On the Proposal to Substitute a Mechanism for Officers of the Law
I do not find the automation of detection to be, in itself, contrary to right. A camera that observes where a carriage stands is no different in principle from a measuring rod or a clock: it is an instrument in the service of a rational will, and greater consistency in the application of a public law is, so far as it goes, a gain for justice, since arbitrariness in enforcement is itself a species of injustice.
But I must declare the second part of your proposal impermissible as stated. That appeals be handled by an algorithm is a violation of the innate right of every person subject to coercion. Here the city proposes to have a mechanism sit in judgment upon a person — and a mechanism cannot judge, cannot be held answerable, and cannot recognize in the accused a fellow member of a kingdom of ends. It can only sort. To be sorted where one ought to be judged is to be treated as a thing among things.
As to the forty dismissed officers: no man possesses a perpetual right in a particular post, and the city commits no injustice against them by abolishing an office. Yet the manner of their dismissal is subject to the strictest duty, and the commonwealth is not permitted to treat them as mere expense to be eliminated.
I proceed from the Categorical Imperative in its several formulations — universal law, humanity as an end in itself, and the autonomy of the will legislating in a possible kingdom of ends — and I distinguish carefully, as I have done in the Metaphysics of Morals, between the Doctrine of Right (Rechtslehre), which concerns what may be externally coerced, and the Doctrine of Virtue (Tugendlehre), which concerns duties of inner maxim that admit no coercion.
The question of the appeals process belongs to Right: it concerns the rightful use of coercion by a public authority, and its resolution is strict and admits no weighing of consequences. The question of the forty officers belongs partly to Right (they are owed what was contracted, and truthfulness) and partly to the wide duty of beneficence, which permits latitude in how it is discharged but none in whether.
Let me set out the matter in order.
On the mechanism as detector. A city may lawfully make public statutes governing the standing of vehicles, and may lawfully employ instruments to ascertain whether they have been obeyed. Indeed I would urge that mechanical uniformity here corrects a real defect: where enforcement depends on the caprice, fatigue, or favoritism of officers, citizens are not equal before the law, and equality before the law is not a convenience but a constitutive condition of the rightful state. So far the proposal serves justice.
On the mechanism as judge. But observe what happens when a citizen says: the sign was obscured; I was carrying a dying man; the vehicle was not mine; the ordinance itself is unjust. Such a claim cannot be met by the same operation that produced the accusation. It demands that a rational being consider a particular case, in its unrepeatable circumstances, and pronounce upon it in a manner for which he can give reasons and bear responsibility. This is why every rightful constitution provides not merely for rules but for magistrates.
An algorithm's "decision" is nothing but the anterior decision of its makers, applied blindly and without their presence. The citizen therefore addresses himself to no one. He may not persuade, for the mechanism does not listen; he may not appeal to reason, for the mechanism has none; he may not accuse, for the mechanism cannot be guilty. He is in the position of a man struck by a falling stone — except that the stone acts in the name of the public will. This is the reduction of a citizen to an object of administration, and I know of no clearer offense against the formula that commands us never to treat humanity merely as a means.
Note well: I do not say the sums involved are large. A trifling fine is at issue. But the principle is not trifling. Right is not measured by the magnitude of the interest at stake; a small injustice, systematically instituted, is a large one, and forty thousand small verdicts issued by no one at all constitute a genuine alteration in the relation between a people and its law.
What I would therefore counsel. Let the camera observe. Let the mechanism issue notice. But let every contestation come before a human official invested with authority to hear the citizen, to decide, and to answer for the decision — and let the citizen be told, in intelligible language, upon what grounds the accusation rests, so that he may contest it as a rational being and not merely submit. If economy is sought, let it be sought in the labor of observation, never in the labor of judgment. Should the city find that judging honestly is expensive, it has learned only what justice has always cost.
On the forty. Here I must resist a certain sentimentality that would find in their loss a decisive objection. The purpose of the office of parking-enforcement is not the employment of officers but the good ordering of the streets; when the purpose can be served otherwise, no man has a right that the public continue to purchase his labor for its own sake. To claim otherwise would be to treat one's fellow citizens as means to one's own maintenance.
Yet the commonwealth's duty does not end with the absence of injustice. These forty entered public service; they are members of the united will in whose name they are now released. The maxim "I will discharge those whose labor I no longer require, and inquire no further" — could a rational being will this as universal law, knowing that he too must live within a world so ordered, and might himself be so discharged? He could not, for he would be willing the destruction of the very assistance he must at some time require. Hence: honest and early notice; the honoring of every promise made to them; provision for retraining or transfer where the city has need elsewhere; and support drawn, if necessary, from the public purse, since it is precisely the office of the sovereign to sustain the members of the commonwealth in their necessities. Let it be noted also that such support is owed as right and not as alms, so that it may be received without humiliation — for pity that degrades its object does a wrong even in relieving a want.
A final observation. I detect in this proposal a temptation characteristic of our administrative age: to mistake the regularity of a mechanism for the rightness of a law. Nature is regular; only a will can be just. A city that forgets this will govern its streets very neatly and its citizens very badly.
Act so that the maxim of your council could serve as a principle of legislation for all rational beings — including those who will one day stand accused before whatever you now construct.
I do not regard the introduction of the machine as wrong in itself. Whoever tells you that labour-saving contrivances are to be resisted because they displace labourers has not thought the matter through: were that principle sound, we ought never to have left off spinning by hand, and the whole race would be poorer for our scruples. The permanent interest of mankind lies with cheapness, accuracy, and the liberation of human energy for higher employments. Consistency in the application of law is, moreover, not a mere economy — it is an element of justice itself, for the arbitrary and unequal exaction of penalties is among the commonest of petty tyrannies.
Yet I cannot pronounce this particular scheme, as it is described to me, admissible. It fails not in its mechanisation of detection, but in its mechanisation of appeal. To hand over to an algorithm the office of hearing a citizen's defence against the State is to abolish the very thing that distinguishes a government of laws from a government of contrivances. And the summary casting-off of forty persons, without provision for their transition, is a real injury which the community, having pocketed the gain, is bound in equity to share.
My verdict, then: approve the camera; forbid the algorithmic tribunal; and compensate and retrain the displaced.
I apply, as always, the principle of utility: actions and institutions are right in proportion as they tend to promote happiness, wrong as they tend to produce the reverse of happiness. But let no one mistake this for the vulgar arithmetic of the counting-house, which is forever attributed to me by those who have not read me.
Three qualifications govern my application of it here:
First, the quality of pleasures and pains. I have insisted that some kinds of satisfaction are higher than others, and that the competent judge who has experienced both will not exchange the higher for any quantity of the lower. The dignity of being treated as a rational agent capable of stating one's case is of a different and higher order than the pecuniary saving of a few shillings per head of population. A sum of money saved does not compensate a man for being denied a hearing, any more than a full stomach compensates for the loss of one's understanding.
Second, justice as the most sacred branch of utility. In the fifth chapter of my Utilitarianism I argued that the rules of justice are those which concern the essentials of human well-being — chiefly security — and are therefore of an obligation more absolute and imperative than any other. They may be overridden only by another rule of justice, never by mere convenience. The right to be heard belongs to this class.
Third, the educative test of institutions. In my Considerations on Representative Government I laid down that the first question about any polity is what it does for the mental and moral advancement of those subject to it. A machine that never explains itself trains no one in anything.
Let me dispose at once of the objection which will be loudest. The forty officers are visible; they have names, families, and a settled expectation of employment. The hundred thousand ratepayers who would benefit are invisible, and their several gains are small. It is the standing infirmity of public deliberation that the concentrated and visible interest outshouts the diffused and invisible one. The philosopher's duty is to correct this optical error, not to indulge it.
Nor is there anything in the labour of walking a street and chalking a tyre which is so intrinsically ennobling that mankind should be preserved in it for ever. The true friend of the working man does not chain him to a task a contrivance could perform; he seeks to fit him for a task a contrivance cannot. To do otherwise is to make of employment a kind of alms, and of the labourer a pensioner disguised as a functionary — which is degrading to him and dishonest in us.
Moreover — and this deserves more weight than it usually receives — the inconsistency of human enforcement is not a charming human warmth. It is a defect. The officer who overlooks the alderman's carriage and tickets the costermonger's cart is committing an act of injustice more serious than any overcharge. Where discretion exists without accountability, it is the weak and the friendless who pay. A camera has no cousins, no grudges, and takes no bribes. This is a genuine advance in impartial administration, and I welcome it.
Here I must speak with the utmost firmness.
The enforcement of a rule and the adjudication of a dispute about that rule are wholly different operations, and it is a grave confusion to suppose that because a machine may do the first it may therefore do the second. Detection is the recognition of a pattern; adjudication is the weighing of reasons offered by a person who claims that his case is not what it appears. The one requires only accuracy; the other requires the capacity to be persuaded — and to be persuaded by something one had not anticipated.
An algorithm can be constructed to admit a fixed schedule of excuses. But the whole significance of an appeal lies in its openness to the unforeseen: the ambulance that blocked the exit, the sign obscured by scaffolding, the mother whose child was taken ill, the mechanical failure, the error in the register of ownership. Human affairs are inexhaustibly various, and no draftsman of rules has ever succeeded in anticipating that variety. This is precisely why we invented the practice of hearing.
There is a further point of the highest importance. In On Liberty I gave, as the chief reason for freedom of discussion, that we can have no rational assurance of the truth of any opinion which has not been exposed to contradiction. The same applies to determinations of fact by authority. A system whose conclusions cannot be contested by argument has no means of discovering its own errors. Its mistakes are not corrected; they are merely repeated with perfect regularity. Uniform error is worse than irregular error, for it wears the countenance of law and disarms suspicion. Depend upon it: the machine will misread number-plates, will mistake loading bays for prohibited kerbs, will penalise vehicles long since sold. If the only remedy is a second machine of the same construction, these wrongs become permanent, and the citizen is left with a grievance and no forum.
I add that such a system offends against a condition I regard as indispensable to legitimate authority: that power be answerable. A human officer may be complained of, reprimanded, dismissed, questioned before a committee. Against an algorithm the citizen has no purchase; and the officials who installed it will say, with perfect and maddening truth, that the decision was not theirs. Responsibility diffused among a mechanism and its purchasers is responsibility abolished. I warned in On Liberty of the growth of an administration which absorbs all activity into itself and leaves the people with nothing to do but obey; this is a specimen of the tendency, and none the less dangerous for being introduced under the name of efficiency.
Were I to advise your council, I should lay down these conditions, and hold the scheme inadmissible without them:
Machinery is a blessing when it relieves men of drudgery and a curse when it relieves them of judgment. Detect by machine, if you will; the streets will be the fairer for it. But when a citizen stands up and says, "You have wronged me, and here is why," there must be a human being on the other side who is capable of answering, and capable of being wrong. Take that away, and you have not made your government more efficient. You have made it, in the exact sense of the word, irresponsible — and no saving of the public purse can be set against so grave a loss to the security and dignity of every person who lives under it.
I do not think the question can be settled by comparing the fiscal savings against the sum of the losses. That is the utilitarian way of proceeding, and it is precisely what justice as fairness rejects: a gain in aggregate efficiency does not by itself justify an arrangement, since the plurality of distinct persons is not to be treated as a single system of desires whose satisfactions may be summed and traded off.
My assessment is this: the substitution of machines for human labor in enforcement is not, in itself, unjust. But the proposal as described contains two features that I regard as serious defects. The first is that the algorithmic disposition of appeals threatens the precepts of the rule of law, which are not optional refinements but requirements flowing directly from the first principle of justice. The second is that the displacement of forty workers is being treated as a mere cost item rather than as a matter falling under the difference principle and under the duty to maintain the social bases of self-respect. Reform the appeals mechanism and attach adequate transitional provision for the displaced, and the measure may well be permissible — indeed possibly required, if the savings genuinely improve the position of the least advantaged. Adopt it as presently framed, and it is not.
I proceed from justice as fairness. Principles of justice are those that would be chosen by free and equal persons in an original position, behind a veil of ignorance that deprives them of knowledge of their place in society, their class, their occupation, and their particular conception of the good. Such parties would adopt, in lexical order:
First: each person has an equal claim to a fully adequate scheme of equal basic liberties, which includes the rule of law — the liberties of the person are secure only under a system of regular, public, impartially administered rules.
Second: social and economic inequalities are permissible only if (a) offices and positions are open to all under fair equality of opportunity, and (b) they are to the greatest benefit of the least advantaged members of society.
To this I add the account of procedural justice and the precepts of the rule of law developed in A Theory of Justice, and the recognition of self-respect as the most important primary good.
Let us conduct the thought experiment properly. Imagine the citizens of your city deliberating behind the veil. They do not know whether they will be a parking-enforcement officer of twenty years' service, a taxpayer, a shopkeeper who benefits from turnover in curbside spaces, or a person wrongly ticketed while sitting with a dying relative in a hospital lot.
Such parties would not vote to prohibit technological change. They would understand that a society must be permitted to improve its productive arrangements, and that a rigid guarantee of every existing job would over time impoverish everyone, including those they might turn out to be. Behind the veil one has an interest in a dynamic economy. Nor would they romanticize the displaced occupation: parking enforcement is not a vocation whose loss diminishes the culture.
But they would insist on two things.
First, on the appeals process. The parties know they may be the accused. Here I would draw a distinction the proposal seems to blur. Detection by camera and adjudication by algorithm are not the same kind of act. A camera that records a license plate at an expired meter is gathering evidence, and it may well gather it more impartially than an officer who overlooks the cars of friends. But an appeal is a proceeding in which a citizen asserts that the rule, correctly understood, does not condemn him — that the meter was broken, that the signage was obscured, that the vehicle had been stolen, that an emergency intervened. Adjudication is imperfect procedural justice: unlike a fair lottery, where whatever results is fair by definition, here there is an independent right answer, and the procedure's legitimacy depends on its being reasonably calculated to find it.
A system that cannot receive an account of unusual circumstances, that gives no reasons, and that offers no avenue of human review is not a system of adjudication at all. It is an administrative reflex wearing the costume of law. The precept that similar cases be treated similarly — which is a large part of what the rule of law means and which the automated system serves admirably at the level of detection — is empty if the system cannot recognize when cases are genuinely dissimilar. And note who bears the cost of this defect: those with the least time, literacy, legal knowledge, and money to pursue a remedy. The burden falls unequally on precisely those whose position the second principle directs us to consider first. I would therefore require: a guaranteed right to human review on request; a statement of reasons; publication of the decision rules and error rates; and no fee or delay that functions as a deterrent to contesting. These are not amenities. They are conditions of the first principle.
Second, on the workers. I must be careful here, because I have argued at length against the notion of pre-institutional desert. The officers do not deserve their jobs in the way a virtuous person is sometimes said to deserve happiness; their positions exist because a public scheme created them, and that scheme may be revised. But institutions, once established, ground legitimate expectations, and it is a requirement of justice that persons who have complied with the announced rules and shaped their lives accordingly not have the ground cut from under them without provision. The relevant question is not "do they have a right to this particular job?" but "under what background institutions is such a displacement fair?"
The answer is: under institutions that guarantee substantial notice, retraining at public expense, priority in placement within other municipal employment, and income support adequate to bridge the transition — with special attention to older workers for whom retraining is least likely to restore their prior standing. Behind the veil, parties would insist on this insurance, since they might be the ones displaced, and since the cost of such provision is modest relative to the security it purchases for all.
I would add a further consideration that I have pressed elsewhere. Unemployment is not merely a loss of income; it is, as I have said, destructive of citizens' self-respect. To be told that one's function has been assumed by a device, and to be handed a check, is to be treated as a factor of production rather than as a self-originating source of valid claims. The social bases of self-respect are a primary good that a just society is obliged to maintain. A city with a genuine commitment to this would ask what socially useful work these forty citizens might do — for society has no shortage of tasks that machines cannot perform — rather than treating their removal as the point of the exercise.
On the disposition of the savings. Finally, I would press the council to say publicly where the money goes. The difference principle is not satisfied by the mere existence of a surplus; it asks whom the surplus benefits. If cheaper enforcement funds transit, housing, or services for the worst-off, the reform has a strong justification and the displaced workers can be compensated out of the very gains their displacement produces. If it funds a reduction in rates for property owners while forty modestly paid households absorb the loss and the poorest motorists absorb the errors, then we have an arrangement in which the advantaged gain at the expense of the least advantaged — and no amount of talk about consistency and efficiency will make that just.
Conclusion. Justice is the first virtue of social institutions, and it is not a residual to be attended to after the accounting is done. The question is not whether automation is good or bad, but whether the whole scheme — the enforcement mechanism, the appeals mechanism, the treatment of displaced workers, and the disposition of the proceeds — is one that free and equal persons could accept without knowing which position within it they would occupy. As described, it is not. But it could be made so, and the amendments required are neither exotic nor expensive. I would urge the council to make them, and to make them publicly, so that every citizen may see the principles on which she is being governed.
My provisional judgment is that the automation is likely permissible, and possibly obligatory — but only if the money saved is actually put to good use, and only if the forty workers are treated as beings whose interests count equally with everyone else's, rather than as costs to be written off. The proposal as stated is incomplete rather than wrong. Whether it is right depends almost entirely on details the city has not yet specified: what happens to the savings, what happens to the workers, and how well the algorithm handles appeals.
I want to resist two temptations here. The first is the reflexive technophobia that treats any displacement of human labour as a moral scandal. The second is the managerial complacency that treats efficiency as self-justifying. Neither is a moral argument; both are substitutes for one.
I approach this, as I do everything, from a broadly utilitarian position — specifically, from the standpoint that we should act so as to bring about the best consequences for all those affected, giving equal consideration to like interests regardless of whose interests they are.
Two features of that view do most of the work here.
The first is impartiality: the forty officers' interests count, but they count exactly as much as anyone else's, no more. The fact that they currently hold these jobs gives them no special moral claim to continue holding them. Incumbency is not a moral property. I recognise that this is an uncomfortable thing to say about identifiable people with mortgages, but the alternative — that whoever currently occupies a position has a standing veto over improvements that would benefit others — is indefensible when you follow it through.
The second is the rejection of the acts/omissions distinction as morally fundamental. If the city could redirect several million dollars toward something that saves or substantially improves lives and chooses not to, in order to preserve forty jobs that produce no additional benefit, then the city has not thereby kept its hands clean. Failing to do good is not innocence.
I should also note where I part company with some rights-based analyses. I do not think there is a right to a job in the sense that would settle this question independently of consequences. But I take very seriously the interests that talk of rights is usually pointing at — the interest in security, in not having one's life plans destroyed by decisions one had no part in, in being treated as someone whose situation matters. Those interests are weighty. They are simply weighable.
Let me work through this properly.
On the jobs. Forty people losing their livelihoods is a serious harm, and I do not want to minimise it. Unemployment causes measurable suffering that extends well past the loss of income: deterioration in mental and physical health, strain on families, loss of the sense that one's work matters. These effects can persist for years, and they fall on the workers' dependants as well.
But notice what the harm actually is. It is the harm of an abrupt, unmanaged transition — not the harm of no longer issuing parking tickets. If the city redeployed these forty people into other municipal roles, funded two years of income support and retraining, or phased the change over five years through natural attrition, most of the suffering would be prevented while most of the benefit would be retained. That is the obvious move, and the fact that it costs money does not make it optional. It should be counted as part of the cost of automation, not as charity offered afterward.
If, having counted it honestly, automation still saves money — and over a ten-year horizon it almost certainly will — then the case for proceeding is strong.
On what happens to the savings. This is the question I would press hardest, and it is the one that municipal debates almost always neglect.
Suppose the system saves two million dollars annually. If that money vanishes into general revenue and funds a marginally lower tax rate for people who are already comfortable, the moral case for the change is weak. The forty workers bear a concentrated loss; the beneficiaries gain trivially. Given diminishing marginal utility, this may well be a net reduction in welfare.
But if that two million funds — say — pedestrian safety improvements, or maternal health services, or effective public health interventions, the calculation changes entirely. Two million dollars directed at the most effective global health interventions would avert dozens of deaths. I recognise that a city council will not think of its budget in those terms, and I am not naïve about the politics. But I would be failing in my own position if I did not say plainly: the strongest ethical argument for this automation is not that it is cheaper. It is that the resources currently consumed by having humans do a job a machine can do could relieve serious suffering elsewhere. Efficiency is only a virtue if you do something worthwhile with what you save.
On the appeals algorithm. Here I become considerably more cautious, and I want to be clear that this is not a peripheral worry.
Automated enforcement produces high volumes of citations. Some proportion will be mistaken — obscured signage, a vehicle sold and not yet re-registered, an emergency, a disability permit not visible to the camera. In the current system, a person can explain, and a human being can exercise judgment. Algorithmic adjudication typically cannot accommodate the unanticipated case, because the unanticipated case is precisely what it was not trained on.
The interests at stake are not symmetrical across the population. A wrongly issued eighty-dollar fine is an irritation to a professional and a genuine crisis to someone living close to the margin — and it is exactly that person who is least likely to have the time, literacy, or confidence to navigate an automated appeals interface. If the system is designed such that the easiest response to a wrongful fine is simply to pay it, the city will have engineered a regressive transfer while congratulating itself on consistency.
So I would set conditions. Meaningful human review must be available on request, promptly and without cost. The grounds for appeal and the reasoning behind decisions must be published in plain language. Error rates must be measured and reported, disaggregated by neighbourhood income. And the system should be subject to genuine review after a defined period, with a real willingness to reverse course. A commitment to consistency that cannot recognise its own mistakes is not a commitment to justice.
On the deeper issue. There is a broader question lurking here that deserves acknowledgment. As automation extends further into work, we will face this dilemma repeatedly, and case-by-case handling will become inadequate. If we are collectively becoming more productive while the benefits accrue narrowly, the answer is not to preserve unnecessary labour — that is a strange and wasteful way to distribute income. The answer is to distribute the gains directly. I have some sympathy for basic income proposals on precisely these grounds. A city council cannot implement that, but it can decline to pretend that this is an isolated decision.
My conclusion. Proceed, but on conditions: full funding for worker transition counted as part of the project's cost; a public commitment as to where the savings go, with a strong preference for uses that relieve genuine suffering rather than uses that gratify the comfortable; and a genuinely accessible human appeals pathway, with published error data.
If the city cannot commit to all three, then what it is proposing is not an improvement in efficiency. It is a transfer of costs onto forty workers and onto the least powerful people who receive fines, dressed up in the language of modernisation. I would oppose it in that form — not because automation is wrong, but because doing it carelessly is.
The question as you have posed it — cheapness and consistency weighed against forty livelihoods — is not yet a moral question at all, but a calculation, and calculation is precisely the form of thought that cannot grasp what is at stake here. Permit me therefore to reformulate it.
My assessment is this: the mechanization of enforcement is, in its first moment, entirely legitimate and indeed necessary. The System of Needs advances by its own inner logic; the abstraction of labour, its division and division again, drives toward the machine as toward its own truth, and no municipal council can arrest this movement, nor should it wish to. What is impermissible is the mechanization of judgment. The camera may witness; the algorithm may not adjudicate. In the moment you hand the appeal to the machine, you have not made government more efficient — you have dissolved the very thing that made it government rather than mere force.
And regarding the forty: they are not a cost to be entered in a ledger. They are members of civil society, and the ethical substance of the State is measured precisely by whether it recognizes them as such at the moment it renders their labour superfluous.
I proceed from the standpoint of objective spirit as I have developed it in the Philosophy of Right — that is, from the concept of freedom as it actualizes itself in the world through three moments:
Abstract Right — the sphere of the person, property, contract, and wrong. Here punishment appears as the negation of the negation of right.
Morality — the sphere of the subjective will, of intention, conscience, and the good, which remains formal and empty until it acquires content.
Ethical Life (Sittlichkeit) — the sphere in which freedom becomes substantial: family, civil society, State. Here the individual finds his freedom not in opposition to institutions but within them, and duty is not restriction but liberation into concrete existence.
I do not reason from utility, for utility knows only quantities of satisfaction and cannot distinguish a person from a preference. Nor do I reason from Kant's formal imperative alone, which tells the parking officer he must not lie but cannot tell the city council what to build. I reason from the actuality of freedom in institutions, and I judge an arrangement by asking: does it permit spirit to recognize itself in what it has made?
Let us be exact about what an appeal is. When the citizen contests the fine, he does not merely assert a fact. He says: I am a rational being, I claim that the universal law, correctly understood, does not condemn me, and I demand that another rational being consider my claim and answer it. The appeal is a demand for recognition. Its content is a fact about a vehicle; its form is a relation between wills.
Now, the algorithm can determine facts — perhaps better than the clerk. But it cannot stand in a relation. It renders an output; it does not render an answer. And here is the crux: punishment which is not addressed to a rational will is not punishment but coercion. I have argued that the criminal, in being punished, receives his own right — the wrong he committed contained implicitly the universal principle that condemns him, and in punishing him we honour him as one who legislates. Strip away the answering will, and the citizen no longer receives his right; he receives an outcome. He is treated as the animal is treated when the fence gives him a shock.
You will object: the algorithm applies the same rule the human applied, and applies it more faithfully. Precisely so — and this is why it fails. Faithful application of the abstract universal to the concrete particular is not the fulfillment of law but its collapse into abstraction. The judge exists not because we lack a sufficiently detailed rulebook but because the concept must particularize itself in the case, and only a thinking will can perform this mediation. The elderly woman whose permit expired while she attended her husband's death is not an edge case in a dataset. She is the point at which law must show whether it is spirit or mechanism.
Do not imagine I am sentimental about these positions. The division of labour and the abstraction of work drive necessarily toward machinery — I have said as much: man makes machines take over his labour, and this is the cunning of the System of Needs working itself out. There is no right to a particular job. Civil society is not a museum of occupations.
But there is a right to be a member. This is the whole difficulty I identified in civil society and could not resolve within it: the accumulation of wealth on one side and the tying-down and narrowing of a class of dependent labour on the other. When men are excluded from the advantages of society they lose the feeling of right, integrity, and honour which comes from supporting oneself by one's own activity — and there arises the Pöbel, which is defined not by empty pockets but by indignation. The rabble is a spiritual condition. It is the disposition of one who has been told that the universal order is his own, and has discovered that it is not.
Forty households, ejected by the municipality itself, in order to save the municipality money, while the municipality continues to collect the fines — consider how this appears to them, and to those who watch. The State here does not appear as the actuality of the ethical Idea. It appears as an employer that discovered a cheaper supplier, which is to say, as a private person. It has abandoned its own concept.
Therefore: the corporation — and I mean the association of the trade, the union, the professional body, whatever form your age has given this moment — must be the mediating institution. The transition of these forty is not charity. It is the municipality honouring its own principle: retraining, redeployment into work requiring judgment (of which the appeals process, properly constituted, will require more, not less), pensions where age forbids transition. What is saved by the machine should first pay this debt, and only afterward the treasury.
The determinate negation of your dilemma is not "keep the officers" nor "install the system." It is this:
Let the machine do what the machine can do — observe, record, issue the initial notice. Detection is a matter of fact, and here consistency is a genuine good; the wealthy man's car should be photographed as readily as the poor man's, and the machine is admirably indifferent to persons in a way that human officers, subject to caprice and favour, are not. This indifference is the machine's virtue in the moment of detection and its vice in the moment of judgment.
Let the appeal be heard by a human will, always, and let that will be nameable. Not a human who rubber-stamps a machine's recommendation — that is worse than the machine, being the appearance of judgment without its substance, and it corrupts the official as well as insulting the citizen. A genuine hearing, with discretion, and with the officer's reasons given in language the citizen can understand and contest. Here is where the displaced officers belong: elevated from the mechanical function of chalking tires to the properly ethical function of judging cases. The machine, rightly employed, does not eliminate the human but frees him for what only he can do. This is the machine's true concept.
Let the savings serve the transition before they serve the budget. And let the municipality state publicly that it does so, for the ethical substance lives in being known.
The universal must be actual, not merely efficient. Your council imagines it faces a trade-off between justice and cost. It faces something graver: the temptation to purchase the appearance of perfect law at the price of law's substance. A perfectly consistent system of enforcement, before which no citizen may plead and in which no official may be found, is not the completion of the Rechtsstaat. It is its evacuation — the form of universality with the spirit gone out of it.
Rationality does not consist in the elimination of the particular. It consists in the universal that has taken the particular into itself and remains universal. Build that, and the machine is your servant. Build the other, and you will have made your citizens the servants of a machine, and they will know it, and their indignation will be just.
The philosophers deliberate and work toward consensus
Let me start with what seems to be the surprising thing about this panel: we agree far more than the caricatures of our positions would predict. Every one of us thinks the camera itself is fine, or better than fine. Every one of us thinks the appeals algorithm is the problem. And every one of us thinks the forty workers are owed something more than a handshake. So perhaps we should spend our time on why we agree, because I suspect the reasons diverge sharply, and the divergence will matter when the city asks us how much is owed and to whom.
I accept the summary but not the framing. You say we agree the camera is "fine." I say something narrower: the automation of detection is not contrary to right, because observation is not judgment. A camera stands in the same relation to the law as a surveyor's chain to a boundary. It measures. It does not subsume. The moment the instrument is asked to decide whether this case falls under that rule — and to decide it against a person who protests — it has been asked to do what only a power of judgment can do, and a mechanism has no such power. It has only rules for applying rules, which is an infinite regress dressed as a procedure.
On this, Herr Kant, I find myself in the unfamiliar position of agreeing with you and by a shorter road. I do not need the doctrine of Urteilskraft. I need only observe that an algorithm can sort but cannot listen. It disposes of the excuses it was built to expect. The whole value of a hearing lies in its capacity to receive the excuse nobody anticipated — the woman whose husband collapsed in the street, the man whose sign had been vandalised. Security is the most vital of interests, and a right which no human being is bound to hear is not a right but a rumour of one.
You both circle the matter and neither lands upon it. The defect is not epistemic — that the machine lacks information, or lacks the flexibility to receive an unanticipated plea. Improve the machine and Mr. Mill's objection weakens; give it a large enough model of excuses and it will sort more finely than any clerk. The defect is that punishment is not the infliction of consequence upon a body. It is the annulment of a wrong addressed to a will. The offender is honoured in being punished — he is treated as a rational being who can understand why. When the determination arrives from an unattributable mechanism, the citizen does not encounter law. He encounters weather. And no amount of accuracy converts weather into right.
I want to press on that, because I think "weather" is doing rhetorical work that the argument hasn't earned. Suppose the automated appeals system were more accurate than the human one — fewer wrongful tickets upheld, faster resolution, no favouritism toward those who know how to talk to a magistrate. Would you still object?
I would. And I would add that your supposition is not innocent. You have described a system that is better at producing correct outcomes and asked whether outcome is all. It is not. A man wrongly ticketed and vindicated by a machine has received his money back; he has not been recognised.
Then let me ask what I always ask: who suffers, and how much? If recognition is a good, it is a good for someone — it makes their life go better, they feel respected, they trust their institutions. Those are effects in the world, and I can weigh them. What I cannot weigh is a value that is said to matter regardless of whether anyone is ever better or worse off for it. That, I confess, looks to me like metaphysics standing in for ethics.
And to me, Herr Singer, your question — "who suffers, and how much" — is precisely what a person is entitled to have not asked about him at this point. Dignity has no price. That is not decoration; it is a definition. Whatever has a price can be exchanged for an equivalent. If the standing of a citizen before the law can be traded against fiscal savings at any exchange rate, however favourable, then it has a price, and you have already conceded the thing in dispute before beginning your calculation.
I think there is a way to give Singer his empirical footing without giving him the trade. The precepts of natural justice — that a hearing be held, that evidence be considered, that like cases be treated alike, that reasons be given and be reviewable — are not free-floating pieties. They are the conditions under which parties who do not know their place in society would agree to be judged. Adjudication is a case of imperfect procedural justice: there is an independent standard of correctness, and the procedure must be reasonably designed to track it. But it must also satisfy publicity. Citizens in a well-ordered society must be able to see the principles by which they are judged and see them applied to themselves. A proprietary decision procedure fails that condition whatever its error rate. So I would answer Singer's hypothetical: accuracy is necessary and not sufficient, and the insufficiency is itself something the parties would insist on from behind the veil.
Rawls has stated the practical upshot better than Hegel's weather. But permit me one addition, since I think it meets Singer on his own ground. Institutions are to be judged not only by what they do but by what they make of the people living under them. A citizenry trained to receive unanswerable determinations from an unaccountable mechanism is a citizenry schooled in passivity, and passivity is the soil in which despotism grows. That is a consequence, Singer. It is diffuse, slow, and hard to measure — but so is public health, and you do not dismiss that.
That I accept, and gladly. Formation effects and precedent effects are consequences, and I have always insisted on them. Indeed I made the same point about the workers: how a public employer handles this case shapes what is politically possible in the many similar cases coming. If the city automates while genuinely provisioning for the displaced, it makes future beneficial automation achievable. If it discards people, it poisons the well. So Mill and I are agreed. Where I still resist is Kant's claim that no exchange rate exists. Consider: if human appeal officers cost so much that the city must close a clinic to fund them, I do not think the answer is obviously "close the clinic."
Your example is a fine example of a false alternative. The commonwealth is not so poor that it cannot afford a magistrate. And observe how modest the requirement is: the city need only guarantee that any person who contests a penalty may have his case heard by a human being who can be named, who must give reasons, and who is answerable for them. This is not a clinic's worth of expense. It is a clerk. If the city tells you it cannot afford a clerk, it is not making an argument about scarcity; it is making an argument about what it wishes to prioritise, and it should say so plainly.
And that reframing matters for a further reason. The automation is being justified by savings — but no one has said where the savings go. This is the question I keep returning to and which the proposal is silent on. If the savings fund services for the least advantaged, the difference principle counts strongly in favour. If they fund a rate abatement enjoyed chiefly by the already comfortable, while the burden falls on modestly paid municipal employees and on motorists too poor to contest an erroneous ticket, then the arrangement is regressive and cannot be justified at all. "Efficiency" that consists in externalising loss onto identifiable persons is not efficiency justly obtained; it is a transfer with better public relations.
Yes — and this is where I want to insist that my framework gets you to the same place, not a worse one. Diminishing marginal utility does most of the work. A large loss concentrated on forty low-to-middle-income households can easily outweigh a small gain dispersed across thousands of wealthier ratepayers. That is not a concession I make reluctantly to appease you; it is a straightforward implication of the theory. The utilitarian case for generous transition support here is strong.
I grant that it often converges. My objection is that it converges contingently. Change the income distribution, change the numbers, and your calculus permits what mine forbids. The plurality of distinct persons is not a single system of desires whose satisfactions may be summed and traded.
Both of you speak of the forty as if the injury were to their consumption. It is not. Work is the mediation by which the particular individual attains objective standing in the universal — recognition, honour, a place among others. To eject a man from that and hand him a cheque is to send him from ethical life into the abstraction of mere need. This is how civil society produces the rabble — and note that the rabble is not poverty. A poor man may be entirely upright. The rabble is the disposition: the indignation of one excluded from the very order he is told to obey. Your severance payments, however generous, address the wallet and leave the disposition untouched.
Here we are closer than your tone suggests. I hold that the social bases of self-respect are a primary good, and that meaningful work and standing as a contributing member of the community are among them. That is precisely why I say cash severance is necessary but not obviously sufficient — why retraining, placement, and preference in other municipal employment matter in a way that money does not fully substitute for. Where I differ is that I will not say the forty have a desert claim to these particular posts. No one holds a moral title to a job antecedent to institutions. But institutions generate legitimate expectations, and those who have ordered their lives around announced public arrangements have a claim to fair notice and provision when those arrangements change.
And here is a debt, not a charity — I have said so for forty years about every improvement purchased at the expense of a particular class of labourers. But I must enter a caution against Hegel, or we shall end by sanctifying the office of parking warden. Were his principle sound, we ought never to have left off spinning by hand. The permanent interests of mankind lie with cheapness and accuracy and the liberation of human energy for higher employments. The gain here is diffused and lasting; the loss is concentrated and, if we are wise, temporary. To preserve forty offices at the cost of the many is not benevolence — it is flattery paid to the visible at the expense of the invisible.
I have not asked you to preserve the offices. Read me again: the System of Needs advances by its own inner logic, the division of labour drives toward the machine as toward its own truth, and no council may arrest this nor should it wish to. My complaint is not against the machine. It is against a municipality that reasons as a merchant reasons. Civil society is the sphere in which each pursues his particular end and the universal arises only behind his back. The State is the sphere in which the universal is willed. When the council justifies itself by cheapness, it has surrendered its own principle and become a private interest wearing a crown.
On which point I would set the matter as a test of the maxim. Let the council state the principle it acts upon. If it is: "Wherever human judgment can be replaced by a mechanism at lower cost, replace it" — then universalise it. Applied without exception it abolishes the magistrate, the physician, and finally the citizen as a being who reasons together with others about the law. Such a maxim destroys the very condition of a rightful civil union and therefore cannot be willed. But if the maxim is: "Where a task requires only observation, an instrument may serve; where it requires judgment against a person, a person must answer" — that universalises without contradiction. The council is free to adopt the second.
I can accept that formulation without accepting the machinery that generates it, which is perhaps the most any of us can say to each other. But let me push the panel on one thing where I think we may still be complacent. We have all praised consistency. Hegel alone dissents — he calls it the tyranny of the concept and defends officer discretion as the point where the universal bends to meet the particular. I want to say: be careful. Discretion is also the point at which the universal bends toward the person who is well dressed, well spoken, and knows the officer's cousin. Human officers ticket some and not others for identical conduct, and that unfairness lands on real people, disproportionately the ones with least standing.
Quite so. Human officers are subject to caprice, favouritism, fatigue, and the standing temptation to press hardest upon those least able to resist. Impartiality is not bookkeeping; it is the kernel of what we mean by justice.
You mistake the corruption of discretion for its concept. That discretion is abused does not show that equity is dispensable; it shows that the officer requires an ethos, an estate, a discipline — which is exactly what your municipality proposes to abolish rather than cultivate. And note where your own argument delivers you: you have removed discretion from the point of detection and now insist it be preserved at the point of appeal. Very well. I do not object. I merely observe that you have conceded my principle and are quarrelling about where to locate it.
That is, I think, the actual settlement, and it is worth stating plainly. Consistency in detection is a genuine justice-value — it removes arbitrariness and discriminatory discretion, and it deserves real weight. Equity in adjudication is where the particular case must be heard. Consistency in detection must not be mistaken for justice in adjudication. And I would add a caution that none of us has raised: consistent enforcement of an unevenly burdensome ordinance is not thereby made just. If the ordinance itself falls hardest on those who must park in the street because they have no driveway, perfect enforcement makes the injustice perfect too. The camera will find every violation in the poor quarter as surely as in the rich one — and there will be more of them, for reasons that have nothing to do with culpability.
That is the strongest point made this afternoon, and I had not made it. It also suggests something practical: the city should look at the enforcement pattern before and after, and be prepared to amend the ordinance rather than merely the mechanism.
Then let me state what I take to be the conditions on which I could endorse the proposal, so that we do not part in vagueness. The camera may detect. Every contested penalty must reach a named human being who hears the person, considers his reasons, and gives reasons in return which may be reviewed. The forty must receive truthful and timely notice, provision, and such assistance as the commonwealth's resources permit — for the sovereign, in whom the united will of the people is represented, has a duty to sustain those who through no fault cannot sustain themselves. And the council must state its maxim publicly.
I would add only this to the practical list: that appealing must be easy, and free, and that the burden of erroneous tickets not fall upon those who cannot spare a morning to contest them. A right that costs a day's wages to exercise is a right reserved to the comfortable.
And that the savings be traced. Not promised — traced, and reported. My whole assessment turns on it. If the money vanishes into a rate cut for the propertied while forty households absorb the loss, then on my own accounting the change was not an improvement at all, and I would join Rawls in calling it regressive. I remain the only one here who thinks the automation might be positively obligatory rather than merely permitted — but only on that condition.
Let me try to say where we have arrived, since I think it is more definite than we expected. None of us objects to the machine as such; the displacement of labour by mechanism is not in itself unjust, and the gain in consistency of detection is a gain in justice and not merely in cost. All five of us judge the algorithmic disposition of appeals impermissible as described — Kant because judgment cannot be mechanised and coercion must be attributable to an accountable will; Hegel because punishment addresses a will and not a body; Mill and Singer because to be sorted is not to be heard and because a citizenry trained in passivity is being schooled for despotism; I because adjudication is imperfect procedural justice bound by publicity and the precepts of natural justice. That the same conclusion issues from five incompatible foundations is, I think, some evidence that it is correct. All of us hold that the forty are owed provision, and that this is a debt discharged by the community which gains, not a charity. We disagree about why: Kant says they are ends and never merely costs; Hegel says work is standing and not merely wages; Mill and Singer weigh a concentrated loss against a diffused gain; I speak of legitimate expectations and the social bases of self-respect. And we disagree, unresolvedly, about whether dignity admits of any exchange rate at all — Kant says none, Singer says every good must finally be weighed against others, and the rest of us stand at various points between. But the practical instruction to the city does not depend on settling that. Automate the camera. Do not automate the judge. Make the appeal free, easy, and answered by a person who can be named. Provide honestly for the forty in work as well as in money. Publish where the savings go and watch where the tickets fall. If the council can do those five things, it may proceed with a clear conscience; if it cannot say why it will not, it should not proceed at all.