Power has traditionally been easy to imagine.
It had buildings, uniforms, offices, signatures, gates, and people who could be identified as decision-makers. Authority occupied a visible place. A person entered an institution, encountered an official, received a decision, and understood, at least in broad terms, where that decision had come from.
This image of power has not disappeared.
But another form of power has grown alongside it.
A recommendation determines what appears on a screen. A scoring system influences whether a transaction appears risky. A platform decides which content receives visibility. A database determines whether a person can be found within an administrative system. An automated process identifies an anomaly and quietly moves a case into a different category.
No official needs to stand in the room.
No command needs to be spoken.
Sometimes, no individual can easily explain the decision at all.
Power remains present in its consequences while becoming increasingly absent from view.
This is one of the defining characteristics of digital society.
From Visible Authority to Embedded Authority
Modern institutions have always relied on systems.
Forms, classifications, registries, procedures, archives, and bureaucratic routines existed long before computers became central to administration. Institutions could not operate at scale without mechanisms that transformed complex social realities into categories that could be recorded and processed.
Digitalization did not invent this logic.
It intensified it.
Decisions that once required visible administrative interaction can now occur within infrastructures that operate continuously and often automatically. Rules become embedded in software. Categories become database fields. priorities become rankings. Institutional assumptions become parameters.
Lawrence Lessig’s influential argument that “code” functions as a form of regulation remains important precisely because digital architecture does more than facilitate action. It also defines what actions are possible, difficult, visible, invisible, permitted, or prohibited (Lessig, 2006).
Architecture governs quietly.
A physical gate communicates restriction through its presence.
A digital system can produce the same restriction through an interface that simply does not provide an option.
There may be no prohibition written on the screen.
The possibility merely disappears.
This marks an important transformation in the experience of authority. Power increasingly operates not only through explicit commands but through the design of environments within which choices are made.
The Quiet Authority of Systems
Digital systems rarely describe themselves as authorities.
They appear as tools.
A dashboard presents information. An algorithm ranks results. A recommendation system suggests content. A risk model produces a score. A database verifies a record.
Each appears technical rather than political.
Yet technical systems participate in decisions about relevance, priority, eligibility, visibility, and risk.
Tarleton Gillespie argues that algorithms should not be understood simply as neutral mathematical procedures. They participate in determining what becomes relevant within contemporary information environments, while the institutional choices embedded within those processes can remain difficult for users to see (Gillespie, 2014).
This matters because classification is never entirely passive.
To classify something is also to create boundaries.
Someone is eligible or not eligible.
A transaction is normal or suspicious.
Information is relevant or irrelevant.
Content is recommended or ignored.
A case is urgent or routine.
These distinctions may be necessary for institutions to function. The problem begins when their institutional origins disappear behind the apparent neutrality of computation.
A decision produced by a person can be questioned as a judgment.
A decision produced by a system can easily be mistaken for a fact.
Distance Between Decision and Decision-Maker
One of the most significant consequences of digital governance is the growing distance between those affected by decisions and those responsible for designing the systems that produce them.
Consider an ordinary institutional interaction.
A person submits information.
The information enters a system.
The system compares it with existing records.
Rules determine whether inconsistencies exist.
A score or classification may be generated.
The result appears on another person’s screen.
At the end of this process, someone may receive a decision.
But where exactly did the decision occur?
Was it made by the officer who accepted the result?
By the programmer who translated policy into software?
By the institution that established the classification?
By the data used to train or configure the system?
By the regulation that defined the original criteria?
Or by the organizational culture that encouraged employees to trust automated outputs?
The answer may involve all of them.
Responsibility becomes distributed across an institutional network.
This distribution can improve efficiency. It can also make accountability difficult.
There may no longer be a single point at which power can easily be located.
Authority becomes infrastructural.
Seeing Without Being Seen
The asymmetry becomes even more significant when digital systems are capable of observing individuals while remaining largely invisible to those being observed.
People generate enormous quantities of data through ordinary activities. They search, travel, communicate, purchase, watch, read, click, work, and interact.
These activities leave traces.
Those traces can be aggregated.
Aggregation makes patterns visible.
Patterns make prediction possible.
Shoshana Zuboff describes surveillance capitalism as an economic logic in which human experience becomes a source of behavioral data that can be used for prediction and commercial purposes (Zuboff, 2019). Her argument points toward a broader question about contemporary power: the unequal distribution of the capacity to know.
Institutions and platforms may know increasingly detailed things about populations and individuals.
Individuals usually know far less about the systems observing them.
The inequality is therefore not only an inequality of information.
It is an inequality of visibility.
Some actors can observe without being observed.
Some systems can classify without explaining their classifications.
Some institutions can predict behavior while those being predicted cannot inspect the mechanisms producing those predictions.
Power becomes strongest precisely where its operation becomes least visible.
The Black Box Problem
Opacity is not always intentional.
Modern digital systems can be extraordinarily complex. They may involve multiple databases, external vendors, machine-learning models, organizational procedures, legacy infrastructure, legal requirements, and layers of technical configuration developed over many years.
No single employee may fully understand the entire system.
This creates an important distinction.
A system does not need to be secret to become opaque.
Complexity itself can produce opacity.
Frank Pasquale describes the problem of the “black box society” in which important decisions may depend upon mechanisms that remain inaccessible to those affected by them (Pasquale, 2015).
The problem is not simply that algorithms exist.
Nor is every automated decision inherently unjust.
The deeper problem emerges when significant decisions become difficult to understand, contest, or attribute.
A person can appeal against an identifiable decision.
It is far harder to appeal against an invisible chain of classifications.
Opacity changes the relationship between citizens and institutions because explanation is part of accountability.
People do not only need decisions.
They sometimes need to understand why those decisions were made.
Power Through Prediction
Traditional institutions frequently acted after events occurred.
Digital systems increasingly attempt to act before them.
Fraud detection tries to identify suspicious transactions before losses occur. Predictive maintenance attempts to detect equipment failure before breakdown. Recommendation systems anticipate what users may want. Public institutions increasingly experiment with analytics intended to identify risk before problems become visible.
Prediction can clearly produce social value.
Prevention is often better than reaction.
But prediction introduces another form of power.
A prediction does not merely describe a possible future.
Once institutions act upon it, the prediction can influence that future.
A person categorized as high risk may receive additional scrutiny.
A piece of content predicted to generate engagement may receive greater visibility.
A neighborhood identified through certain indicators may become the focus of intervention.
A profile considered unlikely to respond may receive fewer opportunities.
The model begins to shape the environment it claims merely to observe.
This creates a subtle feedback loop.
Past data informs prediction.
Prediction informs intervention.
Intervention influences future behavior.
Future behavior becomes new data.
Power therefore moves beyond deciding what has happened.
It begins participating in what is likely to happen next.
Governance Without Commands
This transformation challenges conventional ideas of governance.
Governance is often imagined as a sequence of explicit acts: laws are enacted, regulations issued, decisions made, and instructions communicated.
Digital environments introduce a different possibility.
Behavior can be influenced through architecture rather than command.
An interface can encourage one option by making it easier.
A ranking can influence attention without prohibiting alternatives.
A default setting can shape behavior without requiring consent each time.
A notification can create urgency.
A recommendation can redirect attention.
A scoring mechanism can encourage people to modify behavior even when no formal rule requires them to do so.
There is no need to issue an order.
The environment itself becomes the instruction.
This is why algorithmic governance cannot be treated only as a technical problem. Karen Yeung and Martin Lodge describe algorithmic regulation as part of a broader transformation in the relationship between technological systems and regulatory governance (Yeung & Lodge, 2019).
The important question is not whether algorithms govern society independently.
They do not.
Algorithms are created, deployed, funded, regulated, interpreted, and maintained by human institutions.
The important question is how institutional authority changes once it begins operating through computational infrastructure.
Power does not disappear.
It changes location.
The Problem of Human Judgment
Digital systems are often introduced because human judgment has limitations.
Humans are inconsistent.
They become tired.
They overlook information.
They carry biases.
They cannot process millions of records simultaneously.
Computational systems can help overcome some of these limitations.
Yet institutional dependence on automated systems creates another danger.
Human judgment may gradually become secondary.
An officer may theoretically retain the authority to override a recommendation while becoming increasingly reluctant to do so.
Why reject a result produced by a system that has processed thousands of variables?
Why accept personal responsibility when the algorithm appears more objective?
Over time, the relationship can reverse.
The system was originally designed to assist human judgment.
Human judgment begins assisting the system.
This does not require formal automation.
It requires only institutional trust in computational outputs to become stronger than institutional confidence in human interpretation.
The result is a peculiar form of authority.
The algorithm may have no legal personality.
It may possess no formal office.
It cannot be held morally responsible.
Yet its recommendation can become extraordinarily difficult to challenge.
Karen Yeung identifies concerns surrounding machine decision-making not only in the outputs generated by automated systems but also in decision processes and in systems designed to predict and personalize decisions (Yeung, 2019).
The issue therefore extends beyond technical accuracy.
A perfectly accurate system can still raise questions about authority.
Who should have the right to decide?
Recovering the Visibility of Power
The answer is not to return to a world without digital systems.
Such a world is neither realistic nor necessarily desirable.
Data systems can improve consistency, detect errors, expand access to services, identify patterns invisible to individual observers, and allow institutions to operate at scales impossible through manual processes alone.
The challenge is different.
Digital institutions must preserve the visibility of responsibility.
Every significant automated decision should remain connected to an institutional chain that can ultimately be understood.
Who established the rule?
What information was used?
Why was a particular classification produced?
Can the result be questioned?
Who has the authority to correct it?
Where does responsibility ultimately reside?
These questions may appear procedural.
They are actually questions about power.
Transparency alone will not solve every problem. Publishing source code does little for citizens who cannot interpret it. Explaining a mathematical model does not necessarily explain the institutional assumptions that shaped its design.
Meaningful accountability requires more than technical disclosure.
It requires institutional intelligibility.
People should be able to understand not necessarily every line of code, but the logic through which decisions affecting them become possible.
Power can be complex.
It should not become unreachable.
Human Agency in Invisible Systems
The most important question raised by digital power may therefore concern human agency.
Technology increasingly mediates the environments in which decisions are made. It organizes information before people see it. It determines which possibilities appear first. It categorizes experiences before institutions respond to them.
None of this means that humans have lost agency.
But agency becomes more difficult to exercise when the architecture shaping choices remains invisible.
A person cannot question a category they do not know exists.
A citizen cannot challenge a rule whose operation cannot be observed.
An employee cannot exercise judgment if institutional culture treats computational output as unquestionable.
A society cannot meaningfully govern digital power if that power is continuously mistaken for technological inevitability.
Algorithms do not arrive from outside society.
Databases do not construct themselves.
Metrics do not decide their own importance.
Digital architectures contain choices.
Those choices may be technical, economic, administrative, political, or cultural.
Recognizing them as choices is essential because anything constructed through choices can also be reconsidered.
This is where the question of data justice becomes unavoidable.
Justice in a digital society cannot be concerned only with who owns data or who has access to technology. It must also examine who possesses the capacity to classify, predict, rank, and shape the environments within which others act.
These are emerging forms of institutional power.
Their influence may be enormous precisely because they rarely look like power.
Power Does Not Need to Be Seen
For much of history, power announced itself.
It occupied institutions, appeared in documents, spoke through officials, and exercised authority through visible procedures.
Digital society is producing something different.
Power can now operate through interfaces, databases, rankings, classifications, recommendations, predictions, and automated rules.
It can influence without commanding.
Observe without appearing.
Restrict without explicitly prohibiting.
Prioritize without publicly declaring a preference.
Decide without producing an identifiable decision-maker.
This does not make digital systems inherently oppressive.
It makes them politically significant.
The future of digital governance will therefore depend not only on whether systems become more intelligent, accurate, or efficient.
It will depend on whether societies remain capable of seeing the power embedded within them.
Because the most consequential form of power may not be the authority standing visibly in front of us.
It may be the architecture quietly determining what appears possible before a decision is ever made.
Power has not disappeared.
It has learned how to operate without presence.
References
Gillespie, T. (2014). The relevance of algorithms. In T. Gillespie, P. J. Boczkowski, & K. A. Foot (Eds.), Media Technologies: Essays on Communication, Materiality, and Society (pp. 167–194). MIT Press. https://doi.org/10.7551/mitpress/9780262525374.003.0009
Lessig, L. (2006). Code: And Other Laws of Cyberspace, Version 2.0. Basic Books.
Pasquale, F. (2015). The Black Box Society: The Secret Algorithms That Control Money and Information. Harvard University Press.
Yeung, K. (2019). Why worry about decision-making by machine? In K. Yeung & M. Lodge (Eds.), Algorithmic Regulation (pp. 21–48). Oxford University Press. https://doi.org/10.1093/oso/9780198838494.003.0002
Yeung, K., & Lodge, M. (Eds.). (2019). Algorithmic Regulation. Oxford University Press.
Zuboff, S. (2019). The Age of Surveillance Capitalism: The Fight for a Human Future at the New Frontier of Power. PublicAffairs.
