Chapter 12 — Those Who Refuse to Be Classified
A category appears solid while each person receives its consequences alone. One person challenges a surcharge, another a medical record and a third a refusal of cover. Each file seems particular. The category becomes less obvious when the people concerned compare their experiences and discover that the same rule has organised their difficulties.
In the early 1980s, riders of motorcycles and other powered two-wheelers in France often struggled to obtain insurance at an affordable price. The market gathered them under a familiar figure. The motorcyclist was assumed to be young, reckless and exposed to serious accidents. Insurers presented the premium as the translation of a cost.
Riders did not deny the falls or the vulnerability of a body without the protection of a car. They challenged the way an entire practice was reduced to a category that almost automatically led to a high premium, cancellation or refusal.
The movement of angry motorcyclists did more than ask insurers to correct their prices. It organised training, developed knowledge about riding and tried to create an institution able to cover its members. In 1983, this mobilisation led to the creation of the Mutuelle des Motards [1].
The people whom the market had grouped together as bad risks became member-policyholders. They were insured by the mutual and also belonged to the organisation that owned it. They had not abolished calculation. They had gained a voice over cover, prevention and the use of common resources. Refusing the price assigned to them had led them to take back part of the rule that produced it.
An individual appeal remains essential. An address may be wrong, a claim may have been attached to the wrong person or a model may describe the situation badly. Correcting the file can reduce a premium or reopen access to cover.
That victory leaves the rule intact. To win, the person often has to show that they do not resemble the other members of the disadvantaged category. They leave the bad group without questioning its existence.
A challenge takes on a different meaning when the person no longer says only that their file was misread. They show that the same rule repeatedly produces the same wrong. The question then extends beyond data accuracy. It concerns the way the problem was defined, the evidence accepted, the consequence attached to the classification and the alternatives the institution had set aside.
From an Individual File to a Common Wrong
The effects of an insurance policy are often scattered. One household receives an increase, another a narrower offer and a third no answer at all. An owner learns that the policy will not be renewed. A tenant encounters the consequence later through a rent increase or the sale of the building. Each situation can be explained by a local reason. While the files remain separate, no one knows whether the problem is recurring.
The insurer sees something different. It knows its acceptance rates, non-renewals, territorial differences and the categories in which adverse decisions accumulate. This asymmetry often determines who can recognise a policy in what looks to those affected like a series of isolated misfortunes.
The first task of a collective is to bring together what the organisation treats separately. Letters are compared, testimonies collected, reasons coded and maps annotated. A common wrong does not arise simply because several people suffer. It becomes visible when their experiences can be connected to the same rule.
In October 1987, Julius Joseph, Samuel Parker and John Moore contacted a lawyer and the Milwaukee branch of the National Association for the Advancement of Colored People. The three Black insurance agents had more than fifty years of experience between them. They described practices that restricted access to homeowners insurance in Black neighbourhoods.
The problem did not always take the form of an explicit ban. Some calls went unanswered. Vague criteria were applied differently from one neighbourhood to another. Maps and instructions directed agents away from certain areas. Residents knew that obtaining cover was difficult. They did not yet possess the documents needed to show that the difficulty formed a pattern.
A redlining committee brought the evidence together. Residents described their attempts to obtain motor and homeowners insurance. Former agents explained internal instructions. Paired tests sent otherwise similar applicants to the same providers and compared the responses. Statistics revealed the geography of policies. No single form of evidence would have been enough [2].
The action against American Family involved the NAACP, the American Civil Liberties Union, residents and the United States Department of Justice. The settlement reached in 1995 provided $14.5 million. It also required changes to underwriting rules, the rules used to decide which risks to accept and on what terms, as well as new agent locations in the affected neighbourhoods, development targets, loans and support for home repairs [2].
The mobilisation had not merely demonstrated unequal treatment. It had helped define the material conditions for a return to the ordinary insurance market.
A statistical category does not become a collective by itself. The people it gathers may not know one another. They may not even realise that the same rule applies to them. Older rating classes at least had names. An individual score gives each person a result assembled from many changing attributes. It is harder to connect that result to a shared experience [3].
Associations, meeting places, aggregate data and common accounts are needed to move from the file to the collective wrong. The group may form during the dispute. Its members do not necessarily share a prior identity. They share the fact of being affected by a decision and the wish to intervene in the definition of the problem [4].
In insurance, this intervention often comes late. The product has been designed, the rating system adopted and the model built into routine procedures. Discussion begins when the increase or refusal has already reached the contract. Publishing refusal rates, reasons for non-renewal and territorial effects would allow action earlier, before isolated decisions accumulate into a boundary.
Bringing Other Knowledge into the Room
Challenging a classification sometimes requires learning the language of those who built it. The AIDS movement offers an important example. By the late 1980s, ACT UP activists were not only denouncing slow research and poor access to treatment. They were discussing virology, clinical trials and the rules used to recruit patients. AZT, or azidothymidine, was among the first antiretroviral medicines used against HIV, the human immunodeficiency virus [5, 6].
Activists gained scientific credibility without becoming assistants to researchers. They challenged trials conducted on highly selected populations. Excluding people who had already received some treatments could make analysis easier, but it also reduced the reach of the results and sometimes kept those most in need out of the trial. Activists helped change the criteria used to decide which knowledge and which study populations were acceptable [6, 7].
Counter-expertise means the capacity to examine the questions, methods and conclusions of official expertise. It does not claim that every personal experience is equivalent to a scientific study. It shows that technical skill alone cannot decide which effects deserve to be measured or which solutions should be considered [8, 9].
Different forms of knowledge concern different parts of a situation. An actuary knows claim frequency, selection and the cost of capital. A resident knows which roads become impassable, which work has already been completed and who will be unable to evacuate. A policyholder knows what a deductible means for a household budget, how an illness has changed and why recorded behaviour may not reflect a free choice.
The expert usually answers a precise question. They estimate a cost, assess damage or measure the effect of a variable. The quality of the work depends on how well that question is answered. The group concerned may show that the question is too narrow. It asks what will happen after a refusal, who produced the exposure and why the means to reduce the risk were not financed. Research carried out in laboratories and organisations then meets knowledge produced in direct contact with the situation [4].
This encounter reveals effects that the model left outside its frame. A flood map measures water depth but may not show that a household has no car or includes a dependent person. A health model predicts expenditure but does not count the work transferred to relatives when care is refused. An underwriting rule describes a building without tracing the decades of disinvestment that made repairs unaffordable.
Learning institutional codes opens a door and creates a risk. To be heard, a collective adopts specialised vocabulary, evidence formats and timetables. This translation can distance its representatives from people who do not have the same time or training. ACT UP experienced this tension. Expertise enabled intervention, but it also created divisions within the movement [6].
The answer is not to reject competence. Knowledge must be shared, spokespersons must remain replaceable and decisions must remain open to those who will bear their consequences. Counter-expertise does not succeed when it replaces the institution’s specialists with a small group of permanent specialists from the movement. It succeeds when it changes the questions that can be asked and the conditions under which evidence circulates.
Producing Data against Data
Institutions classify with the information they possess. Collectives often begin by seeking access to refusal rates, territorial differences, variables and decision histories. They then discover that some decisive information does not exist. No database records every call that received no answer or every person who gave up before submitting an application.
Other costs remain outside the files. Time spent on procedures, repairs paid without making a claim, unpaid family work and consequences for tenants are rarely connected to the insurance decision. Counter-data make some of these absences visible. A collective can record refusals, map non-renewals, organise paired tests or follow the effects of a score over time.
These numbers do not speak for themselves any more than the insurer’s numbers do. Cases must be selected, testimonies checked, records cleaned and limitations explained. This work requires resources and reflects values. Its force does not come from a claim to offer pure truth. It comes from challenging the organisation’s monopoly over the description of its own effects [10].
Data activism refers to practices through which citizens produce, reuse, challenge or protect data. Some build databases and display disparities. Others limit collection, obscure a trace or prevent it from circulating [11]. This second form matters. Resistance does not always mean giving an institution more information about oneself.
A person may want an address corrected while disputing its use as a substitute for a history of segregation. A collective may ask for a more accurate map while refusing automatic withdrawal from every area it marks as exposed. An old illness may be recorded correctly without being an acceptable reason for a higher premium. Accuracy does not settle the legitimacy of use.
Algorithmic categories create profiles designed for an operation. They help sell, direct, monitor or decide. They do not necessarily seek to recognise the identity a person claims for themselves [12]. Offering a fuller account of who one really is can therefore leave the main question untouched. The conflict concerns the authority to turn traces into consequences.
Challenging a system does not always mean asking for a more accurate or better trained model. The disagreement may concern the task itself, the information the institution claims the right to use or the consequence attached to the result. Rejecting automation does not mean rejecting all technical tools. It may mean that this decision should not be delegated, or that this classification should not produce this price, scrutiny or refusal [13].
Counter-expertise must therefore be able to reach two different conclusions. The measurement may be poor and require correction. It may also be accurate enough but unsuitable for the proposed use. Rights to be forgotten and zones of non-conversion protect this second possibility. A democracy of risk does not demand that every difference be measured more precisely. It also decides which differences will not become prices.
Producing counter-data takes time, skill and legal protection. People already weakened by refusal often have the fewest resources with which to become experts in their own exclusion. Checking records, understanding a model and lodging an appeal become additional work. When that burden rests entirely on the individual, failure may be mistaken for acceptance [14].
Access to data, support from associations, adequate deadlines and independent expertise determine who can genuinely take part. They prepare the next step, when the knowledge produced must enter a procedure capable of changing the rule.
Entering the Procedure
An ordinary appeal begins after a premium increase, refusal or interruption of a benefit. It may correct an error, but sometimes comes too late to prevent a sale, a loss of care or departure from a neighbourhood. Effective participation must be possible before the decision is fixed.
California’s Proposition 103, adopted in 1988, allows consumer organisations to intervene in the examination of some insurance rate applications. They may recover their costs when their participation makes a substantial contribution to the proceeding [15, 16]. The system does not guarantee low premiums. It prevents the insurer’s expertise from being the only source used to define the relevant evidence, the effects of the rate and the possible alternatives.
This counterweight must also account for its own practices. A reform debated in 2026 proposed clearer standards for substantial contribution, closer review of fees and wider publication of documents. Consumer organisations feared that making funding harder to obtain would reserve technical proceedings for groups that already had substantial resources [17, 18].
Professionalisation can be expensive and can produce its own routines. It may still be necessary when the other party has permanent legal, actuarial and regulatory teams.
A serious challenge brings together several capacities. The rule and its effects must be known. Intervention must occur before the consequence becomes irreversible. The collective must then be able to act on the purpose, data, thresholds and alternatives [19]. Transparency without power merely leaves people better informed about what is happening to them.
Collective action also protects those who do not wish to reveal a diagnosis, conviction or financial condition in public. An association, trade union or municipality can carry the dispute without making every individual the public face of their vulnerability.
Participation is not yet governance. Consultation after a model has been built can absorb criticism without changing the decision. Participation becomes political when it can alter the definition of the harm, the information admitted, the consequence and the composition of the body that decides [20, 21].
From Counter-Expertise to Counter-Institution
An association can document refusals, argue for broader rating classes or defend an obligation to offer cover. It does not necessarily have to establish reserves, buy reinsurance or maintain supply after a bad year. Reserves are funds set aside for future claims. Reinsurance is cover purchased by an insurer to share unusually large or concentrated losses.
This difference does not make the criticism illegitimate. It shows what changes when a collective decides to carry a promise itself.
The Mutuelle des Motards crossed that threshold. From 1981, tens of thousands of people financed the creation of an insurer before the first policy was sold. By September 1983, around forty thousand subscribers had supplied the necessary resources [1]. They were not yet buying cover. They were financing the possibility of offering it.
Collective ownership removed none of the constraints of insurance. The mutual had to estimate losses in a population it knew imperfectly, set contributions, negotiate reinsurance and retain enough capital for difficult years. The members who had rejected a category imposed by the market had to distinguish among motorcycles, uses, experience and riders themselves.
Their advantage did not lie in abolishing classification. It lay in more situated knowledge of the practice and in the ability to debate how that knowledge would be used. The mutual could connect prevention to training that riders could actually attend, examine difficult files with elected representatives and adapt cover to the particular vulnerability of two-wheelers [1].
A crisis at the end of the 1980s also showed the limits of the experiment. Poorly controlled diversification weakened the accounts. Mutual ownership did not protect against strategic mistakes or governance conflict. It did, however, allow an additional assessment, a further contribution requested from members when ordinary resources were no longer sufficient.
Members had to decide whether to finance an institution in difficulty. Other mutual organisations and reinsurers also supported its recovery [1]. Solidarity was no longer an abstract principle. It became money to be paid, a narrowing of the mutual’s activities and a debate over the decisions that had caused the crisis.
A counter-institution becomes credible when it remains answerable for its promise over time. It must say who will pay during a bad year, who will decide how balance is restored and how those classified by the new rules can still challenge them. Owning the institution does not remove the duty to account for its decisions.
Mutuals can be understood as user-owned firms created when ordinary market relations protect users’ interests poorly [22, 23]. Ownership changes who controls the remaining surplus and who holds final authority over the organisation. It does not guarantee equal prices or active democracy. Members do not examine every model, and management can move away from the movement that founded the institution.
Nor is a counter-institution limited to a mutual insurer. A data cooperative, a territorial prevention fund, a public utility or a community organisation able to finance repairs can take back part of the rule. What matters is the combination of three powers. The institution must be able to define needs, administer resources and remain responsible over time for what it has promised.
Few groups can create their own insurer. There are intermediate forms between an individual complaint and that solution. An association maintains a database of refusals. A trade union negotiates cover. A municipality finances work. A regulator gives affected groups a right to intervene. Knowledge can remain close to lived situations while capital and reinsurance are organised on a wider scale. Local autonomy would lose its meaning if it forced the group to bear alone a catastrophe beyond its capacity.
The Boundaries of “We”
Not every category becomes a collective easily. Motorcyclists shared meeting places, associations and a visible practice. AIDS activists had political networks. In Milwaukee, agents understood the internal practices of insurers. Other people may not know that the same rule has brought their situations together.
Individualised scores disperse the wrong. Tenants receive its effects indirectly through rent or the sale of a building. People with old illnesses or convictions may hesitate to organise around a history they are trying to leave behind. Personalisation can turn a collective effect into a series of apparently private results [3].
Public statistics, aggregate data and rights of collective action can help the group discover itself. This representation must include those who were refused. An appeal available only to policyholders already admitted to the pool leaves outside the people whom the rule excluded [24].
The collective has boundaries of its own. It may protect its members by shifting costs to a less organised group. A specialist mutual may improve cover for those who enter and become a closed club. A professional association may concentrate speech among a few experts. A counter-institution remains open only when the next groups it classifies still have the means to challenge it.
This requirement returns us to the difference between consumer choice and citizen judgement. Comparing quotations allows someone to choose among products. It does not decide which differences should remain pooled, which needs require a guarantee or which obligations institutions should accept [25]. Taking back the rule means moving from a choice among offers to a discussion about the architecture of protection.
People who are classified no longer seek only to prove that they do not belong to the bad group. They can turn an assigned risk into a shared experience and build the means to act. A legal victory, a counter-database or a new mutual does not, however, solve the problem of time.
Infrastructure must be maintained, work financed and the benefits of prevention preserved when the policyholder, contract or owner changes. Counter-expertise can alter a decision. Preventing losses over the long term requires resources and an institution able to recognise the value of an event that may never happen. That is the problem of the next chapter.