Start with what you have observed
Set idealism aside and describe people as you have actually found them.
Most people who do this honestly produce a list along these lines: self-interested by default. Bad at weighing long-term costs against short-term gains. Inclined to believe whatever is convenient. Prone to abusing power once they have it, often while sincerely believing they are not. Better at noticing other people's bias than their own.
This is not cynicism. It is the ordinary result of paying attention. And if it is roughly right, it is the most important fact any political system has to be designed around.
Now hand that list a set of tools
Take the people you just described and give them, specifically: the legal authority to take anyone's money or property; the power to confine people who disobey; a monopoly on the legitimate use of violence; the ability to write the rules everyone else must follow; and legal immunity for most of what they do with any of it.
You do not need a philosophy to predict what follows. Your own description of human beings already predicts it, and the record is not ambiguous: capture of regulators by the industries they regulate, civil forfeiture, budgets that only ever grow, wars that outlast their justification, and a durable habit of writing rules that exempt the rule-writers.
These are not stories about unusually bad individuals. They are what the incentive structure produces when it is fed ordinary ones.
The contradiction
The case for concentrating coercive power requires holding two beliefs simultaneously.
The first is the one you just described: people are self-interested, short-sighted, and dangerous with power.
The second is that these same people, once elected or appointed, will restrain themselves, resist the incentives in front of them, and use coercive authority on behalf of strangers rather than themselves.
Politicians are not drawn from a different species. They are drawn from the population you just described, and they take every trait on your list through the door with them.
What changes on the far side of that door is not the character of the people. It is the incentive structure — one that now rewards those traits with other people's money, insulates them from the consequences of being wrong, and makes accountability slow, indirect, and rare.
Your observations about human nature do not pause at the entrance to a government building. If people cannot be trusted to run their own lives through voluntary agreement, they are a strange choice to run everyone else's through legal force.
Which structure limits the damage
The realist question is not "are people good?" It is "what does each structure do with people who are not?"
Coercive systems reward political maneuvering, push costs onto whoever is least able to resist, and concentrate authority in the hands of whoever wins the contest to hold it. Once the stakes are high enough, that contest becomes permanent, which is a fair description of the last several decades.
Voluntary systems reward people for serving others well enough that others keep choosing them. Costs stay closer to whoever created them. Authority is spread across an enormous number of small decisions rather than a few large ones.
The decisive difference is in the reach of a bad actor. Someone dishonest in a voluntary system can hurt the people who chose to deal with them, which is a real harm with a real limit. Someone dishonest who has captured the machinery of legal force can hurt millions who never chose anything.
It would be dishonest to stop there without saying that voluntary systems have failure modes of their own. Free riding is real. Information asymmetry is real. Some coordination problems are genuinely hard to solve without a rule that binds everyone. The philosophy does not claim these vanish. It claims that a structure where bad actors are limited by who is willing to associate with them fails more gracefully than one where bad actors can seize an apparatus built for compulsion and aim it wherever they like.