The Bandit
The bandit occupies the upper-left quadrant of Cipolla's matrix: actions that enrich the actor while damaging others. Bandit behavior follows the logic of self-interest, and this rationality is what makes the bandit tractable. His incentives can be altered; his behavior adjusts in response. Renaissance tax collectors, whom Cipolla studied with particular attention, were frequently bandits in the precise technical sense — enriching themselves through mechanisms that impoverished their communities — but because their behavior followed the logic of extraction, it could be anticipated and constrained through oversight, rotation, and the threat of punishment. In the AI economy, the bandit deploys the technology to extract value from populations that cannot evaluate or resist the extraction.
In the AI Story
The bandit's distinguishing feature is coherent self-interest. He is not pursuing virtue, but he is pursuing something, and that something provides a handle for institutional intervention. Raise the cost of extraction above the benefit, and extraction stops. This is why regulatory frameworks — however imperfect — can constrain AI banditry: fines, liability, reputational cost change the calculation. The bandit does not need to become virtuous. He merely needs to find virtue more profitable than extraction.
The Research Society of Australia proposed in 2023 the concept of Artificial Banditry as the harmful counterpart to Artificial Intelligence — systems deployed for the benefit of their operators at the measurable expense of their subjects. Deepfake fraud, algorithmic manipulation designed to exploit cognitive biases, AI-generated content deployed at scale to capture attention and advertising revenue from users who cannot distinguish generated from human-produced material: each instance fits the bandit quadrant with precision.
The bandit's relative tractability does not mean his damage is small. In aggregate, banditry can destabilize institutions and hollow out trust. But Cipolla's framework insists on the comparison: the bandit is dangerous but manageable; the stupid actor is both more numerous and less tractable. Institutional design that addresses only banditry — through antitrust, transparency, liability — will leave untouched the larger and more damaging category of actors whose behavior is not rational extraction but incomprehending deployment.
Origin
The category derives from Cipolla's work on corruption in pre-industrial European fiscal systems, particularly his studies of tax farming in the Italian city-states. The bandit's archival profile was consistent across centuries: coherent self-interest, predictable behavior, institutional constrainability through incentive adjustment.
Key Ideas
Coherent self-interest. Bandit behavior follows rational extraction logic, which makes it predictable and therefore constrainable.
Incentive tractability. Institutional mechanisms that alter the cost-benefit calculation can redirect bandit behavior without requiring moral transformation.
Artificial banditry. The RSA's extension of the category to AI systems deployed for operator benefit at subject expense.
Less dangerous than the stupid. The bandit's rationality is a handle; the stupid actor's disconnection from self-interest is not.
Appears in the Orange Pill Cycle
Further reading
- Carlo Cipolla, The Basic Laws of Human Stupidity (Doubleday, 2011)
- Research Society of Australia, 'Artificial Banditry' (2023)