The Algorithmic Architecture of Governance

Why well-run institutions can still fail society, and what we can do about it

Gueorgui Vassilev

An introduction to the theory of social algorithmisation

In brief

We tend to judge public institutions by how smoothly they run. Are the forms processed on time? Is the budget spent? Are the procedures followed? These are reasonable questions, but they miss something important. An institution can do all of these things well and still fail to solve the problem it was created to solve.

This paper presents a simple idea for understanding why. Every organised activity in society, from issuing a building permit to passing a national budget, follows a recognisable pattern of steps, rules and decisions. I call these patterns social algorithms. If we learn to see governance as a large system of such algorithms working together, we can start to ask better questions: not only whether each part works, but whether the whole still serves the purpose it was built for.

The key argument is this: the part of the state that sets goals (politics) and the part that carries them out (administration) slowly drift apart over time. Administration can even begin to shape the goals that politics sets. The result can be a system that looks efficient from the inside while drifting further and further away from what society actually needs.

1. A puzzle: busy, efficient, and still not working

Modern states have more capacity than ever before. They have large administrations, detailed regulations and powerful tools for handling information. Yet many people sense that this growing machinery does not always produce better lives, better cities or better services.

A public office can process more cases each year, follow every rule, spend its entire budget and hit all of its internal targets, and still leave the underlying problem unsolved. So the question is:

How can a system in which every part follows the rules still produce results that go against its own purpose, or even harm the society it is meant to serve?

Part of the answer is that efficiency in each small process does not guarantee that the overall design makes sense. But there is a deeper part of the answer, and it concerns the goals themselves.

Institutions are created to achieve particular goals at a particular moment in history. They then often outlive that moment by decades. Leaders change, conditions change, and sometimes the original goal is quietly altered or forgotten. The procedures, however, keep running. Over time, an organisation can also develop interests of its own. It keeps doing what it has always done, but the link between its daily work and the reason it exists grows weaker.

To make sense of this, I propose looking at society through the lens of the algorithmisation of social processes.

2. What is a social algorithm?

When most people hear the word “algorithm”, they think of computer code. Here the word is used more broadly. A social algorithm is simply the recurring way in which people and organisations turn a goal into action: the rules they follow, the decisions they make, the information they pass along and the resources they use.

Put plainly, a social algorithm is the real, repeatable pattern by which a group or institution tries to get something done.

Unlike a computer program, a social algorithm is carried out by human beings. People interpret rules, use judgement, work with incomplete information, have conflicting interests and react to changing circumstances. So social algorithms are never perfectly predictable. That does not mean they cannot be understood. We can still map their structure, just as we can map the route of a river even though no two drops of water travel exactly the same way.

Take public procurement as an example. On paper it has clear stages: approval, budgeting, evaluation of bids, signing the contract, delivery and checking. In practice it also depends on informal arrangements, habits, incentives and personal judgement that never appear in the official rulebook. What matters, then, is not only the procedure as written but the algorithm as it actually works.

A society is many algorithms at once

No institution works alone. A society runs on many interlocking algorithms: those that set goals, those that assign authority, those that collect and spend money, those that gather and pass on information, those that make and carry out decisions, and those that check results and correct mistakes. What we experience as “the system” is the result of all of them interacting.

This is why improving one part does not automatically improve the whole. A procedure can be made faster while creating new costs somewhere else. A regulator can become better at enforcing its own rules while making the activity it regulates less effective. Engineers know this principle well: making each part better does not guarantee that the whole works better. In society, the problem is even harder, because the different parts may be pursuing goals that do not fit together.

Goals are part of the machine

We usually judge an algorithm by whether it achieves its goal. But in public life, goals are rarely fixed and clear. There are usually several of them, some conflicting, some vague, and all open to reinterpretation. The organisation that carries out a goal often has a say in how that goal is later defined.

So a full picture of governance has to include not only the processes that carry out goals, but also the processes that create, justify, interpret and revise them. This brings us to the central distinction of the theory.

3. Energy and information: what keeps an algorithm alive

So far I have described social algorithms as patterns of rules and decisions. But a rule written on paper does nothing by itself. A law that no one enforces, a procedure that no one carries out, a plan without money or people behind it, is just text. To understand why some institutional algorithms work and survive while others fade away, we need to look at what feeds them. My answer is: energy and information.

Society as a living system

In an earlier essay, The Anatomy of Power, I argued that a cell, a multicellular organism and a human society are, at bottom, similar phenomena. Seen through the lens of how they are organised, how energy moves through them and how they are governed by information codes, they show the same basic properties. They are not separate worlds but rising levels of complexity in the organisation of living matter.

Every living system faces the same challenge. It must keep itself organised inside while the world around it tends towards disorder. Physicists call this tendency entropy. A living system survives only by constantly taking in energy from outside and using it to maintain its internal order. And it can only do this because it carries an information code, in the case of the cell its genome, which tells it how to capture energy, how to use it and how to reproduce itself.

This can be summed up in one sentence, which I consider the basic law of living matter:

Living matter survives by preserving an information code that allows it to stay distinct and to reproduce itself, by managing the flow of energy between its organised inner world and the disorganised world outside.

Society obeys the same law. Its information code is made up of its traditions, moral rules, laws and institutional procedures, in other words, its social algorithms. These algorithms decide how society produces and distributes its energy and, just as importantly, how the code itself is created and changed. Like a genome, institutional algorithms are fairly stable, yet they can also mutate and evolve.

What “energy” means in a society

In a society, energy rarely takes the form of heat or fuel alone. It appears as money and budgets, as human labour and time, as capital and equipment, as authority and the ability to compel, and even as public attention and trust. Every institution lives on such a flow. A ministry needs a budget, staff and legal powers. A court needs judges, buildings and the willingness of citizens to accept its rulings. Remove the flow, and the institution stops working, however elegant its rules may be.

Information is the other half. Institutions act as channels through which information from inside and outside society flows towards decision-makers, where it is processed according to established rules. Information also travels the other way, as instructions, laws and orders. An institution that receives distorted information, or none at all, is like an organism with damaged senses: it may still have plenty of energy, but it spends it in the wrong direction.

Seen this way, an institution is a structure together with the algorithms that govern it, kept alive by a constant flow of energy and steered by a constant flow of information.

Algorithms have no budget, living systems do

There is an important difference between an algorithm as an idea and an algorithm at work. As an idea, an algorithm is virtual. It has no energy limits and needs no budget. One can write a law that promises everything to everyone. But an institution that must actually carry out that law is a living system. It has limited energy, and it must constantly choose what to spend it on. It needs, so to speak, a budget calculator.

This is why setting priorities is not an administrative detail but a question of survival. A political goal that ignores the energy needed to achieve it is not a goal but a wish. And an administration that receives goals without the matching energy will quietly reinterpret them until they fit the resources it actually has. This is one of the hidden roots of the drift between politics and administration described later in this paper.

Power as the regulator of energy and information

From this perspective, state power is the instrument through which society directs its energy towards building and maintaining stable structures, and through which it creates and enforces the information codes that shape how people interact. Like any instrument, power is ethically neutral. It can serve the flourishing of human beings or their destruction. Everything depends on the information code that society has built into the institutions through which power is exercised.

One further consequence matters greatly for governance. A prime minister and a factory worker both depend on their ability to process information and act on it. But the consequences of their actions are radically different, because the results of each person’s decisions are multiplied by the energy they control through their position. Hence a simple rule: the quality of algorithmic ability required of a person must grow in proportion to the energy placed at their disposal. The higher the office, the higher the demands on competence and values.

The same logic explains a familiar danger. When someone in power uses their position mainly to satisfy their own energy needs, or those of their group, the flow meant to sustain public structures is diverted to private ones. In Bulgarian this is captured bluntly in the saying that power is “a windfall”. In the language of this theory, it is energy leaking out of the social organism.

4. Two different engines: politics and administration

Politics decides what we want

The job of the political system is to decide what society as a whole is trying to achieve. In a democracy this happens through elections, party competition, parliamentary debate, legislation and other constitutional procedures. These are political algorithms. They take in public preferences, needs, interests and ideas, and they produce mandates, laws, policies and budget priorities.

Society’s needs and preferences → political process → legitimate shared goals

But a goal can be fully legitimate and still be unrealistic, internally contradictory or poorly suited to the problem. Political goals are often set with incomplete information, on mistaken assumptions, or as compromises that are hard to put into practice. Politics has its own ways of going wrong.

Administration turns goals into reality

The job of administration is different: to turn the goals that politics has approved into concrete decisions, services and results. And it is important to speak of administrations in the plural. Ministries, municipalities, agencies and public companies each have their own mandates, rules, cultures, budgets and degrees of independence. Each runs its own partly independent algorithms.

Approved goals → administrative processes → institutional action → results in society

The difficulty is that these many administrations do not necessarily fit together into a coherent whole. Each may do its own job correctly, and yet their combined effect can be overlapping, contradictory or simply not enough.

In theory a hierarchy, in practice a two-way street

On paper, politics sets the goals and administration implements them. In reality, politicians depend heavily on administrators for expertise, institutional memory, information, assessments of what is feasible, and the capacity to actually deliver. Influence therefore runs in both directions.

This is not necessarily a bad thing. Experienced civil servants can prevent unworkable policies, keep the law consistent and protect long-term public interests. The problem begins when this relationship becomes systematically bent by institutional interests, unequal access to information or the instinct of organisations to protect themselves.

5. How politics and administration drift apart

Politics and administration follow different logics. Politicians think about mandates, public expectations and keeping their promises. Administrators must also think about legality, continuity, capacity, professional standards and stability. Most of the time these concerns support each other. Over time, however, they can pull in different directions.

I call this systemic misalignment: the point at which the goals, incentives or working methods of connected institutions are no longer coordinated well enough to produce the results society expects. Importantly, this rarely requires anyone to behave badly. It usually happens gradually, as a normal part of institutional life. Three mechanisms are especially important.

Institutions outlive their reasons

Administrative bodies usually survive the people and circumstances that created them. As leaders come and go, knowledge gets lost. Nobody remembers why a certain rule was introduced. Rules made to solve a specific problem keep operating after the problem has gone. Each reform adds new tasks without removing old ones. Gradually, the administration becomes a layered collection of arrangements that only partly fit together. It still works, but it no longer reflects a clear set of goals. I call this algorithmic inertia.

The measure replaces the mission

A second mechanism is what happens when the tools for measuring progress quietly become the goal itself. An office meant to improve public services begins to judge itself by the number of cases closed. Spending the budget becomes a sign of success, regardless of what the money achieved. Following the rules becomes an end in itself rather than a way to protect the public.

The process keeps running, but it is now tuned to hit a number rather than to achieve the original purpose. I call this algorithmic goal displacement. Its danger is that an institution can show steady improvement on its own indicators while its real contribution to society stays flat or declines.

The world changes, the goals do not

Even when the original goals are faithfully kept, the world around them changes. New technologies, an ageing population, economic shifts, environmental pressures and changing public values can all make yesterday’s goals unsuitable for today. If an institution has no way of questioning its own assumptions, becoming more efficient may actually make the mismatch worse. In such cases the problem is not poor implementation. It is the excellent implementation of a goal that no longer fits reality.

6. When administration starts steering politics

Survival becomes a goal

Every organisation needs continuity, money, staff, authority and stability in order to do its work. These needs are legitimate. But over time, keeping the organisation itself alive can become an unspoken goal. People inside it naturally want to protect budgets, jobs, responsibilities and independence. When these concerns start to dominate, the organisation’s survival competes with the purpose society gave it. In the language of energy and information, the organisation begins to treat the flow of resources that sustains it as an end in itself, rather than as fuel for its public function. No conspiracy is needed for this to happen. It is simply the sum of many people acting sensibly in response to the incentives in front of them.

Whoever holds the information shapes the choices

Administrators usually know far more about how things work on the ground than the politicians who oversee them. This gives them real influence over how problems are understood and which solutions seem possible. Information is selected, summarised and framed in familiar administrative terms. Some options look technically impossible simply because they clash with existing procedures. Others are never raised because the institution lacks the ability or the interest to carry them out. Control over the flow of information is, in effect, control over where society’s energy will be directed. In this way, administration shapes not only how goals are carried out, but which goals are seen as realistic in the first place.

The direction of influence reverses

We normally assume that politics sets the goals and administration follows. But the arrow can partly turn around:

Existing administrative structures → their interests and information → political choices → adjusted goals

Politics then starts adjusting its goals to fit the administration it already has, instead of adjusting the administration to fit the goals society has chosen. This form of misalignment is especially hard to spot, because the political decisions remain fully legal and legitimate. What has been quietly narrowed is the range of options that were ever on the table.

A self-reinforcing loop

Imagine how this can unfold step by step. A government sets a goal without fully understanding what it will take to deliver. The administration reinterprets the goal to fit its existing procedures and resources. The resulting difficulties are reported back to the government, using information produced by that same administration. The government then softens the goal to accommodate those difficulties. The administration takes this as proof that its existing arrangements were right all along. And the cycle begins again.

Over time, politics and administration become very comfortable with each other, while becoming less and less connected to the actual problem society faces. This leads to one of the central warnings of the theory:

Politics and administration can reach a perfect accommodation with each other without ever solving society’s problem.

7. When efficiency becomes dangerous

Doing things right versus doing the right things

A process can be extremely good at reaching its target. That says nothing about whether the target is the right one. A system built to maximise an outdated or poorly chosen target will keep producing it, and will get better and better at it. If that target conflicts with the wider good, every improvement makes things worse.

The efficiency of a process must never be confused with the soundness of the goal it serves.

Three questions to ask of any institution

To avoid this trap, the theory proposes judging institutions at three levels:

Efficiency: Does the process do its job without wasting time, money, information and effort?
Effectiveness: Does that work actually help achieve the goal the institution was created for?
Adequacy: Is that goal still sensible, consistent, achievable and relevant to what society needs today?

These are separate questions. An institution can be highly efficient and still ineffective. It can be highly effective at pursuing a goal that is no longer adequate. A real assessment must look at all three.

Survival is not proof of usefulness

In nature, organisms that do not fit their environment tend to disappear. Institutions face no such automatic test. They survive because they keep securing a flow of energy, through legal mandates, political protection, budget arrangements or simple habit, not necessarily because they work. Ineffective institutions can last for a very long time, especially when the costs of their failure are spread across society while the benefits of their continued existence stay concentrated inside them.

Elections, courts, competition, a free press and reform efforts all help to correct course, but none of them guarantees that problems will be noticed and fixed in time. The fact that an institution still exists and still operates tells us nothing about whether it is still fit for purpose.

More resources can magnify mistakes

This matters most when an institution is given more money, more technology or more data. If its underlying design is flawed, extra resources simply let it produce the wrong results faster. Digitalisation and artificial intelligence are a striking example. They can make administration faster, more consistent and larger in scale without ever fixing its goals or its structure. The result can be the technological amplification of error. Technology multiplies the energy available to an algorithm. If the information code behind it is flawed, it multiplies the flaw as well.

Before speeding up an institutional process, we should first check whether the process, its goals and its place in the larger system are designed correctly.

8. Three levels of governance

Bringing these ideas together, governance can be seen as working on three levels:

Level one: getting things done. These are the everyday procedures: decisions, spending, service delivery and compliance checks. We judge them by efficiency and effectiveness.
Level two: translating and coordinating. These are the processes that interpret political decisions, turn them into mandates for institutions, resolve conflicts between institutions and adjust how goals are pursued. This level also includes the formal and informal ways administration influences politics, which makes it the most important place to look for misalignment.
Level three: checking and adapting. These mechanisms step back and ask whether the goals are still right, whether the institutions can achieve them, whether the procedures work as intended, whether the results match what society expected, and whether something needs to change.

The third level does not have to be a single institution. It can be shared among universities and research centres, independent evaluators, oversight bodies, political leaders and public debate. What defines it is the ability to look at the system as a whole, rather than through the performance figures of its individual parts. This three-level model is offered as a working framework, to be developed further and tested against real cases.

9. From theory to practice: a method

A theory is only useful if it can be applied. Building on these ideas, I propose a practical method, which I call Algorithmic Systems Analysis and Engineering of Socio-Institutional Systems. It works in eight steps:

Rediscover the goals. Find out what the system is officially supposed to achieve, from laws, mandates, mission statements and policy documents. Look at the circumstances in which these goals were set, and separate the original goals from later changes and from the working targets institutions actually use.
Map the institutions. Identify which organisations are involved, who has what authority, who depends on whom for resources, and how they are supposed to coordinate.
Reconstruct how things really work. Trace the real flows of energy and information: money, staff, authority, data, decisions, approvals and feedback. Compare official procedures with everyday practice, using interviews, records and data.
Find the misalignments. Look for conflicting or displaced goals, duplicated or contradictory functions, distorted information, incentives that reward self-preservation, rules that outlived their reasons, and interactions with unintended effects. The aim is not just to spot inefficient procedures, but to understand how they cause one another.
Question the goals themselves. Check whether the goals rest on sound assumptions, are achievable, fit together and match today’s conditions. Where goals are politically chosen, the analyst must not replace democratic choices with personal preferences, but should lay out the trade-offs, side effects and alternatives for decision-makers to consider.
Develop better institutional arrangements: new decision rules, responsibilities, structures, information flows, funding and feedback, so that the system serves its legitimate goals more closely.
Test before committing. Evaluate proposed changes through historical comparison, scenarios, simulation, comparison with other places or carefully designed pilots, always separating improvements in internal indicators from real improvements in people’s lives.
Keep watching. Set up ways to detect future drift in the environment, in institutional behaviour, in political goals and in outcomes. Reform is not a one-off fix but part of an ongoing process of learning and adapting.

10. Why this matters

Seeing politics and administration as two linked but different systems of algorithms offers a fresh way of thinking about what goes wrong in government. Failures are not always caused by bad leaders, incompetent officials or a lack of money. Sometimes the problem lies in how otherwise capable institutions are connected.

Politicians may set goals that are hard to turn into practical instructions. Administrations may have abilities and incentives that do not match those goals. Information may travel along routes that distort the picture of what is really happening. The instinct for self-preservation may keep old arrangements alive long after they have stopped being useful. And all of this can coexist with full legality, careful rule-following and seemingly good performance figures.

The core question is therefore one of consistency: do the goals, the way institutions work and the actual results still fit together?

This question becomes more urgent as technology speeds up public administration. The more capable institutions become at processing information, spending money and carrying out decisions, the greater the damage they can do when they are pursuing the wrong goals or working through poorly designed processes. Truly intelligent governance needs more than efficient execution. It needs the ability to notice when its goals, assumptions or structures no longer match reality.

This does not end political disagreement, and it does not offer a formula for deciding what society should want. Shared goals will always involve competing values, interests and legitimate choices. The contribution of this approach is to make the links between goals, institutional behaviour and consequences more visible, easier to understand and possible to test. It is meant to support democratic governance, not to replace it.

11. Conclusion

Modern government works through many interacting systems of algorithms. Politics sets and legitimises shared goals. Administration turns those goals into decisions and results. Neither system stands still, and they are never perfectly coordinated. They respond to different incentives, constraints, information and time horizons, and so they gradually drift apart.

Administrations can develop a self-protective logic that shapes what politicians aim for, while politicians may adjust their goals to fit institutional limits rather than real public needs. Meanwhile, a changing world can make once-sensible goals obsolete. The result can be a system that runs efficiently while achieving little, or one that effectively pursues the wrong things.

The real challenge, then, is not just to make institutions run better. It is to keep three things in line with one another: the goals society has legitimately chosen, the processes designed to reach them, and the results people actually experience.

Underneath all of this lies a simpler truth. Institutional algorithms live only as long as they are fed by energy and guided by information. Whoever controls these two flows controls which algorithms survive, and therefore what kind of society we live in.

The theory of social algorithmisation offers a way to examine how goals are set, passed on, changed, carried out and judged. The method built on it aims to turn this perspective into a practical tool for diagnosing institutions, redesigning them and testing whether the changes work. Its ultimate ambition is to help build systems of government that can not only carry out their goals efficiently, but also recognise when those goals, or the way they are pursued, need to be reconsidered.

A truly intelligent system of government is not one that simply runs its procedures efficiently. It is one that can recognise when those procedures, or the goals behind them, no longer match reality.

Философският клуб