Racism as Civilisational Compute Wastage: When Human Bias Becomes a Systems Bug

Racism is first a violation of human dignity. No economic argument is required to establish that truth. But systems language reveals another layer of the damage: racism blocks information, talent, trust and opportunity that civilisation needs in order to understand and improve itself.

Racism is therefore not only cruelty toward particular humans. It is civilisational compute wastage; in system language it is a bug.

The metaphor must be handled carefully. Human beings are not processors, and their worth is not measured by productivity. A person deserves equality even if a market never assigns value to their work. “Compute wastage” describes the failure of the social system—not the value of the human being it excludes.

What the systems bug does

A healthy learning system receives diverse information, tests assumptions and updates when evidence changes. Racism does the opposite. It decides that the source of an idea matters more than the idea’s accuracy. It restricts access before capability can be observed. It converts inherited bias into institutional routine.

• Input loss: experiences and observations are ignored because of who reports them.
• Routing failure: qualified people are diverted away from education, employment, funding or leadership.
• Feedback corruption: biased outcomes are treated as proof of biased assumptions.
• Model collapse: institutions learn from a narrow portion of humanity and mistake that narrowness for universality.
• Trust degradation: people withdraw from systems that repeatedly misuse, misread or exclude them.

Misallocation becomes self-reinforcing

Imagine an employer who overlooks strong candidates because a name, accent, postcode or skin colour triggers an assumption. The immediate harm falls on the candidate. The organisation also loses skill, perspective and possible innovation. When thousands of institutions repeat the same behaviour, the loss becomes civilisational.

The World Bank describes social exclusion as both a moral and economic problem, with costs visible in lost earnings, education, health, gross domestic product and human-capital wealth. The OECD similarly warns that discrimination can force people into roles below their skill level and misallocate human and economic capital.

These findings do not mean anti-racism is justified only when it increases output. They show that injustice produces secondary damage across the systems that practise it.

Bias can hide inside apparently neutral processes

A system does not need to contain an explicit racist instruction to reproduce racial inequality. Recruitment criteria may reward networks built through historic exclusion. Credit models may learn from unequal lending. Facial-recognition systems may perform differently across groups. “Culture fit” may become a vague permission to prefer familiarity.

In AI, historical data can turn yesterday’s discrimination into tomorrow’s automated score. Removing a protected characteristic from a dataset does not necessarily remove the proxies around it. Responsible design therefore requires testing outcomes, documenting limitations and giving people meaningful routes to challenge decisions.

Repair requires more than better language

1. Measure access and outcomes. Good intentions cannot reveal who is repeatedly filtered out.
2. Inspect decision points. Identify where discretion, proxies or networks create unequal routes.
3. Correct the data. Document missing groups, measurement bias and historical distortion.
4. Redistribute authority. Inclusion without decision-making power can become decoration.
5. Create appeals and accountability. A person harmed by a decision needs a human route to correction.
6. Preserve dignity. Do not reduce inclusion to extracting more productivity from marginalised people.

The psychology of wasted attention

Racism also consumes cognitive resources. Targets may spend time anticipating prejudice, decoding ambiguity, proving legitimacy and recovering from hostility. Bystanders learn to remain silent. Perpetrators defend an identity built on false hierarchy. Institutions spend energy managing the consequences of a problem they refuse to name.

That is wasted attention at every level. The alternative is not forced sameness. It is a society capable of processing difference without converting difference into rank.

From bug report to redesign

Calling racism a bug should not imply it is accidental or easy to patch. Some discriminatory systems were deliberately built, and others are maintained because they benefit people with power. A proper bug report therefore includes history, incentives, ownership and the conditions under which the failure reproduces.

The repair is cultural, legal, economic, psychological and technical. It requires truthful education, fair institutions, representative leadership, enforceable rights, responsible technology and repeated human courage.

A civilisation becomes more intelligent when it stops throwing away human possibility.

This connects with my work on human feedback as food for AI. A learning system is only as wise as the feedback it is willing to receive—and the humans it recognises as credible sources.

Sources and date note

This article reflects information available on 5 September 2026. For the evidence behind the systems argument, see the World Bank’s social inclusion overview and the OECD’s analysis of discrimination’s effects.