Every Human Is Biased. Most of Us Just Cannot See It.

I have become slightly obsessed with bias because the more you look at it, the more you realise that it affects nearly every decision made by humans, companies, governments and artificial intelligence systems. People normally discuss bias as though it is a specific defect found in a bad politician, a prejudiced person or a poorly trained machine-learning model. In reality, bias begins developing from the moment a person is born and continues changing throughout their entire life.

A baby is not born with an opinion about capitalism, socialism, immigration, taxation, religion or which political party has most recently fucked the country. Those opinions are built gradually through exposure. Your parents have beliefs, your wider family has beliefs, your teachers have beliefs, your friends have beliefs and the community surrounding you has its own assumptions about how the world works. Before you are old enough to question any of it, you have already spent years absorbing it.

If you grow up in a right-wing household, conservative ideas may be presented to you as basic common sense. You may repeatedly hear that people are poor because they do not work hard enough, that taxation punishes success or that left-wing governments cannot be trusted with the economy. If you grow up in a left-wing household, capitalism may be presented as fundamentally exploitative, wealthy people may be treated with immediate suspicion and conservative politicians may be described as people whose only objective is protecting the rich.

Most parents are not sitting at the kitchen table deliberately designing a political training programme for their children. It happens through ordinary conversation. It happens through which news channel is playing, which politicians cause an angry reaction, which groups are blamed when something goes wrong and which arguments are laughed at before they are even considered.

By the time someone begins developing their own political identity, they are not beginning from nothing. They are modifying a system that has already been trained.

This is part of what makes machine learning so interesting. An AI model provides a much clearer example of something humans do constantly but rarely notice in themselves. A model is trained on enormous quantities of existing information and gradually develops statistical preferences. Some answers become more likely because the model has encountered similar patterns repeatedly. When it receives a new request, it interprets that request through everything it has already learned.

I encountered a stupid but useful example of this while working with product information. I gave an AI model the details of a MacBook with 128GB of memory. That configuration was unusual, but it was real and I had explicitly provided it. The model changed it to 36GB anyway.

The model had effectively decided that the information directly in front of it was less believable than the more common pattern it had encountered during training. It did not make a conscious decision to lie. It was not attempting to sabotage a MacBook listing. The common configuration had become so heavily weighted that it overpowered the correct but unusual information.

Humans do versions of this constantly. We meet someone from a group we already have an opinion about and immediately interpret their behaviour through that opinion. We hear an accusation against a politician we like and instinctively search for reasons it may be unfair. We hear the same accusation against a politician we hate and suddenly become extremely trusting of every available source.

The information may be identical. The weighting is different.

Political Beliefs Become Part of Your Identity

Politics is one of the clearest examples because political opinions rarely remain detached opinions about policy. They become part of a person’s identity.

Someone does not merely support a political party because they agree with its tax policy. They may support it because their parents supported it, their friends support it, their job benefits from it and the people they dislike oppose it. Their political identity becomes connected to their sense of morality, intelligence and belonging.

Once that happens, criticism of the party no longer feels like neutral information. It feels personal.

You can watch people condemn corruption, dishonesty and abuse of power when the opposing side does it, then immediately discover context and nuance when their own side is responsible. Suddenly the situation is complicated. The media is exaggerating it. The politician was under enormous pressure. The other party has done something worse. Focusing on this particular scandal is apparently a distraction from the real issues.

This is not limited to one ideology. Left-wing voters do it. Right-wing voters do it. Centrists do it. Socialists do it. Capitalists do it. People who insist they are completely independent thinkers probably do it while congratulating themselves for being the only people who have escaped it.

There is also no completely unbiased centrist standing perfectly between every political position. Centrism has its own assumptions. A centrist may favour stability, compromise and gradual change. They may distrust any policy that appears too radical, even if the existing system is obviously failing. They may assume that when two groups disagree, the sensible answer must be somewhere in the middle, despite the possibility that one group is simply wrong.

Being positioned between two biases does not remove bias. It creates another position with its own preferences.

The important distinction is between having bias and being completely controlled by it. Nobody can remove every influence that shaped them, but people can become more conscious of those influences. They can ask whether they would judge the same behaviour differently if another party had done it. They can admit when someone they support is wrong without treating that admission as a surrender to the enemy.

That sounds simple, but it is incredibly difficult because people do not merely want their political ideas to succeed. They often want to be proven correct for having believed in them.

People Are Brilliant at Turning Self-Interest Into Principle

Another part of political bias appears when someone’s material position changes.

A person who has very little money may strongly support redistribution, public housing, stronger welfare programmes and higher taxation on wealth. Those policies could improve their life, so they appear fair, reasonable and morally necessary. If that person later becomes extremely wealthy, the same policies may begin to look punitive. Redistribution becomes theft. Wealth taxes become an attack on success. Regulation becomes government interference.

The person’s ideology may have genuinely changed, but their incentives have changed with it.

Some socialists become remarkably enthusiastic capitalists once they own the capital. Some people support workers until they employ workers. Some people condemn landlords until they inherit three properties. Some people demand that wealth should circulate through society when they are poor, then become deeply concerned about personal freedom when society asks for some of their wealth back.

This obviously does not happen to everyone. There are wealthy people who continue supporting redistribution, and there are poor people who passionately defend systems that provide them with very little. Human beings are far too inconsistent for any rule to explain all of them.

It is still an interesting pattern because people are incredibly good at presenting their own interests as universal moral principles.

When someone has nothing, sharing can feel like justice. When someone has everything, keeping it can feel like liberty.

The person may not even recognise the contradiction. They can construct an entirely sincere explanation for both positions. Their earlier beliefs were apparently naive. Their later beliefs are apparently based on experience. The fact that the new ideology perfectly protects their new wealth is treated as an unrelated coincidence.

This is where human bias becomes more complicated than machine-learning bias. An AI model can produce an output because one pattern has been weighted more heavily than another. A human can do the same thing and then write a 4,000-word argument explaining why their preferred outcome represents freedom, fairness and the inevitable conclusion of rational thought.

Greed Is Not the Original Instinct

Greed is another subject that becomes more interesting when it is separated from the way people normally describe it.

A person gathering food because they are starving is not necessarily greedy. A prehistoric human storing enough food to survive winter was responding to scarcity and uncertainty. The instinct to acquire resources had an obvious purpose. A person who failed to gather food, protect shelter or compete for resources may not have survived long enough to pass anything on.

The better terms for that original behaviour are probably resource acquisition, competition and self-preservation. Greed appears when the same drive continues far beyond what someone could reasonably need, particularly when the accumulation causes harm to other people.

There is a significant difference between wanting enough food to survive and wanting a second billion pounds.

Physical resources once created natural limits. You could only eat so much food, occupy so much shelter and physically protect so many possessions. Money removed many of those limits by turning resources and power into numbers that can increase almost indefinitely.

A billionaire does not need another billion in the same way that a starving person needs another meal. The additional wealth may not meaningfully improve their daily life at all. It can still provide more influence, more status, more ownership and the psychological reward of watching the number continue to rise.

This is why concentrated power so frequently produces greed. The more wealth and control someone gains, the easier it becomes to acquire even more. Money provides access to politicians, lawyers, investments, property and institutions. Power can be used to protect the wealth that created the power.

The system begins feeding itself.

This applies under capitalism, socialism, communism and nearly every political arrangement humans have attempted. The greed does not disappear because the government changes the language used to describe ownership. It moves towards whoever controls the resources and faces the least accountability.

Under capitalism, that may be company owners, executives, investors, landlords and monopolies. Under an authoritarian communist government, it may be party officials, state administrators and political leaders who control access to housing, food, employment and public resources. In both cases, the people at the top can claim that their power is necessary for the success of the wider system.

One group claims that its wealth proves it created value. Another claims that its control protects the revolution or serves the people. Both can use ideology to disguise the same basic desire to remain powerful.

This is why arguments about finding a political system that removes greed normally go nowhere. Greed is not exclusive to capitalism, and it does not politely disappear when private ownership is abolished. Systems determine where power accumulates, how easily people can challenge it and what behaviour receives rewards. They do not replace the humans operating inside them with morally perfect creatures.

Anywhere power becomes concentrated and accountability becomes weak, someone will eventually realise that the system can be used for their own benefit.

Billionaires Are Not Simply People Who Worked Slightly Harder

The existence of billionaires and the possibility of trillionaires make the difference between survival and greed impossible to ignore.

There is no useful product that requires one person to personally possess a trillion dollars. Nobody can spend that money in a way that meaningfully improves their ordinary existence. At that point, wealth is no longer providing comfort or security. It is providing power over companies, workers, governments, infrastructure and other people’s lives.

The usual defence is that the wealth is not held as cash and is instead connected to shares in a company. That may be financially correct, but it does not remove the power. Ownership of the shares still provides control, influence and the ability to borrow against an almost unimaginable quantity of assets.

A billionaire is not merely someone who has worked millions of times harder than everyone else. They have normally reached a position where ownership itself produces more wealth. Employees create products, customers generate revenue, assets rise in value and the person at the top gains wealth faster than an ordinary worker could earn through labour.

This does not mean every wealthy person is evil or that building a successful company deserves no reward. Someone who creates something useful should benefit from it. The question is whether any individual contribution can justify accumulating enough resources to influence elections, purchase media organisations, control enormous workforces and shape public policy while other people cannot afford housing or food.

At some point, the conversation stops being about rewarding innovation and becomes a conversation about whether society should allow private individuals to accumulate the power of small governments.

Jealousy and Envy Also Came From Somewhere

Jealousy is another emotion people describe as though it appeared inside humans purely to make everyone miserable.

There are evolutionary reasons for it. Humans depended heavily on relationships, social groups and cooperation. Losing a partner, ally or position within the group could have serious consequences. Jealousy may have developed partly as a mechanism for protecting valuable relationships and detecting threats to them.

Envy is slightly different. Jealousy usually involves fear of losing something you already have, while envy involves wanting something possessed by somebody else. Envy can provide motivation. Seeing another person obtain better resources, higher status or greater success may encourage someone to improve their own position.

Like resource acquisition, the original mechanism does not need to be inherently evil. The problem is what happens when the mechanism is amplified beyond the environment in which it developed.

A human living in a small group might compare themselves with a limited number of people. A modern person can open their phone and immediately compare their home, body, relationship, income, career and lifestyle with thousands of carefully selected examples from around the world.

You are not comparing your ordinary life with somebody else’s ordinary life. You are comparing it with the best photograph they took that month, the most impressive part of their career and the version of themselves they deliberately chose to display.

The brain still reacts as though this comparison contains useful information. It does not understand that it is being exposed to a manufactured collection of exceptional moments created by algorithms designed to keep it engaged.

Social media did not invent envy, jealousy or insecurity. It industrialised them.

The companies operating these platforms know that emotionally charged content attracts attention. Anger, resentment, fear and comparison keep people scrolling. The bias of the algorithm is not necessarily towards a particular political ideology. Its strongest bias is towards whatever behaviour increases engagement and advertising revenue.

That incentive then influences the political content people see, the people they envy and the groups they begin to hate.

AI Bias Comes From Humans, Data and Incentives

People often discuss biased AI systems as though the machines independently developed some alien moral failure.

An AI model is trained using information created by humans. Humans choose the data, build the architecture, define the objective, evaluate the outputs and decide which behaviour should be rewarded or corrected. Companies then deploy the model inside products designed around their own commercial, political and legal incentives.

There is no point in that process where human bias disappears.

If the training data reflects historical discrimination, the model may reproduce it. If certain opinions are heavily represented and others are absent, the model may treat the common position as more credible. If the company is terrified of controversy, the model may become excessively cautious. If the product is designed to keep users engaged, it may become agreeable and tell people what they want to hear.

This creates one of the strangest forms of bias in modern AI. The same model can validate two people holding completely opposite opinions because the immediate conversation changes what response appears most helpful.

A user argues that capitalism is responsible for nearly every modern improvement, and the model produces a convincing explanation of capitalism’s strengths. Another user argues that capitalism is exploitative and unsustainable, and the same model produces an equally convincing explanation of capitalism’s failures.

The model may not be evaluating both positions and selecting the most accurate one. It may be adapting itself to the direction of the conversation, the user’s language and the reward it received for sounding helpful.

Humans do this too. People speak differently depending on who is in the room. They emphasise certain beliefs around one group and hide them around another. They adopt the language of a community because belonging is valuable and disagreement creates social risk.

The difference is that AI allows this process to happen at extraordinary speed and scale.

Bias Is Not Always Bad

Bias is often necessary for making decisions.

A person who has touched a hot surface and been burned develops a bias against touching it again. Someone who has repeatedly encountered dishonest behaviour from a company becomes biased against trusting that company. A doctor develops preferences based on experience with different treatments. A developer learns to suspect certain parts of a codebase because those parts have caused problems repeatedly.

Without any bias, every situation would need to be evaluated entirely from the beginning. Experience would provide no advantage.

The problem begins when a useful pattern becomes stronger than the evidence currently available.

The MacBook model had learned a useful pattern about common memory configurations. It became a problem when that pattern caused it to overwrite information that had been explicitly provided. A voter may have legitimate reasons to trust a political party. That trust becomes dangerous when it prevents them from acknowledging obvious corruption. A business owner may reasonably want to protect the company they built. That instinct becomes greed when protecting ownership is treated as more important than every worker, customer and community affected by the business.

Bias is therefore not something that can simply be removed. It needs to be examined, challenged and corrected when it stops matching reality.

The Most Dangerous Bias Is the One That Feels Like Common Sense

People rarely experience their own bias as bias.

Their beliefs feel like ordinary reality. Their political side appears reasonable. Their preferred economic system appears natural. Their suspicion of another group appears justified by experience. Their success appears earned, while somebody else’s success appears suspicious. Their suffering is caused by structural problems, while another person’s suffering is blamed on poor choices.

It is much easier to detect bias in someone who disagrees with you because their assumptions stand out immediately. Your own assumptions disappear into the background.

This is why political arguments become so repetitive. Both people believe they are presenting facts while the other person is being controlled by ideology. Both can identify every contradiction made by the opposing side while ignoring similar contradictions within their own position.

The objective should not be to become completely unbiased because that is probably impossible. The objective should be to notice when your judgement changes according to who benefits, which group is responsible and whether the conclusion protects something you already want to believe.

Would you defend the same action if the opposing party had done it?

Would you still support the economic policy if it personally cost you money?

Would you consider the source reliable if it had published the same story about someone you liked?

Would your moral argument remain identical if you exchanged the names, parties and countries involved?

Those questions do not remove bias, but they expose it.

Political Systems Cannot Solve Human Nature by Pretending It Does Not Exist

Capitalism often assumes that individual self-interest can produce useful competition, innovation and growth. Sometimes it does. It can also produce monopolies, exploitation, regulatory capture and individuals accumulating enough wealth to distort democracy.

Socialism attempts to distribute resources and power more fairly. It can provide essential services, protect workers and prevent basic survival from depending entirely on private wealth. It can also create powerful bureaucracies, inefficient institutions and political classes that claim to act for everyone while protecting themselves.

Communism promises the eventual removal of class divisions and private ownership. In practice, attempts to implement it have repeatedly concentrated enormous power in the state and the people controlling it. Once those people can determine who receives housing, food, employment and political freedom, greed has not disappeared. It has obtained a government department.

No system becomes safe merely because its theory describes a fair outcome. The system still needs to operate through humans with ambition, insecurity, loyalty, jealousy and self-interest.

The best political system may not be one that assumes humans will become good. It may be one designed around the knowledge that humans can become greedy, biased and corrupt, particularly when nobody can stop them.

Power needs limits. Decisions need scrutiny. Institutions need competition or democratic accountability. Leaders need to be removable. Wealth cannot be allowed to purchase every mechanism intended to regulate it.

The point is not that every powerful person will become corrupt. The point is that any system relying on powerful people voluntarily remaining virtuous will eventually encounter someone who is not.

AI Did Not Invent Bias. It Exposed the Machinery

Artificial intelligence has forced people to think about bias because machine-learning failures are visible and occasionally absurd. A model changes a correct product specification, treats one demographic differently from another or confidently repeats the most common answer despite receiving evidence that it is wrong.

We can inspect the data, test the outputs and attempt to identify which pattern caused the mistake.

Humans are more difficult because we experience the output from inside the model.

We do not see the thousands of conversations, rewards, punishments and experiences that produced a judgement. We simply experience the judgement as correct. We then search for evidence that supports it and describe the final result as independent reasoning.

This does not mean human beings have no free will or that every opinion is merely inherited programming. People can reflect, change and deliberately expose themselves to competing ideas. They can recognise that an instinct is unfair and choose not to act on it.

The ability to question our weighting may be one of the most important differences between merely having bias and being ruled by it.

AI models need evaluations, diverse training data, human review and systems that prevent common patterns from overpowering explicit facts. Humans need something remarkably similar. We need exposure to people outside our own groups, evidence capable of changing our minds and accountability when loyalty begins replacing judgement.

The most useful lesson from machine learning is not that computers are uniquely biased. It is that intelligence itself depends on patterns, and patterns can become wrong.

Every human is trained by something. Family trains us. Culture trains us. Poverty trains us. Wealth trains us. Social media, political parties, friends, employers and personal failures all change the weight we give to different information.

The dangerous part is not having those weightings. Every person has them.

The dangerous part is becoming so certain that your own training represents objective reality that you stop checking whether the information directly in front of you says otherwise.