Optimisation was supposed to be a tool. You applied it to a specific problem, solved it, and moved on. That is not what happened. Optimisation became an environment — the default condition of infrastructure, work, price, attention, and urban space. We are not optimising particular things anymore. We are living inside systems that continuously optimise everything, including us. The question worth asking is not whether this is efficient. It is what we are losing in the process.
The Invisible Architecture
The price of your Uber changes while you are opening the app. The route your delivery driver takes is recalculated every thirty seconds. The content you see is ranked by systems that predict what will hold your attention longest. The job application you submitted was filtered before a human read it. None of this is visible. All of it is shaping the conditions of your day.
James C. Scott argued in Seeing Like a State that the impulse to make social life legible — to render it readable, measurable, and manageable — consistently destroys the informal complexity that makes systems actually work. The village that looks chaotic on a map is functional in ways the map cannot capture. The optimised version, built from the map, often is not. His examples were agricultural collectivisation and urban planning. The mechanism he described applies equally to the algorithmic systems now running daily life.
As explored in “The Quiet Expansion of Algorithmic Life,” the choice architecture of contemporary life is increasingly pre-computed. You make decisions inside a system that has already narrowed your options. That is not freedom constrained. Instead, systems pre-shape freedom, making it harder to see and harder to resist.
Goodhart’s Law at Scale
There is a principle in social science called Goodhart’s Law: when a measure becomes a target, it ceases to be a good measure. Credit scores, school performance rankings, engagement metrics, productivity dashboards — each starts as a useful proxy for something real. Each, once optimised for, begins to distort the thing it was meant to represent.
Algorithmic systems often overlook intangible contributions such as creativity, empathy, and collaborative problem-solving. The overemphasis on efficiency leads to neglecting relational dynamics and subjective worker experiences. Some forms of labour—mentoring, informal leadership, and creative contributions—are difficult to measure. As a result, algorithmic evaluations often exclude them.
This is not a bug. It is a structural consequence of measurement at scale. You can only optimise what you can quantify. Systems exclude what they cannot quantify. Over time, this also removes it from the decisions those systems make. Algorithmic management can undermine job quality when it is built solely around efficiency. The same logic applies to optimised cities, optimised healthcare triage, optimised content — wherever the metric replaces the judgment, the unmeasurable becomes invisible.
Solutionism and Its Costs
The critic Evgeny Morozov coined the term solutionism. He uses it to describe the belief that people can treat social problems as engineering problems. This view assumes that friction, inefficiency, and disorder are failures that people can fix. The belief is not malicious. It is, in many cases, genuinely well-intentioned. It is also consistently wrong in the same way: it mistakes the symptom for the problem.
A city with unpredictable street life is not a failure of urban planning. It is the condition under which serendipity, community, and informal economy occur. An optimised version — legible, sensor-monitored, flow-controlled — may move traffic more efficiently while destroying the conditions that made the street worth being on. As explored in “Cities Are No Longer Selling Places. They Are Selling Experiences,” the most valuable urban qualities are precisely those that resist measurement: atmosphere, encounter, the feeling of a place.
Dynamic pricing is another version of this problem. Uber’s surge pricing is, in the narrow sense, efficient — it allocates rides to the people most willing to pay for them at the moment of highest demand. It also means that an algorithm sets the price of getting home after a concert or leaving a dangerous neighbourhood late at night. That algorithm does not know who you are or why you need the ride. Efficiency is optimised. Fairness is not a variable in the system.
Systems That Pre-Shape Behaviour
The deepest problem with optimisation as an environment is not that it makes bad decisions. It is that it shapes behaviour before decisions are made. Algorithmic systems co-constitute organisational norms, workflows, and relationships. Instead of treating technologies as neutral tools, we should see them as actively shaping how work is organised, how people understand performance, and how workers relate to each other.
Shoshana Zuboff’s analysis of surveillance capitalism makes the same point at a larger scale: the data collected to predict behaviour is used to modify it. The recommendation system is not neutral. Training focuses on outcomes that serve the platform, and the system shapes your choices in that direction. You believe you are choosing. In a meaningful sense, you are being steered.
Foucault described power operating not through prohibition but through the construction of “optimal” behaviours — not forbidding things, but making certain things feel natural, efficient, correct, while others feel deviant or costly. The nudge, the default setting, the ranked result: these are small exercises of that kind of power, accumulated at scale and executed algorithmically.
The Narrowing
Optimisation is not the enemy. Applied to specific problems with clear objectives and human oversight, it is genuinely useful. The issue arises when this becomes the default mode of organising everything. Efficiency stops being an outcome to achieve and becomes a governing principle to obey.
What looks like improvement is often a narrowing of possibility. The optimised route is faster. It is also always the same route. The optimised content feed shows you more of what you already responded to. The optimised workspace tracks output and misses the conversation that changes someone’s thinking. Each individual optimisation is defensible. In the aggregate, the world contains less slack, less serendipity, and less room for things that matter but resist measurement.
Optimisation is no longer a tool. It is an environment. And environments, unlike tools, are not easy to put down.
Key Sources
- CEPS, “Algorithmic management isn’t the problem — what it optimises is”: https://www.ceps.eu/algorithmic-management-isnt-the-problem-what-it-optimises-is/
- arXiv / New Technology, Work and Employment, “The Rise of Algorithmic Management”: https://onlinelibrary.wiley.com/doi/10.1111/ntwe.12343
- arXiv, “Algorithmic Management and the Future of Human Work”: https://arxiv.org/html/2511.14231v1
- The Guardian, Evgeny Morozov on solutionism: https://www.theguardian.com/commentisfree/2013/feb/23/evgeny-morozov-technology-solutionism
- Wikipedia, Goodhart’s Law: https://en.wikipedia.org/wiki/Goodhart%27s_law
- Princeton University Press, James C. Scott, Seeing Like a State: https://press.princeton.edu/books/paperback/9780300078152/seeing-like-a-state
Subscribe to EuroLuminant for independent European journalism.
