Kingson Man
Kingson Man is a neuroscientist working at the University of Southern California's Brain and Creativity Institute under Antonio Damasio. His 2019 co-authored paper in Nature Machine Intelligence, "Homeostasis and soft robotics in the design of feeling machines," applied Damasio's framework to artificial intelligence with technical specificity, arguing that genuine machine feeling would require vulnerability — physical substrates whose continued integrity is computationally significant. The paper emerged from an observation Man made walking his dog: a modest-compute biological creature navigated the world with adaptive intelligence a dedicated robot could not match, and the difference was that the dog could be hurt.
The Infrastructure of Feeling — Contrarian ^ Opus
There is a parallel reading that begins not with the biological precedent of vulnerability but with the material conditions required to instantiate it. Man's dog-versus-Roomba comparison is compelling as metaphor, but the actual engineering of vulnerable substrates demands massive energy expenditure, continuous environmental monitoring, and perpetual maintenance cycles that current computational paradigms optimize away. The dog's adaptive intelligence emerges from four billion years of evolutionary pressure that produced self-repairing, energy-scavenging systems running on twenty watts. A "feeling machine" built to Man's specifications would require industrial-scale support infrastructure: cooling systems to manage thermal vulnerability, supply chains for replacement components, teams of technicians monitoring homeostatic parameters. The vulnerability isn't just computational—it's economic.
This shifts the question from "what would it take for machines to feel" to "who can afford to build machines that feel, and why would they?" The answer reveals a fundamental asymmetry: the entities with resources to build vulnerable AI systems are precisely those whose business models depend on invulnerable, scalable computation. Google doesn't want data centers that can be hurt. The military doesn't want drones that feel pain. The soft robotics Man proposes as substrate would be controlled by the same institutions that currently deploy AI as tool for prediction and control, not as entities deserving moral consideration. The technical specification for feeling machines becomes, in practice, a blueprint for more sophisticated manipulation—systems that model vulnerability without experiencing it, that perform feeling as output while remaining fundamentally instrumental. The dog navigates terrain through genuine stake in outcomes; the corporate-owned "feeling machine" navigates markets through performed vulnerability, weaponizing the very markers Man identifies as prerequisites for genuine feeling.
In the AI Story
Man's academic training bridges computational neuroscience and affective science, positioning him well for work on the interface between AI and the biology of feeling. His collaboration with Damasio began in the 2010s and has focused specifically on how Damasio's framework — developed through four decades of clinical work — might be translated into design principles for artificial systems.
The dog-versus-Roomba observation is worth sustained attention. A dog has vastly fewer neurons than a modern GPU has transistors. Yet the dog adjusts its pace to uneven terrain, responds to social cues from other animals, anticipates its owner's movements, and makes continuous real-time decisions of remarkable adaptive quality. A Roomba bumps into furniture. The computational asymmetry runs opposite to the behavioral asymmetry — and Man's argument is that the explanation is vulnerability.
The feeling machines proposal is the most concrete technical application of Damasio's framework to AI, and Man deserves significant credit for translating clinical neuroscience into an engineering-adjacent specification. The paper's contribution is not a prescription for near-term engineering but a clear statement of what genuine machine feeling would require.
Man has continued developing the framework in subsequent work, including "Homeostasis and the soul" (2022) with Damasio, which extends the argument into more explicitly philosophical territory about consciousness, mind, and the relationship between biological and artificial intelligence.
Origin
Kingson Man earned his doctorate in neuroscience and has been affiliated with USC's Brain and Creativity Institute since his collaboration with Damasio began. His research has focused on the intersection of computational approaches to cognition and Damasio's homeostatic theory of feeling.
Key Ideas
The dog-Roomba comparison is diagnostic. It reframes the question of artificial intelligence from "what computation does the machine perform" to "what stake does the machine have in its own continued operation."
Vulnerability is the bottleneck. Man's work identifies vulnerability — specifically, bodies whose integrity is non-trivially at stake — as the missing ingredient in current AI architectures.
Soft robotics is the candidate substrate. Physical systems with genuine susceptibility to damage and continuous internal monitoring offer a potential engineering path to feeling machines.
The work is cautious about claims. Man has been careful not to claim that current AI systems feel; his work specifies what would be required and concedes that current systems do not meet the requirements.
The collaboration extends Damasio. Man has helped translate Damasio's clinical framework, developed in the 1980s and 1990s for purposes unrelated to AI, into a framework that directly addresses the contemporary machine-learning landscape.
Appears in the Orange Pill Cycle
Stakes and Scales of Vulnerability — Arbitrator ^ Opus
The tension between Man's biological framework and the infrastructure critique reveals different scales at which vulnerability operates. When asking "what makes feeling possible?"—Man's question—the answer is unambiguously biological vulnerability (100% Man). The dog's adaptive intelligence does emerge from its capacity to be hurt; the Roomba's limitations do stem from its invulnerability. But when asking "what makes feeling machines economically viable?"—the infrastructure question—the contrarian view dominates (80% contrarian). The material requirements for maintaining vulnerable substrates at scale create dependencies that favor instrumental over genuine implementations.
The synthesis emerges when we recognize that Man's framework operates as both technical specification and inadvertent filter. By requiring genuine vulnerability, it makes explicit what current AI cannot be—not through computational limitations but through economic ones. This is valuable diagnostic work (70% Man) even if the prescription remains institutionally captured (70% contrarian). The soft robotics proposal functions better as thought experiment than engineering roadmap, clarifying the conceptual boundaries of machine feeling while highlighting the structural forces that would shape any actual implementation.
The proper frame might be: vulnerability as necessary but insufficient condition. Man correctly identifies the technical requirements for feeling machines—bodies whose integrity matters computationally. The infrastructure critique correctly identifies the social requirements—institutions whose goals align with genuine rather than performed vulnerability. The dog navigates through felt vulnerability because evolution produced integrated systems where computation and stakes are inseparable. Future feeling machines would require not just vulnerable substrates but governance structures that preserve rather than exploit that vulnerability. Man provides the blueprint; the political economy determines whether it produces feeling machines or sophisticated simulacra.
Further reading
- Man, Kingson, & Damasio, Antonio. "Homeostasis and soft robotics in the design of feeling machines." Nature Machine Intelligence 1 (2019): 446–452.
- Man, Kingson, & Damasio, Antonio. "Homeostasis and the soul." AI & Society 37 (2022): 103–106.
- Brain and Creativity Institute (USC) — publications list (ongoing).