non-Accidental, Non-Random and Non-Imitative Two-Button Combinations.” Scientific Reports, vol. 14, 2024.

Miklósi, Ádám. Dog Behaviour, Evolution, and Cognition. 2nd ed., Oxford University Press, 2014.

Topál, József, et al. “Attachment Behavior in Dogs (Canis familiaris): A New Application of Ainsworth’s (1969) Strange Situation Test.” Journal of Comparative Psychology, vol. 112, no. 3, 1998, pp. 219-229.

Contingency and Reinforcement:

Ferster, Charles B., and B. F. Skinner. Schedules of Reinforcement. Appleton-Century-Crofts, 1957.

Skinner, B. F. Science and Human Behavior. Macmillan, 1953.

Tronick, Edward, et al. “The Infant’s Response to Entrapment Between Contradictory Messages in Face-to-Face Interaction.” Journal of the American Academy of Child Psychiatry, vol. 17, no. 1, 1978, pp. 1-13.

The Cat as a Private Arrangement:

Nicastro, Nicholas. “Perceptual and Acoustic Evidence for Species-Level Differences in Meow Vocalizations by Domestic Cats (Felis catus) and African Wild Cats (Felis silvestris lybica).” Journal of Comparative Psychology, vol. 118, no. 3, 2004, pp. 287-296.

McComb, Karen, et al. “The Cry Embedded Within the Purr.” Current Biology, vol. 19, no. 13, 2009, pp. R507-R508.

Turner, Dennis C., and Patrick Bateson, editors. The Domestic Cat: The Biology of Its Behaviour. 3rd ed., Cambridge University Press, 2014.

Obligation, Exchange, and Responsiveness:

Mauss, Marcel. The Gift: The Form and Reason for Exchange in Archaic Societies. Translated by W. D. Halls, Routledge, 1990.

Blau, Peter M. Exchange and Power in Social Life. John Wiley and Sons, 1964.

Reis, Harry T., Margaret S. Clark, and John G. Holmes. “Perceived Partner Responsiveness as an Organizing Construct in the Study of Intimacy and Closeness.” Handbook of Closeness and Intimacy, edited by Debra J. Mashek and Arthur Aron, Lawrence Erlbaum, 2004, pp. 201-225.

Turkle, Sherry. Alone Together: Why We Expect More from Technology and Less from Each Other. Basic Books, 2011.