Paternalism and Mathematics

From intuition to models

For a long time classical economics assumed that a person behaves rationally and consistently. Behavioural economics showed that this is too strong an assumption: people are subject to cognitive biases, temporal inconsistency, attention errors, and context effects. This is where the mathematics of paternalism appears: if decisions systematically deviate from long-term interests, can the permissible degree of intervention be described formally?

Optimal paternalism

The idea of optimal paternalism lies not in maximal control but in finding a point of compromise between improving welfare and respecting autonomy. In models with bounded rationality the intervention can be very subtle: a change of default option, the price of a signal, the order in which information is provided. Formally such problems are described through dynamic optimisation, optimal control, and dynamic programming.

Game theory and the principal-agent model

The paternalist almost never has complete information about a person's preferences. This is precisely why principal-agent models are useful. The principal designs a contract, rule, or choice environment so that the agent voluntarily chooses the desired action. This makes it possible to analyse social programmes, incentives for healthy behaviour, transport systems, and digital interfaces as problems of strategic interaction rather than of morality alone.

Mechanism design

Mechanism design asks: how can the rules of the game be constructed so that even under self-interested behaviour by participants a socially desirable result emerges? For the theme of paternalism this matters especially where the point is not a prohibition but the design of an institution: systems of benefits, resource allocation, voting, access to services, or the rules of a digital platform.

Bayesian persuasion

In soft paternalism what proves decisive may be not the prohibition but the signal. The theory of Bayesian persuasion shows how the sender of information influences the receiver's actions through the structure of data disclosure. Applied to paternalism, this means that the very organisation of a risk report, a warning, or an interface can change decisions without direct coercion.

Fairness and constraints

Formal optimality does not guarantee fairness. An intervention that improves the average result may worsen the situation of particular groups. That is why modern models increasingly include fairness constraints: limits on the distribution of benefits, risks, and errors across different types of agents.

Why this section matters

The section "Paternalism and Mathematics" sets Paternus apart from an ordinary encyclopaedia. It shows that the dispute about freedom and care can be discussed in the language of models, computational complexity, optimisation, and the structure of information. This is especially valuable for researchers in public policy, economics, AI, and digital design.

Below are formulations from works on Bayesian persuasion and dynamic programming.