Hans Christoph Grigoleit1Hans Christoph Grigoleit is a Professor at the University of Munich, where he holds a Chair for Private Law, Commercial Law, Corporate Law and Theory of Private Law. He received his legal education at the Universities of Tuebingen and Munich (1984–1989). He then achieved an L.L.M. from the University of Miami (1990) and was admitted to the New York Bar in 1991. Grigoleit obtained his doctorate in 1996 and his habilitation in 2003 at the University of Munich. In his Ph.D. thesis, he dealt with “Pre-Contractual Information Liability” (1996), while his habilitation study concerned “Shareholder Liability” (2003). After a call to the University of Regensburg (2003–2009), he returned to the University of Munich in 2009.
* * *
A part of the series, Personalized Law.
I. The Obvious Potential of Personalization: Purpose-Bound Fine Tuning and Real-Time Communication
The potential of adjusting legal rules to personal characteristics is obvious: while the reason of law coincides with the purposes of its norms, the fulfillment of these very purposes depends, in many ways, on personal characteristics of the individuals to which legal provisions relate. The technological developments that are commonly referred to under the label of big data open up a universe of information that might be used to identify personal characteristics potentially relevant for the fine-tuning of legal purposes. A greatly significant side effect of big data in the context of law personalization is the possibility to communicate an infinite amount of information in real time to an infinite number of individuals. This paves the way to communicate in real time commands of infinite specificity to individuals who then may act in accordance with those commands.
Consequently, personalized law surely is “a worthy conversation.” However, it will be a herculean task to exploit the potential of personalization in the vastness of the oceans of law, and even more so in its microstructure. Like many fundamental ideas, the notion of personalization—as convincing as it may be from an abstract point of view—must overcome a great deal of counterintuitive findings when it is tested in concrete legal issues of all sorts.
It is therefore a visionary and pioneering achievement of Professors Omri Ben-Shahar and Ariel Porat that they have—after some groundwork—elaborated on the potential of legal personalization in a more than comprehensive way. In their book, Personalized Law: Different Rules for Different People, they have kept track of the personalization paradigm all over the legal map—from practical observations over doctrinal aspects in many areas to crucial issues of procedural and ethical quality.
The following remarks cannot exhaust the extensive line of thought put forward by Ben-Shahar and Porat. This Essay will also not venture to bridge the vast gaps in the practical fields of collecting, processing, and communicating the required data—including the related data protection issues. Nor will it tackle the intricate equality and discrimination issues arising from a tendency toward personalized law. Regarding the latter concerns, Ben-Shahar and Porat have argued quite profoundly from a general point of view that personalization may contain valid justifications for unequal treatment and that discrimination might to a certain extent be neutralized by the design of the algorithms governing personalized law.2One may note, however, that one equality caveat against high-degree personalization cannot be dispelled by individual justifications of unequal treatment: personalization tends to entirely dissolve the model function of general (and in its generality, uniform) law. If everything is special, nothing is equal.
Rather, this Essay argues that the debate on personalized law would generally benefit from more specific reasoning and background. To achieve this, I will propose or highlight some distinctions concerning the reference point of personalization (Part II). Moreover, I will take a closer look at the procedural consequences that would result from administering personalized law by algorithms—in particular, concerning the allocation of functions as between the legislature and the judiciary (Part III). I will conclude my line of thought by presenting six summarizing statements (Part IV).
II. Necessary Distinctions Concerning the Reference Point of Personalization
In order to discuss the new potential of personalization in a more specific manner, it is useful to distinguish between different reference points. I will contrast conventional and new personalization (Part II.A), ex ante and ex post personalization (Part II.B), unilateral and bilateral personalization (Part II.C), and infinite and gradual personalization (under Part II.D).
A. Conventional vs. New Personalization
In view of the obvious advantages of personalization, it is not surprising that conventional laws and legal doctrines already account for the personal characteristics of the parties and for background specificities in many (mostly quite coarse) ways.
1. Focus on substantial changes toward personalization.
Wherever conventional rules are open to person-oriented adjustments, we might make ourselves aware of this tendency and of potential steps toward refining the standard personal distinctions. We might also facilitate conventional procedures of personalization and make them more accurate via big data—for instance, if we allow criminal law judges to make use of informatics tools in order to predict the probability of recidivism when determining sanctions.
However, such findings basically do not exceed the structure of conventional law, and they are arguably insignificant in the current personalization debate. In order to identify personalized law as a reform issue, it is worthwhile to focus on developments where personalized law brings about substantial changes. The mere fact that we allow judges to make use of informatics tools in order to improve the quality of their reasoning would not account for such a substantial change. As a simple test, one may determine in what respect the intervention of the informatics tool in question reaches beyond the use of conventional calculators.
2. Justifying the need to go beyond conventional laws and doctrines.
The criterion of substantial change complicates the discussion on personalization in the sense that it requires a detailed doctrinal analysis in the discussed fields of law by which the potentials of conventional personalization should be specified. For instance, as regards the implied warranty of habitability—and with respect to other implied warranties—one should inquire whether the price and the recognizable interests of the parties might not already lower or increase the level of warranty under conventional doctrines. Moreover, in the area of criminal sanctions or calculation of damages, it might be very much in line with conventional concepts to (merely) increase an already significant level of personalization within the boundaries of conventional doctrines, whether or not the judge is provided with informatics tools for determining certain criteria of law. Another example of a conventional means to achieve personalized results is to employ the intuition and experience of professionals in person-related decisions—as it is common practice, for instance, with regard to hiring university teachers.
One may add from a general perspective that, in probably all jurisdictions, equity principles or legal methods allow for adjusting rules to individual circumstances, which place a case or a group of cases outside the scope of the rule’s purposes. Under the German principles of legal reasoning, such an operation is called (quite expressively) “teleological reduction,” and it is fully recognized. Obviously, this technique and similar operations allow for a wide range of conventional acts of legal personalization carried out by the courts. Such gateways for flexibility in fine-tuning rules and their purposes should be considered in any move toward personalization. They reduce the contrast between conventional (uniform) rules and the vision of personalization.
3. Defining the essence of personalized law.
If we are to exhaust the potential of conventional laws and doctrines, there is one crucial question: Where does the tendency toward personalized law cross the borderline of conventional (uniform) law? Arguably, three features might be characteristic for the substantial divergence from conventional law which Ben-Shahar and Porat envisage for their perspective on personalized law: First, they tend to put forward a high degree of personalistic granularity that may not be achieved by conventional legal criteria and their conventional application by the courts (discussed infra in Part II.D). Second, they evidently suggest that discretionary decisions, which involve personalized parameters and would conventionally be carried out by the human minds of judges, are handed over comprehensively to algorithmic tools. Third, they propose that personalized rules might be generated in advance and communicated to an individual in order to allow them to adapt their conduct accordingly (discussed infra in Part II.B).
Whilethese three features seem to play an important role in many of the ideas unfolded by Ben-Shahar and Porat, they evidently do not need to be fulfilled cumulatively to qualify for their vision of personalized law. However, it might improve the debate on personalized law if one would clearly identify, concerning each use of personalized law, which specific feature of the (new) personalized law is employed to compensate for which specific shortcoming of conventional law.
B. Ex Ante vs. Ex Post Personalization
I have just mentioned ex ante determination and communication as an important feature of Ben-Shahar and Porat’sideas of personalized law. This aspect and its distinction from the ex post adjudication of conventional law deserves some more attention.
1. Ex ante personalization: requirement to communicate personalized standard before legally relevant action.
Concerning some of the presented cases of personalization, it is obvious that the respective rule must be communicated to the addressee before they engage in legally relevant conduct in order to enable them to take account of the exact content of the rule in advance.
This holds true, for instance, with regard to the personalization of speed limits. If a car driver shall be obliged to follow a speed limit specified according to their personal risk potential and to any outward background conditions of his ride, they must be informed about it from the outset. Similarly, if the criteria of legal capacity are determined according to the characteristics of the individual, the result must be available to the individual in question as well as to the other party of the contract—at the latest when they enter into the contract. Moreover, regarding default clauses, both parties need to be informed about the exact substance of any default governing their contractual relationship. This is so in contract law because every legal rule is potentially price sensitive, and therefore the parties must have the opportunity to consider personalized rules in their decisions to set or accept prices. In contractual settings as well as in one-sided legal dispositions (e.g., wills), the character of default clauses presupposes an opportunity for the parties to opt out. While this does not require the addressee’s precise knowledge concerning the personalized default rule (in contrast to speed limits), they must at least have an opportunity to consider the content of the rule in advance. Mandated disclosure is also a clear example of ex ante personalization—probably one of the model areas where personalization is the most reasonable.
2. Ex post personalization: no communication required before legally relevant action.
With regard to other personalization proposals, however, the precise result of a legal personalization tool does not necessarily have to be available from the ex ante perspective of an individual acting in the context of the rule. As an example, one may refer to the calculation of damages. If one applied a personalized rule, neither the plaintiff nor the defendant would have to know in advance about the exact details of the calculation. Even according to conventional damages rules, the eventual size of a damages award is subject to many eventualities, which cannot be fully anticipated. Any preventive tendency should be oriented toward future behavior control.
Similar considerations arise when applying personalized patterns to criminal law sanctions. General principles in many jurisdictions require that the criminal conduct and the sanction must be specified by (written) law prior to the criminal conduct (i.e., nullum crimen sine lege or nulla poena sine lege). However, the resulting demands to the elements of sanctioning allow for broad leeway and potentially any reasonable personalization, in particular as the (conventional) level of personalization already seems to be quite high in many jurisdictions (discussed supra in Part II.A). The preventive tendency of criminal sanctions should be directed forward—just like the parallel issue of damages calculation.
The specification of the standard of negligence would (likely) be another example where ex post personalization suffices. While the tortfeasor should have the general capacity to avoid the conduct which gives rise to their liability, they do not need to be given access to the exact formula of the standard of care, because negligence implies the absence of a deliberate decision. Here, again, ex post establishment of the standard of care by the courts is in line with the conventional law and practice.
3. In particular: the speed limit example.
In this context, it might be worthwhile to note that the example of the speed limit, which appears in high frequency in the personalized-law model of Ben-Shahar and Porat, is not representative for specifying the general standard of care. This is so because speed limits are—as conventionally regulated all over the world—ex ante specifications of the standard of care, abstract in that they focus on speed and neglect of all other conditions of traffic. This ex ante specification serves the purpose to clearly communicate the limitation to the driver through a simple rule which relieves them (to some extent) from relying on their own judgement.
However, speed limits are always qualified. The speed limit is not a “valid right” of the driver—more restrictive commands may be derived from the general standard of care ex post. A below-average driver may, therefore, under the general standard of care, be required to drive slower than the speed limit. Apart from that, the speed limit is an example where the implementation of a personalized rule would be comparably simple—the command is directed at one individual, depends on one clear parameter (personal traffic risk), and consists of one number on which the acting naturally concentrates. This may explain the popularity of the speed limit as an example in the personalization debate, but it also questions its conclusiveness for the overall idea. To conclude the critique on the speed limit example, one may add that a high degree of granularity may not be of outstanding importance in this case; when weighed against the interests of individuals in a perfectly fitting speed limit, the demands of an appropriate degree of safety should be prioritized. This degree might as well be achieved by “defensive” uniform rules or by personalized distinctions based upon coarse criteria (discussed infra in Part II.D).
4. Consequences resulting from the requirement of ex ante communication.
The conditions of ex ante and ex post personalization differ fundamentally in procedural respects: ex ante personalization and ex ante communication involve a new process of defining the law and of communicating it to the individual concerned. The demands of advance specification and notification require that the analysis of the individual case is—to a certain (and potentially large) extent—shifted from the courts to a new, and probably not yet fully appreciated, process. If any legislature shall keep full control of that process, it must be equipped with a significant capacity to anticipate relevant circumstances of individual cases (discussed infra in Part III).
In contrast, if the specification of a personalized rule can be left to a subsequent legal dispute, the distribution of function as between the legislature and the judiciary can be more or less retained in the conventional way: the personalization formula and its potential algorithmic implementation might be under the control of a legislative function, and the courts may then apply the personalized rule (with some discretion) to the case.
C. Unilateral vs. Bilateral Personalization (and Price Sensitivity)
Personalization is comparatively simple if a certain legal issue concerns only (or predominantly) the interests of one party. A unilateral perspective can be taken, for instance, if the personalization of the preferences of the testator is at stake. Similarly, as regards mandated disclosures in the law of consumer protection, the relevant information within a given quantity of information may be determined according only to the individual consumer’s informational needs.
If a legal issue, however, touches bilateral interests, personalization becomes more complex, as any legal change according to the characteristics of one party must also be appropriate with regard to the (personalized) interests of the other party. This holds true in particular for any personalization of implied, default, or mandatory contract law rules. In this field, an intervention on the grounds of personalization must account for the fact that any rule of contract law is potentially price sensitive. If, for instance, one considers adapting the mandatory or default warranty period to the risk inclination of a buyer, it is obvious that there must be a mechanism for price adjustment. It will not be trivial, in practical terms, to open up personalization for such a mechanism—particularly if we envisage a granularity of some high degree.
D. Infinite vs. Gradual Personalization
Paying more attention to the aforementioned criteria—to conventional personalization, to the difference between ex ante and ex post personalization, and to the challenges of bilateral personalization—might also call into question any tendency toward infinity in granularity. In contrast, the potential of graduality comes into play.
1. Limits to infinity.
The proposition that infinite personalization of the law might be achievable or even meaningful is a chimera—not only from a practical point of view but also from a theoretical perspective. Not even algorithms have the capacity to process infinite amounts of information on personal and background characteristics.
Therefore, it is worthwhile to note that any tendency toward (a more) personalized law merely refers to improving the degree of granularity. Moreover, as the observations on conventional doctrines (supra in Part II.A) have indicated, the dichotomy of (conventional) uniform law on the one side and (potentially infinite) personalization on the other side might be oversimplified.
2. Correlation between degree of personalization and complexity.
The tendency toward legal personalization potentially increases the complexity of the law and thereby increases transaction costs concerning decision-making and legal enforcement. As the potential of personalization is infinite, so is the potential of increasing transaction costs. Any development toward personalization must consider this increase in transaction costs. Any step toward personalization should depend upon whether the marginal utility of the increased degree in legal granularity outweighs the marginal cost of the increase in legal complexity and the corresponding increase in legal transaction costs.
It seems likely that this test and the complexity issue will set considerable limits on high-degree personalization. It is also evident that the cost-benefit analysis will be intricate and require detailed and area-specific inquiries. While it is true that personalization and the corresponding data resources may also help to deal with complexity,3This is the bottom line of Ben-Shahar and Porat, with some quite abstract examples. One may mention that the traffic-death rate per inhabitant in Europe is only 25% of that in the United States while traffic rules are quite uniform in Europe as well. this general observation does not pass as sufficient proof for area-specific efficiency of high-degree personalization. For instance, abandoning the numerus clausus principle in property law in favor of infinite distinctions might have a tendency to overstrain an average human mind and the average human capacity of trust and reliance.
It does not need to be mentioned that supporting the trust and reliance of individual actors with legal certainty is a fundamental feature of legal efficiency. However, it is quite complex to measure this feature concerning specific issues of law. In general, a tendency toward infinite granularity of the law does not seem to be a perfect match to the human mind and to human expectations. It is uncertain whether (and how much) the increase of information resources will provide for sufficient compensation.
3. Cutting down on the degree of personalization and on the relevance of algorithmic power?
One potential conclusion, which one might draw from the complexity issue, would be to abstain from high-degree personalization and focus on a few significant and clear features or on the design of a menu of stepwise settings to be chosen according to the crudely personalized features of the case. Such a coarsening approach would also facilitate “matching people in a personalized manner,” which Ben-Shahar and Porat recommend to account for the complexity issue with regard to group activity and its coordination. The matching proposal, however, amounts to standardizing rules in a personalized context and confirms the relativity of the overall exercise.
Restricting personalized law to crude features and giving up on the (unrealistic) paradigm of infinite granularity may also, to some degree, cut down on the dominance of algorithms and artificial intelligence. In the personalization debate, this dominance seems to be ominous and enticing at the same time. While it is realistic that we will see improvements from the use of algorithms with regard to the reliability of information processing and of avoiding biases, this does not mean that personalization will generally improve with the degree of granularity and—correspondingly—with the measure of algorithmic power. Rather, just like the degree of personalization, the general necessity and the specific extent of the use of algorithmic power should be justified by balancing its marginal utility and marginal cost. In this balancing process, apart from the complexity cost accruing to the individual parties, one should consider the procedural issues attached to designing the algorithmic tools of personalization, which I will now turn to.
III. Procedural Background: Creating, Controlling, and Communicating the Parameters and Algorithms
The parameters of any development toward personalized law do not identify themselves. Nor do algorithms design themselves to account for characteristics of individuals and of the circumstances in which they act. Moreover, one should bear in mind that—at least concerning some of the areas of a potential personalization reform (see supra Part II.B)—real-time generating and communicating of personalized rules must be warranted before the addressee engages in legally relevant conduct. In the light of these findings, the tendency toward personalized law will bring about major changes to the structure of power distribution in the judicial system. One may argue that these effects can be accepted in the light of the benefits of personalized law, but they must at least be taken into account.
A. The Legislature: Raising Complexity, Reducing Transparency and Increasing Expert Influence
If the structure of democratic powers shall be retained, the competency for defining the relevant parameters and for designing the algorithms would—at least in general—have to rest with the legislature. At this level, one important issue will be that the personalized rules will be more complex and less transparent than conventional laws. The relevance of this issue depends upon the degree of personalization that is pursued. The higher the degree of personalization, the more complex and less transparent the rules and their making will become. By the same token of transparency, the degree of personalization may also undermine public control of lawmaking and the related discourse in society. One may expect adverse effects to the democratic process.
Furthermore, according to the degree and complexity of personalization and to the corresponding demands of setting up algorithms, expert participation and influence will grow in the process of legislation. This will shift power from parliament either to the administration or to private actors.
B. The Judiciary: Loss of Decision-Making Power
The more personalized (and specified) rules are designed on the legislative level, the less leeway for judicial determination. Therefore, on the level of the judiciary, the tendency of personalized law goes along with a loss of decision-making power corresponding to more specific legislative commands. One may call this tendency an appropriation of quasi-adjudicative powers by the legislature.
Such a shift of powers is not detrimental per se. We should take the shift into account, however, and the consequences should be thoroughly explored. One should note in particular that the full view on actual cases that have been decided by a court provides for valuable insights in retrospect that the legislature may not easily anticipate. The case for fixing a personalized rule may seem less certain in the context of the comprehensive set of facts presented before a court.
Another issue of a tendency toward high-degree personalization is that—inasmuch as a legal dispute will be decided on the grounds of a nontransparent algorithm—the decision-making process would be dehumanized as compared to a court decision based upon a conventional rule. This procedural shift might affect the acceptability of decisions even if the outcome is crystal clear and—from the perspective of ideal legal reasoning—less flawed or “better” than in cases of purely human decision-making.
C. Procedural Dimension of the Distinction Between Ex Ante and Ex Post Personalization
As I have mentioned above in Part II.B, another characteristic feature of personalized law would be the procedural distinction between ex ante and ex post personalization, which relates to the demands of ex ante generating and communicating personalized rules. The demands of ex ante personalization deepen the power shift between the legislature and the judiciary. Moreover, (only) ex ante personalization requires a communicative device, which makes the result of the personalization process available to the addressee before they engage in legally relevant conduct.
- It might be useful to diverge from an all too general (uniform?) approach to personalized law and to account—on an area-specific basis—for some fundamental distinctions concerning the legal, practical, and procedural background.
- The debate on personalization should thoroughly explore the potential of conventional laws and doctrines which already allow for a great deal of personalized considerations. Any move beyond the borderlines of conventional laws and doctrines should specify what sort of shortcoming of conventional laws and doctrines is supposed to be remedied by which feature of the reform: Is the move toward personalization directed at (1) high-degree granularity, (2) comprehensive “algorithmizing” of balancing decisions, or (3) ex ante implementation of legal rules?
- The debate on personalized law should clearly distinguish between areas where ex ante personalization/communication is and is not essential. Only ex ante personalization requires a new process of defining the law and communicating it to the concerned individuals.
- Legal personalization is comparatively simple if a certain legal issue relates only (or predominantly) to the interests of one party. Personalization becomes more complex if the process of personalization must take account of the (personalized) interests of the other party. In the latter case, a process of bipartisan fine-tuning or of price adjustment must be provided.
- Infinite personalization of the law is a chimera. Any tendency toward legal personalization only refers to improving its degree. Increasing the degree of personalization tends to increase the complexity of the law and thereby transaction costs concerning decision-making and legal enforcement. Any development toward personalization must consider this increase of transaction costs. Any step toward personalization should be made dependent upon whether the marginal utility of the increased degree in legal granularity outweighs the marginal cost of the increase in legal complexity and the corresponding increase of legal transaction costs. A potentially meaningful reaction might be to focus personalization generally on few significant and clear features. Personalization will not necessarily improve with the degree of granularity and—correspondingly—neither with the use or measure of algorithmic power. Just like the degree of personalization, the general necessity and the specific extent of the use of algorithmic power should be justified by balancing its marginal utility and marginal cost.
- Corresponding to the degree of personalization and algorithmization, lawmaking will be more complex and less transparent. Expert participation and influence will grow in the process of legislation, shifting power to administrative functions or to private actors. Increasing the degree of personalization also reduces the leeway of judicial determination. The legislature may not easily compensate (in its algorithm-drafting exercise) for the resulting loss of valuable case-related insights. The specific demands of ex ante personalization further exacerbate the power shift between the legislature and the judiciary. One may argue that these effects can be accepted in the light of the benefits of personalized law, but they must at least be taken into account.
* * *
Hans Christoph Grigoleit is a Professor at the University of Munich, where he holds a Chair for Private Law, Commercial Law, Corporate Law and Theory of Private Law. He received his legal education at the Universities of Tuebingen and Munich (1984–1989). He then achieved an L.L.M. from the University of Miami (1990) and was admitted to the New York Bar in 1991. Grigoleit obtained his doctorate in 1996 and his habilitation in 2003 at the University of Munich. In his Ph.D.-thesis, he dealt with “Pre-Contractual Information Liability” (1996) while his habilitation study concerned “Shareholder Liability” (2003). After a call to the University of Regensburg (2003–2009), he returned to the University of Munich in 2009.
* * *