Faculty of law blogs / UNIVERSITY OF OXFORD



This post is part of a special series including contributions to the OBLB Annual Conference 2022 on ‘Personalized Law—Law by Algorithm’, held in Oxford on 16 June 2022. This post comes from Timothy Endicott, who participated on the panel on ‘Personalized Law’.

Law regulates the life of a community impersonally, through rules that apply generally to classes of persons. As Ariel Porat and Lior Jacob Strahilevitz say, ‘Law is impersonal’. They view it as a regrettable contingent fact about law—a matter of how it happens to have developed (before the emergence of algorithmic big data analysis), which could and should be changed.

I think it is a truth about the nature of law. ‘Personalized law’ may sound like a potentially better kind of law, but I think that personalizing law actually means getting rid of it. Some laws ought to be got rid of. But it cannot be a good general strategy for governance, even in the 21st century.

In their exciting book, ‘Personalized Law: Different Rules for Different People’ (OUP 2021), Omri Ben-Shahar and Ariel Porat call for ‘a jurisprudential transformation of to each their own law’. If each of us had our own law (in the sense of particularized norms imposed on each of us personally) we would not have any law (in the sense of general norms of a political community). Except, perhaps, one general norm of the community, giving force to the particular commands of the machine.

In ‘The Death of Law? Computationally Personalised Norms and the Rule of Law’, Karen Yeung and I argue that the worthwhile potential for algorithmic personalization of governance is strictly limited, and can only legitimately be pursued within the framework of the rule of law. Here I will offer some reasons for concluding that moving toward ‘to each their own law’ is not the right direction of travel.

But I should emphasize that I am not making a general argument against Ben-Shahar and Porat’s book. They actually advocate two different things under the rubric of personalized law. The dramatic prospect, which I will call ‘radical personalisation’, is captured in their slogan ‘to each their own law’, and it means getting rid of general legal rules. The less dramatic prospect, which they also endorse, is to craft many more general legal rules: general rules that are well tailored to relevant considerations. That is often obvious good sense, and sometimes non-obvious good sense, and in such cases it is undoubtedly valuable. So there is much good sense in Personalizing Law. Let me illustrate with an example from the vignette of Abigail and David, with which Ben-Shahar and Porat introduce the idea of personalized law:

David finds a parking ticket on the windshield of his Acura. The fine is $97. An acquaintance mentioned to him a few days earlier that he received a parking ticket of $39 in the same neighborhood. Hmm, David didn’t realize that his friend’s income was so much lower.

I am all in favour of tailoring parking fines to the wealth of the driver, if it can be done effectively. Now, suppose that it is done, as suggested in this vignette, by reference to income: suppose, for example, that David is in an upper income band, and his friend is in the basic band. There is a basic fine for a particular parking offence, with a multiplier of 2.5 (as in the vignette) for drivers in the upper band of income. This scheme uses a general legal rule. I think that it might be a better rule (I say ‘might’ because this would depend on some crucial questions of craft in the making and administration of the scheme, and some potentially difficult questions of data-sharing between the income tax office and the parking enforcement agency, and so on). But it does not amount to giving ‘to each their own law’. David’s fine is determined by a general rule.

To take a paradigm, real-life example of such better-differentiated general legal rules, consider the difference between a poll tax (with each liable person required to pay the same sum) and an income tax like the UK income tax (differentiated, first of all, by income, and then further differentiated by progressive rates, by personal allowances (eg for marriage and civil partnership), and by relief for such outgoings as pension contributions, charitable donations, family maintenance payments etc). In terms of effectiveness in securing public revenue and in terms of social justice and the need for public confidence in the system, the UK income tax is so much better than a poll tax, that it is well worth incurring the huge resulting cost of the collection system, and the costs resulting from the greater scope for fraud. Insofar as Ben-Shahar and Porat advocate targeted schemes like the UK income tax rather than crudely undifferentiated rules such as a poll tax, they have an irrefutable case (and they have much to say on the complex and crucial details of what forms of targeting, in which schemes, might secure just and effective law at reasonable cost). But the result is not ‘to each their own law’. The result is better general rules.

Now contrast a radically personalised tax scheme that really does give ‘to each their own law’. It would be a very different matter from the tailoring of tax law. It would get rid of tax law. It could be carried out by conferring discretion on human tax collectors to assess each particular person’s ability to contribute to public expenditure, without applying a general rule. As with so many schemes of taxation through human history, that massive discretion would be a disastrous abandonment of the rule of law. Not only because of the scope for corruption on the part of the tax collectors, but even if the tax collectors were honest: subjecting you or me to the unregulated say-so of the tax collector as to what you can afford, or what I can afford, would subject us to arbitrary government.

Computer personalization of state revenue would solve the problem of corruption (at least, if we can assume integrity on the part of the revenue administration and the programmers). Here, it seems to me, lies the great attraction of computer personalization to those theorists who advocate it: it seems to offer the advantages of massive discretion (what Ben-Shahar and Porat call the ‘precision benefit’), without the drawback of the arbitrariness of governance by the whim of a human official. The tax demand might be generated by a big-data analysis of the affordability to the taxpayer of contributing to state revenue. It might be done through analysis of the individual’s preferences and dispositions in expenditure and other transactions, using commercial and financial behaviour as a basis for an algorithmic assessment of ability to pay. Such a radically personalized tax seems to promise a revenue system that would be much more sensitive than an income tax (which does not reflect real wealth), or even a wealth tax (which would not reflect outgoings). The state would extract revenue on the basis of what really matters to the person: affordability.

But the opaque technique of an algorithm designed to assess affordability would not solve the problem of arbitrariness in the determination of my liability to tax, or yours. I call it ‘arbitrariness’ because the particular tax assessment would land in my personalized tax account as an imposition by the state with no rationale—except the unregulated say-so of the machine. If we replaced the complex general rules of income tax law with a particularised assessment by big data analysis—if we got rid of tax law—we would give up the intelligibility of the state’s imposition of tax liability, and therefore we would give up responsible government. Tailoring tax law to the relevant considerations is a good idea. Getting rid of tax law would be a disaster.

So it is worth asking Ben-Shahar and Porat’s basic question:

‘The basic question this book wrestles with is the right scope of personal­ized law.’

That scope is patently limited. The best default principle is that a political community like the UK ought to be governed by law, and getting rid of law through computerised personalization is only justified when doing so would help to achieve the goals of the law. And this is just where Ben-Shahar and Porat make a very general, very explicit claim:

‘Plain and simple: compared to uniform law, personalized law promotes more effectively the law’s underlying goals. Any goals, of any law.’

That general claim—that personalization promotes the law’s goals—is supported by an analogy with medical care: ‘Personalized medicine tailors treatments to each patient based on their anatomy, physiology, and environment of use.’ Ben-Shahar and Porat conclude that lawmakers could achieve the same massive ‘precision benefit’ that good medical care achieves by personalization. Omri Ben-Shahar asks, in his contribution to this blog, ‘If medicine, education, or parenting can treat, teach, or nurture better when personalized and adjusted to the subjective, why not law?’.

The answer is that law does not treat, teach, or nurture. Personalized medicine bears a crucial difference from personalized law: in medical care, precisely because the purpose is to treat the patient, there is no inherent value in uniformity of action. It may be efficient to locate a variety of medical services on a single site, or to prescribe a uniform treatment (such as a standard dose of a medicine) in order to achieve delivery of second-best treatment by medical personnel (or by patients administering their own treatment) who are not able to make fine-grained, expert judgments. But these values of uniformity in medical care always reflect costs of service delivery, so that personalized medicine is generally desirable, if the problems of cost can be met.

Legal regulation, by contrast, often has the achievement of uniformity as its goal. Ben-Shahar and Porat recognise this, citing the rule as to which side of the road to drive as the ‘canonical’ example: ‘it would be a disaster to replace [a rule that vehicles travel on the right] with a rule that allows some people to drive on the left.’ The purpose of law is not to diagnose and to cure a person of illness. So personalization in law is not generally valuable; it is only valuable when it is a way in which the state can pursue its overall purpose of ordering a community (and ordering the community’s relation with each individual person) for the public good.

Very often, for a variety of reasons, that purpose rules out personalization. Consider, for example, minimum voting age requirements (discussed in ‘The Death of Law?’, section IV). The voting age ought to correspond to the maturity that is needed for the responsible exercise of the right to vote. But that can only be a very rough correspondence, because maturity is a very personal matter, and adolescents grow up in different ways, at different rates. We could personalize the right to vote, with an algorithmic inference through big data analysis as to whether each individual has reached an appropriate level of cognitive and social maturity. But personalizing the franchise in such a way would involve a new kind of arbitrary government: the franchise would be based on considerations on which the community ought not to act. The use of a single, uniform voting age is a better approach than personalizing the franchise, because it is needed to attain the goals of the law.

And in fact, Ben-Shahar and Porat recognise that the purposes of law cannot be promoted by radical personalization. They offer a thought experiment in which each citizen has a ‘personalized constitution’, with a personalized level of protection for different fundamental rights such as freedom of speech or criminal process rights, tailored to each citizen’s preferences: ‘Yes, even voting rights could be subject to personalization, since some people value these rights less and refrain from voting’. They do not propose that we should take that step. On the contrary:

‘… in areas where the primary value of the rights is public and where equal participation by people is critical for a democracy, personalization in the distribution of rights might erode their foundations.’

That crucial point has to be understood, in order to understand Ben-Shahar’s and Porat’s approach. I think it is sound, and it supports a conclusive refutation of radical personalization as a general strategy. Moving toward ‘to each their own law’ is not the right direction of travel, because arriving at that destination would not achieve the goals of the law.


Timothy Endicott is Vinerian Professor of English Law, University of Oxford.


With the support of