Forgiveness: repair or subsidy?

It is easier to forgive an enemy than to forgive a friend.

— William Blake

Forgiveness enters the conversation already loaded with morality. It is rarely discussed as an operation inside a relationship; more often it shows up as a demand, a virtue or an accusation: if you love, forgive; if you respect yourself, don’t; the strong know how to forgive, the weak put up with everything. One word comes to stand for giving up revenge, restoring trust, keeping the relationship, freeing yourself from anger, forgetting the damage, cancelling the consequences, and sometimes simply the exhaustion of a person who no longer has the strength to continue a conflict. The conversation quickly loses its definition.

In one situation the lie is disclosed by the offender himself, the damage is repaired, the defence of the convenient version stops, and new behaviour gradually produces data for trust. In another the person apologises, asks to leave the past in the past, and repeats the same move at the next opportunity. Both cases use the word “forgiveness,” but only the first produces repair.

Three decisions instead of one

Economists call it moral hazard when protection from consequences changes behaviour. The term requires neither malice nor inner calculation. A bank confident of a bailout may take more risk; a partner who is forgiven with no new conditions may again choose the convenient violation. The cause is the changed price of the action.

Forgiveness cannot be separated from consequences. Popular culture offers a false fork: either forgive or take revenge. Forgiveness is compatible with leaving, with changing the rules of access to money, body, home or information, with declining to restore trust, and with deciding not to continue the joint project. Forgiveness, trust and continued cooperation are three different decisions, often glued into one moral lump.

In business it is more obvious. A company can understand a supplier who blew a critical delivery, accept compensation and wish him no harm. That does not mean the contract is automatically renewed on the old terms. There will be a backup supplier, a penalty, a smaller order, or an end to the relationship. A breach does not turn a counterparty into an enemy, but it changes the risk assessment. Memory in a relationship does the same job: it is not obliged to poison the future, but it is obliged to update the forecast.

The offender often wants not so much forgiveness as an instant return to the previous level of trust. “I said sorry,” “how long are you going to keep bringing it up,” “if you forgave me, don’t go back to it” sound like a request for generosity, but frequently mean a demand to delete the data. Trust does not work like a court sentence annulled by an act of mercy. It is a forecast of future behaviour, and a forecast is restored only by a series of new observations. One honest conversation after ten years of reliability and one honest conversation after ten years of lying carry different weight.

Forgiveness in a world with noise

A repeated game rarely runs without errors. A person forgets an agreement, misreads a message, snaps after a sleepless night, or makes a move the partner takes for betrayal although the intention was different. The strategy “answer every violation with equal punishment” looks good in a clean model and can start an endless chain of retaliation in a world with noise: one random glitch triggers a countermeasure, the countermeasure is taken for a fresh violation, and cooperation falls apart with no original villain.

In models of the evolution of cooperation, the strict tit-for-tat strategy can be displaced in a noisy environment by a more generous version: it answers cooperation with cooperation, notices repeated defection, but sometimes lets a single bad move pass. Forgiveness here is not a moral reward. It is a mechanism that keeps a random error from destroying cooperation that benefits both.

The key word is single. A generous strategy tells noise from a stable type by repetition. If the violation continues, the probability that we are looking at chance drops fast. To forgive identically every time is to stop updating your model of the partner. To never forgive is to assume the world is error-free.

But a system where any mistake becomes a life sentence is not viable either. People err out of weakness, fear, shame, poor skill at defending themselves, and an inability to see the damage in advance. A confession that triggers an execution instead of a repair teaches people to hide better, not to be more honest. Healthy forgiveness has to leave a way back, but the way back runs through changed behaviour, not through a well-phrased request to forget.

Repair has to produce new data

The difference between regretting the harm and regretting that the harm was discovered becomes visible after the apology, when the boring part begins: compensation, transparency, limits, patience and repetition. A series of experiments and longitudinal studies gave this difference a clear name — the “doormat effect”: forgiveness supported self-respect when the offender showed through his actions that the injured party would again be safe and valued; without such corrections the same forgiveness was associated with lower self-respect and less clarity about oneself. A person who is genuinely repairing does not demand that trust return faster than he can produce grounds for it. He understands that his shame does not pay for someone else’s risk.

Forgiveness has to do with changing the system, not with the size of the emotion. An affair that grew in a hidden zone of a life calls for a new architecture of transparency and boundaries, not for a single vow. Regularly wrecking the budget calls for a different regime of money and responsibility. Addiction, aggression or vanishing when things get hard make forgiveness without treatment, outside help or strict conditions a way of postponing unpleasant decisions.

The injured party has incentives too. Sometimes withholding forgiveness becomes a way of holding power. The violation has been admitted, the damage compensated, the behaviour changed, but the past goes on being used as a permanent lever in every new argument. Such memory does not protect boundaries; it creates a debt pit with no procedure for repayment. Good repair needs two symmetrical things: consequences for the violation that protect against repetition, and an end point to the punishment that protects against permanent war.

The general question of whether one must forgive is too thin. Sharper markers are needed: what was violated, what damage occurred, has the person acknowledged that damage, have the conditions changed, has new data for trust appeared, who bears the cost of restoration, and what follows a repetition. Research on post-transgression trust supports exactly this order: closeness makes forgiveness easier mainly when the expectation of safety is restored after the offence, not simply because the relationship mattered before.

Without forgiveness a relationship becomes brittle: any damage may turn out to be final. Without consequences it becomes defenceless: any damage can be inflicted again. Between these extremes lies adult cooperation, where a person is not required to be error-free but is required to take part in repairing what he broke.

Main sources

Finkel, E. J., Rusbult, C. E., Kumashiro, M., & Hannon, P. A. (2002). Dealing with betrayal in close relationships: Does commitment promote forgiveness? Journal of Personality and Social Psychology, 82(6), 956–974.

Luchies, L. B., Finkel, E. J., McNulty, J. K., & Kumashiro, M. (2010). The doormat effect: When forgiving erodes self-respect and self-concept clarity. Journal of Personality and Social Psychology, 98(5), 734–749.

Strelan, P., Karremans, J. C., & Krieg, J. (2017). What determines forgiveness in close relationships? The role of post-transgression trust. British Journal of Social Psychology, 56(1), 161–180.

Nowak, M. A., & Sigmund, K. (1992). Tit for tat in heterogeneous populations. Nature, 355, 250–253.