'Why should a careful person pay for someone else's risk?' asked Kavya, studying a hypothetical insurance proposal. Her colleague Sameer answered with another question: 'Which risks count as someone else's before you know whether you will bear them?' She considered the reply evasive. The proposal charged different prices according to increasingly detailed estimates of individual risk. Surely that was fairer than placing unlike people in the same pool. 'It may be more actuarially differentiated,' he said. 'Whether that makes it fairer depends on which conception of fairness you think insurance should serve.'
Kavya explained her position. If two participants predictably impose different expected costs, charging them identical amounts makes one subsidise the other. Sameer agreed with the accounting but questioned the conclusion. Risk pooling deliberately redistributes resources from those who do not experience a loss to those who do. The disagreement concerned which differences should be priced in before outcomes occur. 'We can distinguish sharing uncertainty from sharing known disadvantage,' he said. 'But once prediction becomes more detailed, the boundary between those categories moves. Better information changes the institution's moral problem as well as its calculation.'
'Would you rather the insurer remain ignorant?' she asked. He distinguished possessing information from being permitted to use every piece of it for pricing. Institutions routinely decide which distinctions are relevant to a purpose. A rule limiting classification need not claim that the underlying differences do not exist. It can express a commitment to broader solidarity. Kavya objected that suppressing information could destabilise the pool. 'Yes,' Sameer replied. 'A norm must confront the incentives it creates. But an incentive problem does not disappear by renaming the most finely divided price the uniquely natural one.'
He described adverse selection in a simplified voluntary market. If lower-risk participants regard a pooled price as too high and leave, the remaining pool becomes more expensive to cover. Prices may rise, encouraging further departures. Kavya said this proved her point. Sameer thought it proved a narrower one: pooling rules must be considered alongside participation rules, subsidies, and the distribution of information. A voluntary arrangement with unrestricted exit behaves differently from a broadly inclusive arrangement. 'The mechanism tells us why one design may unravel. It does not identify the only legitimate social objective or every possible design capable of sustaining it.'
Kavya then raised moral hazard: protection against loss could weaken incentives to avoid it. Sameer accepted that insurance could influence behaviour, but resisted treating the technical term as an accusation of bad character. Some responses reflected deliberate carelessness; others reflected reduced financial barriers to useful activity. Their significance depended on the service and the contract. 'We should specify the behaviour, its cause, and its consequences,' he said. 'Otherwise the phrase becomes a moral label applied to any increase in use, even when increased access is part of what the arrangement is meant to achieve.'
'You keep widening the question,' Kavya said. 'A company still has to set a price.' Sameer agreed. Expected costs, administrative expenses, uncertainty, and the conditions of competition mattered to the business. But the company's viable price and society's preferred distribution of access were different questions. Public arrangements might need to address the gap. 'If we ask one commercial contract to deliver every social objective without supporting it, we may create an impossible demand. If we let commercial feasibility determine every entitlement, we may exclude precisely those people whose needs are greatest.'
They considered a participant whose elevated risk arose from an inherited condition rather than a voluntary choice. Kavya acknowledged the intuitive case for solidarity but worried about drawing a clean line between chosen and unchosen risks. Occupation, housing, environment, habits, and resources interacted. A behaviour that appeared freely chosen might reflect limited alternatives. Sameer agreed that responsibility was difficult to partition. 'That difficulty should make us cautious about presenting risk scores as measures of desert. A model estimates a probability under its assumptions. It does not establish how much of a person's exposure they morally deserve to bear.'
Kavya pointed out that uniform pricing could conceal inequalities too. If the pooled product remained expensive, poorer participants might subsidise others while still struggling to obtain adequate protection. Sameer accepted the objection. Solidarity was not established by a single equal price. Contributions could instead reflect capacity to pay, with funding mechanisms designed explicitly for redistribution. Each arrangement introduced administrative and political choices. 'The important thing is to name the subsidy and defend it,' he said, 'rather than pretend either that no redistribution occurs or that every redistribution is unjust simply because it can be identified.'
What, then, did greater predictive accuracy achieve? Kavya said it reduced uncertainty and made costs more transparent. Sameer agreed that this could improve planning and prevention. But increasingly precise classification might also make high-risk people predictably unaffordable to insure on narrow commercial terms. The institution could become more accurate at identifying those it was least able to include. 'That is not a defect in arithmetic,' he said. 'It is a tension between risk segmentation and protection. The model can expose the tension; it cannot select our response merely by becoming more accurate.'
They discussed prevention. If information identified hazards that could be reduced, the pool might benefit from financing changes before losses occurred. Yet prevention required attention to who controlled the relevant conditions. An individual could not always alter a dangerous workplace or an unreliable building. Rewarding only personal compliance might ignore structural sources of risk. Kavya observed that the insurer could not repair every social institution. Sameer agreed again. The limit did not make the wider causes irrelevant; it showed why insurance should be understood as one institution within a larger arrangement of regulation, infrastructure, and public responsibility.
By the end, Kavya no longer thought identical prices were necessarily fair, but she also doubted that the most individualised price settled the question. Different arrangements balanced inclusion, incentives, financial viability, and responsibility differently. Sameer warned against interpreting this plurality as permission to avoid decisions. A policy still required explicit commitments and evidence about likely effects. 'We should ask who remains covered, who can leave, which behaviour changes, and how losses are funded. Fairness becomes more serious when we specify what the rule does to the pool rather than merely describing the rule as equal or personalised.'
The proposal's title promised perfect fairness through perfect knowledge. Kavya crossed out the word perfect. Information could improve a design while destabilising the solidarity that had made the design useful. A pool existed because some uncertainties and burdens were shared; dividing it ever more finely could eventually remove the sharing without removing the vulnerability. The task was not to preserve ignorance or prohibit differentiation. It was to decide which risks should be individualised and which should remain collective, then build an arrangement capable of honouring that decision. The calculation could inform the choice, but the choice would still belong to people.