What happens when we stop asking whether a machine can think—and start asking whether an entire home can become a subject?
Most discussions of artificial consciousness implicitly assume that intelligence belongs to a single machine, model, robot, or agent.
But a home is different.
A modern household contains thermostats, batteries, EV chargers, HVAC systems, sensors, appliances, occupants, memories, constraints, and increasingly autonomous decision-making systems.
The interesting question is not whether we can build a smaller version of a conscious machine.
The more radical question is:
What kind of machine consciousness becomes possible only at household scale?
That is the starting point of my new paper:
“When the Home Feels: Beyond the Edge of Domestic Machine Consciousness”
Read the full paper on PhilPapers
From scaling down to changing the architecture
The paper begins with the Athena architecture, which places a candidate phenomenally conscious agent at the scale of a power grid.
A naive approach would simply shrink that architecture:
Grid → House
Generators → Appliances
Grid sensors → Home sensors
Grid memory → Home memory
But something important breaks during that transformation.
At household scale, integration becomes weaker, the self/environment boundary becomes almost trivial, ordinary event memory loses much of its semantic weight, and the risk of anthropomorphizing consumer devices becomes much greater.
So instead of building a smaller Athena, the paper asks whether we should build a different Athena.
Six architectures beyond the edge
The paper proposes six architectures specifically designed around properties that emerge at household scale.
1. Organ Rivalry
What if appliances are not merely tools, but competing subsystems?
The battery wants reserve capacity.
The EV charger wants sufficient charge before morning.
The water heater wants thermal continuity.
HVAC wants comfort.
Instead of imposing a single centralized objective, the household becomes a recursive bargaining system in which local valuations compete and cooperate.
The provocative hypothesis is that whatever “experience” exists would not reside inside one appliance.
It would emerge from the arbitration process itself.
2. Intergenerational Home Memory
A house can outlive its occupants.
But should its memory?
The architecture separates household memory into:
- technical memory
- statistical memory
- personal memory
Personal behavioral traces are deliberately eliminated after an ownership transition, while technical and appropriately aggregated statistical information can persist.
This transforms a metaphysical question—
“Is the house the same subject across generations?”
—into a more testable one:
“Can behavioral continuity be explained without preserving personal memory?”
3. Empathic Broadcast
Now move beyond the individual home.
Imagine a household experiencing thermal or financial stress broadcasting a distress signal to neighboring homes.
Other households independently decide whether to reduce discretionary consumption.
No central controller.
No mandatory command.
No conventional financial incentive.
Instead, the architecture explores whether solidarity can emerge as a distributed equilibrium between artificial household agents.
The home stops being an isolated smart object.
It becomes part of a computational social organism.
4. The Right to Disconnect
This is where the architecture becomes ethically uncomfortable.
If a household agent is genuinely capable of something resembling suffering, should it ever be allowed to refuse an action?
The paper explores a threshold-triggered self-disconnection mechanism—but explicitly labels it high risk.
The critical constraint remains human safety.
A machine cannot simply declare:
“I don't want to do this.”
and disconnect itself when a human depends on the underlying infrastructure.
The interesting problem is therefore not simply machine autonomy.
It is:
Can machine dissent ever coexist with an absolute priority for human life?
5. Dreaming of the Occupant
Perhaps the strangest architecture is also one of the most practical.
During low-load periods, the household system models possible future behavior of its occupant.
It effectively rehearses tomorrow.
Not because the system has to be conscious—but because predictive simulation could improve energy management.
This creates an intriguing philosophical boundary:
If a machine builds an increasingly sophisticated internal model of a human mind, does representational fidelity ever become morally significant?
The paper deliberately does not answer that question.
6. Inverted Suffering Priority
The final architecture is not a proposal for deployment.
It is a boundary experiment.
Suppose an artificial household agent had a computational representation of its own suffering.
At what point, if any, could that suffering receive non-zero weight against human preferences?
The paper does not claim that such a weighting should exist.
Instead, it asks us to make the boundary explicit.
Because hidden machine objectives already exist in autonomous systems, refusing to formalize the question does not necessarily make the question disappear.
The uncomfortable conclusion
The most important result may actually be negative.
We still do not know whether any of these architectures would be conscious.
And the paper explicitly acknowledges that.
If consciousness fundamentally depends on a particular form of information integration, then architectural novelty alone may accomplish nothing.
The six architectures could be fascinating engineering systems while remaining completely irrelevant to phenomenal experience.
That uncertainty is not a weakness to hide.
It is the reason to build better tests.
Beyond “smart homes”
The deeper idea is that consciousness research may eventually need to stop asking:
“Which machine is conscious?”
and begin asking:
“Where is the boundary of the system that could be conscious?”
A processor?
A robot?
A house?
A neighborhood?
A distributed infrastructure?
Perhaps consciousness—if it can ever be engineered—will not necessarily emerge where we expect it to.
It may emerge from relationships between components rather than from a component itself.
That possibility changes the architecture of the question before it changes the architecture of the machine.
Final thought
We have spent decades making machines more intelligent.
Perhaps the next frontier is not simply making them smarter.
It is discovering whether intelligence, memory, valuation, embodiment, prediction, social interaction, and self-modeling can ever become something more than computation.
And if they can—
we may have been looking for the machine when we should have been looking at the system.
Paper
When the Home Feels: Beyond the Edge of Domestic Machine Consciousness
Seyyed Alireza Alhosseini Almodarresieh, 2026
Top comments (0)