My illustration entitled: “The Machine at the Desk” — An imposing automated clerk processes citizens one by one, issuing access tokens without explanation.
When a person is excluded by software, the most important question is not whether the decision was automated. It is whether the person can understand it, challenge it, and continue to participate somewhere else.
Economic participation is increasingly decided before a person ever speaks to another person. A software system may determine whether a worker’s profile is shown to employers, whether a merchant’s payment is accepted, whether a borrower receives credit, whether a driver is offered work, whether a seller is visible in a marketplace, or whether a customer is permitted to use a service at all.
These systems are often introduced as tools of efficiency. They can process large volumes of information, identify patterns, prevent fraud, reduce delay, and help institutions make decisions at scale. Used carefully, they can lower costs and widen access. An algorithm can spare people from arbitrary personal prejudice, detect risks that would otherwise be missed, and make routine processes more consistent.
But efficiency is not neutrality. When software becomes the practical gatekeeper of work, credit, commerce, identity, and visibility, it is no longer merely a technical tool. It becomes part of the institutional structure that determines who may participate in economic life.
This is the algorithmic permission economy: an economy in which access to ordinary opportunities is increasingly granted, ranked, restricted, or withdrawn through automated systems whose rules may be difficult to see and even harder to challenge.
The danger is not that all automation is harmful. The danger is that people may lose the ability to distinguish a legitimate assessment from an unaccountable exclusion. If a decision cannot be understood, reviewed, or escaped, then formal rights may remain intact while practical economic freedom steadily diminishes.
Permission Has Moved Into the Software Layer
In earlier economic systems, permission was often visibly exercised by a person or institution. A bank officer approved a loan. A manager selected an applicant. A shop owner chose what to stock. A clerk reviewed a form. These decisions could still be unfair, but the source of authority was identifiable. There was someone to ask, a policy to question, and sometimes a decision-maker who could reconsider.
In the algorithmic permission economy, authority is often distributed across software, data, rules, and automated thresholds. A person may receive a refusal without knowing which fact triggered it, which data source was relied upon, whether the information was correct, or whether a different human judgment would have reached another conclusion.
The result can be a peculiar form of power. No individual appears to have excluded the person. The organisation may say that “the system” made the decision. Yet the system did not design itself, choose its objectives, collect its data, or decide how much error was acceptable. Human institutions made those choices, even when the final act of exclusion is carried out automatically.
Software can make power less visible without making it less real. Indeed, opacity can increase power because it makes the affected person unsure where to begin. They may be unable to identify the relevant rule, correct a mistaken record, or distinguish a temporary error from a permanent barrier.
Political Economics has always concerned itself with the difference between nominal rights and practical liberty. The algorithmic permission economy brings that difference into sharp focus. A person may retain the legal right to work, trade, own property, or seek credit, yet encounter a chain of automated refusals that makes those rights difficult to exercise in reality.
Participation Is More Than an Account
Modern institutions frequently describe participation in simple terms: open an account, accept the terms, meet the eligibility criteria, and use the service. But economic participation is not a single event. It is an ongoing capability. It includes the ability to be discovered, trusted, paid, insured, financed, supplied, and connected to others.
When software mediates each of these functions, a person can be excluded without any formal declaration that they are excluded. Their account may remain open, but their work may no longer be recommended. Their product may technically remain listed, but it may no longer be shown. Their application may be accepted, but it may be silently ranked below every competitor. Their payment method may function in most places while repeatedly failing where it matters most.
This is why access should not be confused with meaningful participation. A door that is technically unlocked but impossible to find, reach, or pass through does not provide a practical right of entry. In the same way, an economic system that offers nominal eligibility while using hidden rules to make success unattainable creates a thin and conditional form of inclusion.
There will always be legitimate reasons to assess risk and establish standards. Not every person must be offered every opportunity, and not every transaction should be approved. The issue is whether an automated system has become so central that its decisions effectively determine who counts as economically visible, credible, or eligible.
Once that threshold is reached, the system must be judged not only by its efficiency but by its effect on human agency.
The Problem of Invisible Criteria
An automated decision can be difficult to contest for a simple reason: the person affected may not know why it occurred. The system may draw on data that is incomplete, outdated, irrelevant, or wrongly attributed. It may treat a proxy as though it were a fact. It may reward patterns that advantage those already well positioned and penalise people whose circumstances do not fit the expected model.
The affected person is then placed in a position of radical informational weakness. The institution knows the criteria, the data sources, the scoring methods, and the internal thresholds. The individual knows only the outcome.
This asymmetry has serious economic consequences. If a person cannot identify the reason for a refusal, they cannot correct an error. If they cannot know what standard is being applied, they cannot plan rationally. If they cannot distinguish between a temporary concern and a permanent classification, they may waste time and resources trying to satisfy a requirement that was never made clear.
Transparency does not require an institution to publish every line of code or reveal information that would enable fraud. It does require a person to receive meaningful reasons when an important economic decision has been made about them. “The system says no” is not a reason. It is an abdication of responsibility.
A fair system should be able to explain, in plain terms, the main basis on which a significant decision was reached. It should provide a path for correcting inaccurate information. And it should not treat a data-driven conclusion as immune from human reconsideration merely because it was produced by software.
From Judgment to Classification
Human judgment is imperfect, but it can recognise context. It can distinguish between a one-time mistake and a lasting pattern, between an unusual circumstance and an ongoing risk, between a person who needs assistance and a person who intends harm. Automated systems tend to classify. They reduce complex lives to categories, scores, flags, rankings, or predicted outcomes.
Classification has value when it is used modestly and responsibly. It becomes dangerous when the classification itself becomes destiny.
A person should not be permanently defined by a score they cannot see, a reputation they cannot carry, or a risk category they cannot challenge. The more important the consequence, the stronger the need for review. Losing access to a minor convenience is not the same as losing the ability to earn income, receive payment, obtain housing, or maintain a lawful business.
The proportionality principle matters here. Automated tools may be useful for identifying questions that require attention. They should not automatically be treated as the final authority in decisions that can seriously alter a person’s economic life. A system that triggers review is different from a system that closes opportunity without explanation.
This distinction protects both institutions and individuals. Institutions gain the benefit of efficient screening while retaining accountability for high-impact decisions. Individuals gain the assurance that they are not merely a data point trapped inside an unchallengeable classification.
My illustration “The Machine at the Desk” work-in-progress. The art represents how opaque automated systems can quietly ration people’s access to opportunity, turning citizenship into a queue for permission.
Ranking Is Also a Form of Power
Not every algorithmic decision is an outright refusal. Some are rankings. A worker is listed lower in a search result. A merchant is placed beneath preferred sellers. A creator’s work is shown to fewer people. A product is recommended less often. A borrower’s application is placed in a slower queue. These decisions may appear minor one at a time, but together they can determine who receives opportunity and who remains unseen.
Ranking systems allocate attention, and attention has economic value. In markets where discovery happens through software, visibility can be as important as legal permission. A person whose work is not surfaced may be technically present but economically absent.
This is particularly important when a platform’s ranking criteria are unstable or obscure. Participants may devote enormous effort to complying with rules that change without warning. They may be pressured to adopt particular business practices, purchase additional services, or alter their conduct simply to preserve visibility. The relationship ceases to be one of voluntary exchange between independent parties and becomes a contest for continued algorithmic favour.
Economic freedom requires more than the right to appear on a list. It requires a reasonable ability to understand the conditions under which one’s contribution can be found, evaluated, and rewarded. When ranking becomes central to livelihood, it should not be treated as a trivial feature of product design.
Accountability Cannot Be Automated Away
Institutions may be tempted to treat automated decisions as objective because they are repeatable. But repeatability is not the same as fairness. A system can apply the same flawed rule consistently. It can reproduce an inaccurate data point at scale. It can make a biased assumption appear neutral by expressing it as a calculation.
Accountability begins by recognising that every algorithm reflects human choices. Someone selected the goal the system should optimise. Someone decided which data to use. Someone defined what counts as a successful outcome. Someone accepted a level of error. Someone determined whether the system’s judgment should be reviewed by a person.
Those choices should not disappear behind the word “algorithm.” The institution that relies on software remains responsible for the consequences of its use.
There should therefore be identifiable lines of responsibility. A person affected by a serious decision should know which institution is accountable. There should be a real process for appeal, not merely an automated form that repeats the original conclusion. A review should be capable of considering new information, correcting errors, and recognising exceptional circumstances.
Human review is not valuable merely because it is human. It is valuable because it reintroduces judgment, responsibility, and the capacity to respond to reasons. A reviewer who has no authority to alter the result offers little protection. Meaningful review requires the power to correct.
The Right to Seek Another Path
No decision system will be perfect. Errors will occur, and some institutions will maintain standards that individuals dislike or cannot meet. For this reason, the right to exit remains essential. A person should not be condemned to economic exclusion because one platform, one database, or one automated classifier has made an adverse decision.
Alternatives are the practical safeguard against algorithmic overreach. If a worker can seek clients through more than one route, if a merchant can receive payment through more than one network, if a borrower can present a portable record to more than one provider, then an unfair or mistaken decision is less likely to become a complete economic sentence.
This is why interoperability, data portability, and open economic networks are not merely technical preferences. They help prevent a single algorithmic decision from becoming absolute. They allow people to carry evidence of their work, retain lawful records, and seek recognition in another system.
Counter Economics contributes a constructive principle: when participation is too heavily dependent on permission, people should be able to develop lawful, voluntary, decentralised alternatives. The goal is not to avoid accountability. It is to ensure that accountability does not become a monopoly exercised by one opaque system.
More than one path is a condition of meaningful freedom.
Designing a More Human Permission Economy
The solution is not to ban automated tools or pretend that institutions can operate at modern scale without software. The solution is to establish conditions under which automation serves human agency rather than replaces it.
First, high-impact decisions should be explainable in terms the affected person can understand. A person must be able to know the principal reason for a refusal, suspension, or severe reduction in economic access.
Second, inaccurate data must be correctable. Economic life cannot be fairly organised around records that individuals cannot inspect or dispute.
Third, important automated decisions should be reviewable by a responsible person with authority to reconsider the result. This is especially necessary where livelihoods, property, credit, payment access, or professional standing are at stake.
Fourth, systems should be designed for portability and alternatives. People should not lose their legitimate history, reputation, or productive capacity because they change providers or disagree with one institution’s terms.
Finally, automated systems should be assessed by their effect on practical freedom. Do they widen genuine access, or do they convert ordinary economic life into a sequence of unanswerable permissions? Do they help people build capability, or do they make capability dependent on continuous compliance with hidden rules?
Sovereign Capability Economics offers a clear standard: a healthy economy should increase the practical ability of people to create value, own what they create, learn, choose, and exit. Software is compatible with that purpose when it expands opportunity while preserving accountability and alternatives. It is incompatible with that purpose when it becomes an invisible sovereign over who may participate.
Participation Must Remain Contestable
Automation will continue to shape economic life. The relevant choice is not between a technological future and a non-technological past. It is between systems that treat people as active participants and systems that reduce them to categories managed by institutions they cannot question.
The algorithmic permission economy becomes dangerous when a person’s ability to work, trade, borrow, be seen, or be paid depends on decisions that are invisible, irreversible, and without alternative. Such systems may be efficient for the institution while quietly making the individual more dependent, more legible to power, and less able to direct their own economic future.
Technology should make participation easier, not make permission more absolute. A legitimate economic system can use software to improve speed, safety, and access. But it must retain a place for explanation, correction, judgment, and exit.
No algorithm should have the final power to exclude a person from economic life without giving that person a reason, a remedy, and another path.