People are agents, clearly. So are corporations and governments, insofar as they pursue goals (like 'maximizing shareholder value' or 'defending territory'). Even a plant can be said to have agency, since it 'wants' to grow toward the sun. Not all agents need to be selfish — e.g., a non-profit — but any system that can be called selfish (like a neuron) will necessarily be an agent.

In reality the brain is a tangled mess of agents operating on many different levels, often simultaneously; in Hofstadter's phrase, it's a heterarchy rather than a hierarchy.

What I'm going to argue today is that agency is a fundamental property of the brain. Not only is agency the function of the brain — and thus it's very reason for existence — but it's also built into the brain's fabric and architecture. Because even neurons have agency, in the form of (metabolic) selfishness, higher-order brain systems don't need to create agency 'from scratch' out of mindless robotic slaves. They inherit agency pretty much for free.

The brain is thus uniquely hospitable to agents, who can be said to take root and grow in the brain quite readily.

Similarly, agents grow best on a substrate rich in agency. Computers, though technically capable of supporting agency, aren't particularly hospitable to it. The brain, in contrast, is already teeming with agency (in the form of billions of selfish neurons), and is thus uniquely fertile.

Modular views of the mind date back to the early days of AI. In 1959, Oliver Selfridge proposed a 'Pandemonium' architecture for an AI system, full of little independent 'demons' that had simple, goal-oriented jobs. Dennett also refers to his modules as 'demons' in Consciousness Explained, following Selfridge. In The Society of Mind, Minsky refers to them simply as 'agents.'

At the level above simple modules, but below the self, are poised what I will call sub-personal agents. These are systems like drives or instincts — hunger, lust, curiosity, greed, Addictions — that have agency recognizable even to lay-people. We don't need neuroscience to reason about these agents because we can 'feel' them, through introspection, pulling at our psyches — faintly or insistently, gently or violently. And indeed, people have been reasoning about these systems, as agents, for thousands of years.

Sub-personal agents also have immense explanatory power. This is most visible in the life of an addict. The addict 'himself' often doesn't want to keep up the Addiction, but he keeps doing it anyway. Thus the addict is often described, even by himself, as powerless, and perhaps the best, most parsimonious explanation for his behavior is that there's literally another agent inside his brain — his inner addict — realized as a particular cabal of neurons and modules.

When you take an addictive drug for the first time — nicotine, let's say — a new agent begins to bud around that source of pleasure (i.e., the neurotransmitters that flood your brain while smoking). The agent starts out small and weak. But the more you feed it, the bigger it grows, until there are many neurons, many modules, and even other brain-agents under its influence, feeding off the nicotine and craving it in ever larger doses, co-opting your planning and reasoning skills so it can scheme about how to get more of it.

It's this same sense — abstract, non-explicit — in which these agents engage in 'reasoning,' 'negotiation,' 'bargaining,' joining 'alliances,' and other forms of coalitional politics. When two sub-personal agents are bargaining, for example, they're not using words to do it, but the process is nevertheless the kind of thing that can be put into words — and thus these agents can be very 'persuasive'.

Obsessions, compulsions, Addictions, and other "inner demons" aren't the only agents with real power to control and explain our behavior; our brains are host to 'benevolent' agents as well. Our Consciences, for example. These are agents that live inside our brains, who are being trained throughout our lives, but especially in childhood, by our interactions with parents, authority figures, and other moral teachers, and by various rewards and (especially) punishments.

It's an interesting feature of our brains that society (or perhaps "elite society") can install these types of agents — God, the Conscience, a sense of morality — to look after its own interests. This is reminiscent of the way the UN will install weapons inspectors or election observers inside otherwise-sovereign nations.

In this view, the self is a social agent. It's both externally- and internally-facing, its role as much public relations as executive control.

If we accept that the brain is teeming with agency, and thus uniquely hospitable to it, then we can model the self as something that emerges naturally in the course of the brain's interactions with the world.

In other words, the self may be less of a feature of our brains (planned or designed by our genes), and more of a growth.

Most normal human brains, in normal environments, will naturally grow a single agent (the self) at the top of their agent hierarchy. But what if the brain or the environment isn't quite normal? Are we capable of growing other selves or Person-like agents — even multiple ones in a single brain?

Joan of Arc heard a few different voices which helped her to "govern" herself. "Whatever I have done that was good," she said, "I have done at the bidding of my voices."

But what's common to all of these phenomena is that they seem to involve separate entities — agents who aren't wholly 'us' — living inside our brains. God knows, they may even be sentient. There's certainly nothing in principle that would prevent a brain from hosting two separate sentient creatures. And while I can't say for sure that it's true, the mere possibility of it should give us pause.

It turns out there's a community — on the internet (where else?) — trying to intentionally cultivate these kinds of agents in their brains.

Exorcism is a form of psychological therapy in which the disease is treated as an agent.

In other words, an exorcist is a healer who takes the intentional stance toward a person's inner demons. Instead of looking for a medicinal cure (physical stance), and instead of addressing the patient's 'self' (psychological stance), the exorcist addresses the patient's ailment directly. This could entail any number of things: negotiating with it, reasoning with it, bribing it, showing it love and compassion, making it swear an oath, threatening it, or commanding it in the name of a higher power.