Homo Arbiter

The new always happens against the overwhelming odds of statistical laws and their probability, which for all practical, everyday purposes amounts to certainty; the new therefore always appears in the guise of a miracle.

— Hannah Arendt, The Human Condition (1958)¹

The Sentence That Returned

Jason Pepper finished serving his twenty-four-month prison term three days after the court of appeals reversed his sentence. It was June 2005. He began supervised release while the case went back to the district court, where the punishment for his part in a methamphetamine conspiracy would have to be decided again. By the time of that hearing, almost another year had passed. The conviction remained. There was also a life being lived outside the proceeding, and some of it had become evidence.²

Pepper had completed a five-hundred-hour drug-treatment program in prison. At the resentencing he testified that he no longer used drugs, had enrolled full-time at a community college and earned A's in the previous semester, and was working part-time. His father, who had had virtually no contact with him during the five years before his arrest, described their renewed relationship. A probation officer recommended the original term in light of Pepper's assistance to the government, rehabilitation and low risk of recidivism. The judge adopted the testimony as findings of fact and again imposed twenty-four months.

These were familiar signs of rehabilitation, which is part of their interest. There was no revelation that the original crime had been imaginary. Pepper had begun doing things a sentence is ordinarily supposed to leave possible afterward: working, studying, keeping in touch with someone who had lost touch with him. His father could testify to a relationship which the conviction could neither describe nor cancel. The court had to decide what weight to give that testimony.

The government appealed again. The Eighth Circuit rejected the use of rehabilitation after the original sentencing as a basis for the further reduction and directed reassignment to another judge. The Supreme Court sent the case back for reconsideration in light of Gall, an intervening sentencing decision; the Eighth Circuit returned to its position. By the second resentencing hearing, in October 2008, Pepper was a night-crew supervisor at a warehouse retailer, had been recognized by management and was anticipating a promotion. He had married and was supporting his wife and her daughter. The new judge refused a further variance based on his rehabilitation and, after other adjustments, imposed sixty-five months in January 2009. The reduced allowance for his assistance also mattered: exclusion of rehabilitation was not the only difference between the sentences. Reopening a judgment had reopened favorable determinations as well as unfavorable ones.

The appellate court acknowledged the progress. It commended Pepper for it. Then it held that the progress was not a permissible ground for the variance he sought. One can imagine a database failing to receive new facts, but that was not what had happened here. The facts had been received, discussed and praised. The difficulty was what the court was permitted to do after knowing them.

In 2011 the Supreme Court rejected the categorical exclusion. When a sentence has been set aside on appeal, a court at resentencing may consider rehabilitation since the first sentence, and in an appropriate case that evidence may support a sentence below the advisory Guidelines range. Federal sentencing law called for consideration of the defendant's history and characteristics, as well as such purposes as deterrence and protection of the public. Later conduct could bear on all of these. Pepper was to be resentenced with appropriate weight given to the evidence. The holding did not require a reduction, confer forgiveness or authorize every prisoner to reopen a final sentence.³

That last limit leaves an awkwardness which a celebration of individual judgment would too readily lose. The prisoner whose sentence was imposed correctly might work just as hard at rehabilitation and never have another occasion to present it to a sentencing judge. The Sentencing Commission and the court of appeals had regarded this unequal opportunity as a reason against considering the evidence. Access to a second hearing can depend on something other than the merit of what a person has done since the first. Two lives can change in similar ways while only one returns to the place empowered to recognize the change.

The Supreme Court's answer was that differences in procedural opportunity arising through ordinary appellate review were not the unwarranted sentencing disparities Congress had instructed courts to avoid. This allowed a court already resentencing someone to use pertinent evidence. It did not equalize the occasions on which such evidence would be heard. The law could correct one exclusion without completing its account of the people who never reached the hearing. I do not think that incompleteness is a reason to prefer the exclusion. It is a reason to resist making the successful hearing stand for everyone left outside it.⁴

There was another difficulty with the exclusion, one less easily defended in the name of equality. The Eighth Circuit allowed later conduct supporting increased punishment to be considered. The Supreme Court pointed to that practice, including the possibility of offenses committed after an original sentence. These were not offenses attributed to Pepper. They exposed what was wrong with the rule applied to him: the passage of time could bring the state another reason to punish, while being denied the capacity to bring a reason for restraint. If later evidence was inadmissible because it had not existed at the first sentencing, the exclusion should have reached both. The statutes supported no such general exclusion, and certainly no permission to attend only to the adverse part of an intervening life.⁴

Justice Thomas, dissenting, wrote that he did not see what purpose further incarceration would serve. He nevertheless would have upheld the sentence. He rejected the remedy through which Booker had made the Guidelines advisory and believed the mandatory scheme should govern here, since Pepper had admitted facts permitting a sentence longer than the one imposed. The disagreement concerned the legal authority to depart, not an inability to recognize rehabilitation. A judge could see the force of the later life and believe himself bound to leave the punishment in place.⁵

This is already more than a problem of putting a human in the loop. There were people throughout Pepper's proceeding, exercising judgment, writing reasons and disagreeing in public. A constitutional account which merely required a name at the end of the decision would have found much to approve. It would still have had to ask which institution could change the rule of consideration, whether the rule was defensible, and what the person subject to it could require that institution to hear. Compassion in an official is no substitute for lawful authority. Lawful authority is no excuse for declining to examine the arrangement that makes compassion irrelevant.

A record is complete enough only in relation to a question. The institution must still establish its authority to decide it. The record can establish a conviction without settling the appropriate sentence at a later hearing; establish a sentence without settling every later use of the conviction; preserve a justified act without authorizing its repetition. We have followed written obligations across places and financial claims across changes of ownership and circumstance. In each passage, something could remain accurately recorded while the actions available under it changed. The institution which receives the record must account for that change. Its possession of a reliable history does not make it the inheritor of every power ever exercised through the history.

The Slippery Right

Kant called pardon the most slippery of sovereign rights. The phrase is exact in a way that praise of mercy often is not. A ruler who can spare can favor; severity remains available for those who have failed to acquire the ruler's interest. An exception may relieve a person and strengthen the dependence of everyone else upon the person granting it. Kant's warning belongs to his treatment of public right. His ethical discussion also recognizes a duty to be forgiving without requiring meek submission to repeated wrong. Neither passage supplies a doctrine of databases. Together they make it difficult to treat every limit on punishment as one undifferentiated act of kindness.⁶

Jeffrie Murphy presses the objection further. If the reasons for leniency are reasons of justice, why is the applicant being made to ask for mercy? If they are not, why should this applicant receive what another does not? An eloquent account of transformation does not answer the second question. It may reveal one of the ways unequal treatment acquires a defensible appearance. The person with a stable job, a supportive family and someone able to arrange the papers possesses resources for presenting a changed life which another person may lack. This need not make the testimony false. It should make us cautious about treating the ability to produce it as the measure of entitlement.⁷

Pepper had substantial evidence to offer. Others may have less, for reasons which range from their circumstances to their conduct. Some will have made little effort. Some will remain unpleasant. A public rule cannot require the administrator to admire everyone it protects. If the period governing a particular use of a record has ended, the person who invokes that limit is asking the institution to observe a boundary on its own power. Requiring proof of regeneration before honoring the boundary would replace a right with an audition.

The distinction becomes harder where an institution has duties toward people who are not making the application. A completed sentence does not establish that a person can safely be entrusted with every responsibility, and a criminal judgment is not the only source from which a protective restriction can derive authority. An employer or licensing body may have a legitimate reason to examine conduct closely related to the proposed work. A victim may retain a claim to compensation or protection. None of this follows merely from the word desert. Nor does it disappear merely because punishment has ended.

Desert addresses a reason for punishment. It does not distribute an unlimited power to punish among all later recipients of the record. Douglas Husak's distinction between deserving punishment and justifying its imposition all things considered matters here. Even when the first question is answered affirmatively, the second remains. The further extension is ours: a judgment imposing a finite punishment does not, by itself, authorize a sequence of unrelated institutions to impose additional disabilities. Some later burdens will be protective, some punitive, some defensible on another ground, and some will be an old condemnation surviving in a procedure nobody has recently defended. We have to ask which is which.⁸

Calling everything punishment would evade this work as surely as calling everything risk management. Consider a restriction on work involving a particular vulnerability. If the institution invokes protection, it must explain the connection between the conduct and the responsibility, consider pertinent current evidence, and defend the scope of the restriction. It should also have to consider whether supervision, a narrower assignment or another less restrictive arrangement would adequately protect those exposed. These alternatives will not always suffice. Where they do not, the interests of the people who would bear the danger belong in the decision. But an institution cannot use their existence to dispense with the inquiry and then invoke the original offense as its answer to every question about duration.

There is a political convenience in this dispersion. Each recipient can describe its own decision as small and prudent. None need intend the combined exclusion from ordinary participation. The person encounters the decisions together, and their accumulation can impose a condition no single office believes it has imposed. A new court to pronounce once more upon the person's character would not necessarily help. What needs examination is the set of uses being made of the conviction, including the authority each institution claims for its part in the result. A true record has become the occasion for governing, and governing has to answer for more than the truth of its occasion.

The Mercy Threshold is the point at which a true record must cease to determine a person's future automatically. It is implemented through designed forgetting and through judgment by an institution that can be named, challenged, and made to answer. The name must not obscure the allocation. Ordinary release belongs in declared, purpose-specific rules. The person meeting their conditions need not obtain an exceptional grant. Verification can continue, and the record can remain available for a justified archival, adjudicative or protective purpose; what ends is its presumptive authority for the later adverse use to which the threshold applies.⁹

After that boundary, the institution seeking renewed adverse use bears the burden. It must establish a present connection to the proposed decision, explain why current evidence or a less restrictive means is inadequate, and limit the authorization in purpose and time. Independent review matters because the holder of a record cannot be the final authority on whether the limit placed upon its use still binds. The affected person needs notice before the act where practicable and an effective route to challenge. The new act and its grounds must be preserved in a receipt. An exceptional authorization neither restarts the threshold nor restores general use in other domains.⁹

These limits are compatible with retaining evidence against the institution itself. The order that imposed a burden, the information relied upon and the means of challenging the act may remain necessary long after ordinary use of the person's history has ended. Retention should follow effects, obligations, review and systemic significance, with personal details restricted where they need not travel. To destroy the only evidence by which an injured person can make an institution answer would be an extraordinary interpretation of protecting that person from memory. A sealed record and a publicly circulating classification are different arrangements. So are remembering an official act and making its subject answer for it again.

The limit must also follow the use through its transformations. A holder cannot retire an entry while continuing to act upon it through a score, an affiliate's classification or a trained system's output. Material reliance on expired information remains within the rule. Establishing that reliance in a trained system can be difficult; the institution may be unable to isolate the old record's contribution. It must disclose that uncertainty and support independent testing. The difficulty does not restore default permission to use what has lost authority.¹⁰

There are good reasons to bring inquiries to an end. A person who has prevailed in a dispute should not have to meet the same wealthy adversary indefinitely under successively revised descriptions of the case. Others plan around settlements. Evidence deteriorates, and the work of reopening can itself become a sanction. A right to question power cannot mean that every obligation is perpetually provisional. Pepper concerned the evidence admissible in an already reopened sentencing; it did not make change in a person's life sufficient to set every judgment aside. The constitutional distinction is between the finality of a specified determination and a claim that the determination has settled every future relation of its subject.

Ordinary rules will still leave cases in which an otherwise available consequence should yield. Public clemency belongs here, with an authority, a scope and reasons for departing. This is power, including when it is exercised generously. Patterns of relief should be open to scrutiny for favoritism; conflicts should be disclosed; refusal must remain appealable or confined to its own domain rather than acquiring the force of a universal veto. Where exceptions reveal a recurrent fault in the rule, the fault belongs before those empowered to change the rule. A government should not congratulate itself for repeatedly granting relief from a condition it can amend.

The decision-maker must not become an insurer of the recipient's future. A grant can be corrupt, reckless or unsupported, and its author should answer for those defects. A defensible grant can also be followed by harm. If every later failure is charged back as proof that relief should have been refused, the safest course will be to grant it only to people whose futures were already easiest to trust. That would preserve the appearance of mercy by withdrawing much of what makes mercy a risk. The people exposed by release deserve protection and a serious account of the decision. They do not gain it from a rule which confuses a bad outcome with proof of an indefensible judgment.

Nor does the public authority inherit the victim's forgiveness. Arendt's pairing of forgiving and promising concerns relations among people: a release from the continuing hold of what was done, and a limited continuity made through an uncertain future. A schedule governing the use of a record does another kind of work. It can end an institutional consequence while the wrong remains unforgiven. The person who suffered it may have ample reason to withhold reconciliation. A public order should be able to release one person without requiring the other to announce that the past has ceased to matter.¹¹

The wronged person may regard public release as a failure to take the injury seriously. Their refusal to forgive can preserve a judgment they are unwilling to surrender for the convenience of everyone hoping to move on. It is possible to recognize this and still deny that the refusal authorizes every later public burden. Conversely, someone may forgive while needing protection against renewed contact or further harm. An institution must attend to the remaining danger and the claims it is charged to protect. It cannot infer safety from forgiveness any more than it can infer a sentence from resentment.

There need be no reconciliation at the end of this arrangement. One person can regain a freedom while another declines a relationship. The law may preserve evidence, enforce a remaining obligation and prevent coercion without requiring the encounter to become harmonious. What the book calls the Gift Zone protects this space from compulsory accounting; it does not protect violence or exploitation from law. A wronged person may give more than an institution can demand. The institution must not make its own observance of a limit conditional on obtaining that gift.

Before the Harm

Pepper's rehabilitation mattered partly because it bore on the likelihood of another offense. Prediction was already inside the judgment, serving a legitimate purpose. To say that institutions must wait for completed wrongdoing before acting would exclude much of what we reasonably ask them to do. A credible warning about a danger can arrive before an injury. Sometimes waiting for certainty means deciding that someone else will bear the experiment.

An inquiry can begin with evidence insufficient for a final adverse determination. Between the two lies the possibility of acting temporarily while finding out. A temporary restriction may be defensible while a serious danger is examined, with a short duration, a limited reach and prompt access to review. Continuing it requires more than repeating the initial warning. The institution must show what has been learned, which uncertainty remains material and why the imposed burden is still warranted. Otherwise a precaution can become indefinite precisely because the institution has failed to complete the work for which it was granted time.

The demands vary with the decision. Credit is extended before repayment is known; an institution need not await a default before assessing the risk of lending. A prediction may supply important, even decisive, evidence under a lawful and defensible rule. It still cannot determine whether the institution was entitled to acquire the inputs, use them for this purpose or impose this consequence. Restricting someone's liberty raises different requirements from declining a proposed loan. The differences ought to be argued in terms of the interests at stake, the harm threatened, the reliability and applicability of the evidence, and the alternatives available. Placing all these acts under one threshold would be a way of avoiding political judgment while insisting upon its importance.¹²

Where a risk justifies intervention, the risk and the intervention are the present matters to be established. A forecast must not be redescribed as an offense already committed. The more severe or durable the burden, the stronger the required justification and the more consequential the opportunity to challenge it. A recipient should be able to dispute mistaken inputs, the application of a population estimate to this case, the permissibility of the rule, and the breadth of the response. None of these objections needs to be framed as a promise that harm is impossible. Requiring such a promise would make precaution immune from challenge by definition.

The Right to Disappoint the Model belongs within these distinctions. It gives no right to escape a justified safety restriction by being difficult to predict. It gives no institution authority to treat departure from a profile as, in itself, a wrong to be corrected. A person may behave just as expected and remain free. He may supply compelling evidence that an expectation was mistaken, and still owe an obligation undertaken before the evidence arrived. What matters is that his standing to address the decision is not a permission issued by the prediction. The account can be useful without possessing jurisdiction over every reason for departing from it.

Arendt's idea of appearance helps describe what would otherwise disappear. A person addresses others and undertakes something among them; he does not merely add another attribute to a description. A public institution must often classify him, but the classification is made for a purpose and can be called upon to explain its relevance. To demand a total biography before accepting a particular challenge would reverse the asymmetry the book has sought to establish. The exercise of power should carry enough of its history to be contested. Its subject should not have to make every private association available as the price of entering the forum.

An institution can be entitled to refuse the challenge after examining it. It can also be required to cease a restriction while uncertainty remains. This second possibility is the one a preventive system has most reason to resist. An agency charged with avoiding harm can always describe an imaginable event which further control would make less likely. The constitutional question is how much authority that possibility supports, for how long, and at whose expense. If no remaining uncertainty can ever be tolerated, the person has been assigned the task of proving a future. No hearing can be fair under that assignment.

Who Commissioned the Answer

The arbiter's work begins before a difficult case reaches the desk. Someone has determined the purposes for which information will count, what will be treated as a relevant loss and which official can alter the consequence. Some of those decisions may be old enough to look like features of the world. Others may be embedded in a purchased service whose recommendations arrive without the discussions through which their objectives were chosen. The point of identifying a principal is to reach that authority. Merely finding someone prepared to sign the result can leave it untouched.

A model may compare possible fee schedules with an accuracy no committee could match. It may show how much revenue, congestion or participation changes across them, and uncover an option its designers had missed. The committee then has more to judge. A threshold excluding a class of participants does not acquire legitimacy because the exclusion can be predicted precisely. The institution needs to defend the purposes for which it accepts that exclusion, including to those who do not belong to the constituency whose benefit it has optimized. An objection can concern the mandate even when every estimate in the report is right.

The same difficulty enters through apparently personal objectives. A system can learn what someone tends to choose and offer a persuasive account of the satisfaction the choice will bring. The conduct from which it learned occurred under particular alternatives and dependencies. It may reveal a preference; it does not establish that the person has authorized every institution to pursue that preference for him. He can keep a promise that makes him unhappy or reject a profitable role because he objects to its purpose. Assistance might clarify those reasons better than unaided reflection. Its competence supplies no warrant for deciding in advance which reasons an institution is entitled to exclude.

When the principal's choice disappears behind the model's recommendation, purpose has been laundered through a result. A technically excellent system can make the disappearance more persuasive. The ordinary question—why are we doing this?—comes back dressed as an objection to expertise. The affected person is asked to challenge a calculation when the dispute concerns the end for which it has been commissioned. A record of consequential power must therefore preserve the authority for the mandate as well as evidence of its execution. The reviewer needs a way to reach both, even when different institutions control them.

Homo Arbiter names an office with authority to judge such questions and give its judgment effect. It need not reside in a solitary person. A court, panel or other body may distribute the work, using computational assistance at every stage. The Roman terms from which the title borrows distinguished varied forms of entrusted evaluation; they did not oppose an unthinking iudex to an arbiter free to invent fairness. That historical complication remains salutary. A constitutional office needs powers specified closely enough to restrain it and sufficiently real to let it depart from the answer it receives.¹³

It can agree with the system. Consistency may protect applicants against an official's taste, haste or prejudice, and computational comparison may expose a pattern the officials themselves have normalized. Nor does every official possess authority to invalidate the rule being applied. Where the question exceeds the office's powers, the person needs access to a body capable of deciding it. The chain must not terminate in a collection of conscientious explanations that every office has done only what the office before it required. A constitutional limit without a place where it can alter conduct is an opinion held about an institution, rather than a constraint upon one.

The office also needs occasions through which its members acquire and exercise judgment. We encountered this difficulty in the making of productive systems: retaining access to an answer is different from sustaining people able to decide when the answer should change. Here the distinction concerns public power. An institution cannot remove the time, evidence and practice needed for independent assessment, then invoke the official's signature as proof that such assessment occurred. Nor should it make every departure professionally dangerous while treating conformity as self-justifying. The existence of people who can give reasons must be provided for before their reasons become inconvenient.

This is not a claim that no artificial participant could ever hold an office. A future polity might have to recognize beings whose capacity for standing, obligation and injury was not established by the constitution under which we now live. It would have to confront those claims. Present performance does not settle them, and human standing need not await their resolution. No person earns a right to be heard by defeating a machine at the task the institution happens to prize. The political question concerns those subjected to an arrangement, including those whose contribution the arrangement no longer needs.

San Marco

At San Marco, the tapering surfaces beneath the dome gave the mosaicists spaces to fill. Gould and Lewontin made those surfaces the opening of their argument against explaining every biological feature as a direct adaptation for its present use. In their account, the structure supplies the spaces; the decoration occupies and elaborates them. To begin with the fitness of the decoration and explain the structure backward from it would mistake an achieved use for a cause of origin. The architectural terminology became contentious—the surfaces are pendentives, which Gould included under the broader term spandrels—and the analogy did not settle the biological disputes it entered.¹⁴

The distinction nevertheless permits a further question. Once the work exists, what gives its origins authority over its value? The structural account can explain why a surface was available without explaining why anyone should care about what was subsequently made there. It can identify a dependence without exhausting a significance. An achievement can owe its possibility to conditions that were never directed toward achieving it.

Suppose consciousness arose in something like this fashion, through an organization selected under pressures other than the production of felt experience. That remains a conjecture; the alternatives include direct adaptation and relations between organization and experience the metaphor does not explain. Grant the conjecture for the sake of the difficulty. If capacities once associated with conscious life become available through a different productive architecture, why should that genealogy establish that the life has become dispensable? An explanation of how we came to be here has not yet supplied an authority entitled to decide what may be done to us.

It is tempting to answer by finding another service consciousness can perform. Once calculation travels, perhaps imagination remains; once invention travels, perhaps remorse. The defense begins to resemble a negotiation in which the terms of remaining alive are revised after every technical advance. I would refuse those terms. The worth of an experience is not a residual claim on the tasks for which an owner has yet to find a cheaper supplier. A society can release people from work without releasing itself from obligations to them. Whether a particular faculty entered the world as an adaptation, a consequence or something we do not yet understand does not decide those obligations.

The objection is that this judgment is made by the beings whose position it defends. Of course it is. We should distrust the ease with which a species can describe its own advantages as the purposes of the universe. But the rival judgment, that only contributions to a specified output deserve preservation, has a point of view too. It contains a choice of output and a distribution of the authority to choose. Declaring that choice neutral because it can be optimized does not make it less political. It makes the people injured by it harder to recognize as people with an argument.

There is no need to demonstrate a permanent human monopoly on feeling in order to reject that procedure. Were other beings shown to have morally consequential experience, the circle of claims would widen. It would not follow that the existing claimants had lost theirs. Standing cannot be allocated as the prize for whichever kind of being currently performs best, unless we are prepared to accept that the weakest members of every kind possess it only at another's pleasure. A political order committed to equal standing has already refused that allocation among its citizens. Better machinery does not revoke the commitment.

The mosaics at San Marco did not have to explain why arches needed them before they could matter. In a constitutional order, the corresponding refusal is directed at an authority: people need not establish their indispensability to its operations before their lives can limit its purposes. The authority was created within a world of claims. It cannot reverse the relation merely because it has learned to proceed with less assistance from some of the claimants.

The Next Exclusion

A record can be accurate, an institution competent, its officers conscientious, and its continued demand unjustified. Pepper's case makes the distinction unusually visible because the courts discussed the later life while disagreeing over the permission to act upon it. The Supreme Court changed that permission within a resentencing proceeding. The other prisoners remained outside that proceeding. Their absence cannot justify excluding his evidence; his hearing cannot answer for their absence.

We need both capacities: to reconsider a consequence when the governing grounds have changed, and to let a specified matter end. They protect different people at different times. A citizen may need a hearing which the current procedure refuses; the same citizen may need protection from an adversary who will never accept that a hearing is over. An institution must be able to identify what has become final and what has not. A conviction is not a completed jurisdiction over a life.

Much of the practical demand will therefore fall on arrangements which make restraint ordinary. A purpose-specific limit should take effect without a performance of repentance. A temporary intervention should end unless its continuation is justified. A correction should reach the uses that depended upon the corrected premise. A public refusal should not extinguish the claims it had no authority to decide. The people making these limits effective may be assisted by systems more capable than they are at searching the evidence. Their task is to give the limits institutional force, including when doing so relinquishes information or control the institution would find useful.

Usefulness is the hard case. There would be little need for a constitutional boundary if the holder wished to stop at it. An old record can remain predictive, a fuller profile can improve selection, a further restriction can reduce a risk. Those benefits belong in the argument. They do not entitle their beneficiary to settle the argument alone. Beyond the applicable threshold, an institution seeking another adverse use must justify that act under the public rule, before an independent authority and within a bounded purpose. If it cannot, non-use follows. The person is not required to become admirable enough to compensate for the institution's lost advantage.

This is what the office must sometimes enforce: a limit which leaves the past intact and the holder of the record without permission to make it consequential again in this way. There may be no forgiveness, no reconciliation and no persuasive prediction of a better life. The institution must relinquish the use without waiting for the person to become someone it would have chosen.

Source notes

Footnotes

1. Hannah Arendt, The Human Condition (1958), 2nd ed. (University of Chicago Press, 1998), p. 178. See also chapter V, sections 24 and 28, on disclosure and appearance. The book's constitutional use of these ideas does not adopt unpredictability as a definition of freedom. ↩

2. Pepper v. United States, 562 U.S. 476 (2011), opinion of the Court, Part I, including the testimony, findings, remands and separate substantial-assistance adjustments. The initial prison term ended three days after the June 2005 appellate reversal. The 2008 hearing, December memorandum and January 2009 sentence are separate events. Footnote 5 records later imprisonment and release pending the Supreme Court appeal. This account does not invent Pepper's private motives or the experience of his family. ↩

3. Pepper, opinion of the Court, II.A–B and II.D, footnote 17, and III. The Court also invalidated 18 U.S.C. § 3742(g)(2), following the remedial approach of United States v. Booker, 543 U.S. 220 (2005); Booker's constitutional and remedial holdings are distinct. A limited remand may restrict the relevance of later evidence. Neither case establishes a universal right to merciful human discretion or to reopen a final sentence. ↩

4. Pepper, II.C.2, especially the discussion of USSG § 5K2.19, 18 U.S.C. § 3553(a)(6), United States v. Stapleton, 316 F.3d 754 (8th Cir. 2003), and United States v. Jones, 460 F.3d 191 (2d Cir. 2006). The comparison with later offenses describes the evidentiary rule, not Pepper's conduct. The surviving inequality of opportunities for a hearing is this chapter's interpretation of the limited holding. ↩ ↩²

5. Pepper, Thomas, J., dissenting. His policy judgment about further incarceration coexists with rejection of Booker's remedial approach and his view that applying the mandatory scheme on the admitted facts would not violate the Sixth Amendment. The contrast is not between an informed majority and a dissent ignorant of rehabilitation. ↩

6. Immanuel Kant, The Metaphysics of Morals (1797), trans. Mary Gregor (Cambridge University Press, 1996), Akademie 6:337 and 6:460–461. The institutional extension to later uses of records is the present author's. Kant's argument is not explained as a response to costly verification, and the ethical duty to be forgiving supplies no administrative release schedule. ↩

7. Jeffrie G. Murphy, chapter 5 of Jeffrie G. Murphy and Jean Hampton, Forgiveness and Mercy (Cambridge University Press, 1988), pp. 162–186. The application of the justice/favoritism objection to unequal resources for narrating rehabilitation is reasoning about the proposed institution, not a finding about the causes of Pepper's outcome. ↩

8. Douglas N. Husak, “Why Punish the Deserving?”, Noûs 26, no. 4 (1992), 447–464, especially 459–462. Desert and an all-things-considered justification for punishment are distinct. This chapter's further distinction among punishment, collateral burden and justified protection is not attributed to Husak as a modern record-use doctrine. It does not assert that every collateral consequence is legally classified as punishment. ↩

9. The definition and purpose-specific burden are the constitutional proposal developed in the Codex, Chapter 9 and Appendix D.3. They are not a statement of uniform existing law. Expiration, sealing, aggregation and separation can limit active use while preserving appropriately restricted evidence. They confer no universal expiry date. The formal companion does not prove that record-composition costs rise monotonically, that memory becomes unboundedly expensive or that one unique temporal threshold follows mathematically. Appendix D.3, ‘Required records’ and ‘Institutional response,’ supplies the implementation. Where prior notice would defeat an emergency response or protected investigation, notice must follow promptly unless an independent, time-limited order permits further delay. The exception alters notice timing; it does not remove independent review, the bounded purpose and duration, the route of challenge, the receipt, or the prohibition on resetting the threshold. ↩ ↩²

10. Binchi Zhang, Zihan Chen, Cong Shen and Jundong Li, “Verification of Machine Unlearning Is Fragile” (ICML, 2024), PMLR 235, 58717–58738. The result concerns the studied verification strategies, not the impossibility of all unlearning or auditing. Appendix D.3's rule concerns material reliance after the applicable threshold; fresh probability drift, untraceable influence and foreclosure without an expired input remain distinct unresolved problems. The published fragility of the studied verification methods supplies no assurance that all influence can be found or removed. Appendix D.3, ‘Required records,’ includes correction lineage through derived scores, cached classifications, affiliates and downstream recipients where traceable; its following paragraph requires disclosed uncertainty and independent testing where a trained system's dependence cannot be isolated. No particular unlearning method is prescribed. ↩

11. Arendt, The Human Condition, sections 33–34, pp. 236–247; P. F. Strawson, “Freedom and Resentment,” Proceedings of the British Academy 48 (1962), 1–25, especially 9–11. The distinction between public release and a wronged person's forgiveness is the manuscript's institutional argument. The unreconciled relation is a philosophical possibility, not a reported encounter or a claim that private forgiveness extinguishes every legal entitlement. ↩

12. The treatment of prevention is a normative argument. Pepper, II.B, already treats future criminal conduct and protection of the public as relevant sentencing concerns; it does not establish this chapter's proposed procedures for credit, investigations or temporary precautions. Domain-specific evidence, lawful authority, proportionality and effective challenge govern those distinct uses. Prediction may support legitimate action without proving completed wrongdoing. ↩

13. Adam Boóc, “Comments on the Concept of Arbiter in Roman Law,” Journal on European History of Law 10, no. 2 (2019), 133–138. The Roman analogy distinguishes forms of entrusted evaluation; it does not establish a mechanical iudex or an unconstrained arbiter. The dependence of an office on practiced judgment continues Volume II, Chapter 17, without importing an employment prediction. ↩

14. Stephen Jay Gould and Richard C. Lewontin, “The Spandrels of San Marco and the Panglossian Paradigm: A Critique of the Adaptationist Programme,” Proceedings of the Royal Society B 205 (1979), 581–598, opening; Stephen Jay Gould, “The Exaptive Excellence of Spandrels as a Term and Prototype,” PNAS 94 (1997), 10750–10755. The priority of structure in the analogy is not a claim that San Marco's builders had no decorative intentions. Neither article establishes consciousness as an evolutionary byproduct. The claim about political standing is a normative judgment, independent of that conjecture. ↩