There was a moment in the late-nineteenth and early-twentieth centuries when automobiles were being made with steam, electric, and internal combustion engines. As early as the 1840s, inventors had created “[r]echargeable batteries that provided a viable means for storing electricity onboard a vehicle.”Footnote 1 Porsche’s first car was an all-wheel-drive electric car that “set several records” in time and distance competitions, and by 1897 a fleet of taxis built by the Electric Carriage and Wagon Company of Philadelphia was running in New York City.
The internal combustion engine won out because of a conjunction of factors. Henry Ford figured out how to produce and sell the Model T at a price half that of standard electric vehicles. Charles Kettering made crankshaft starters obsolete. And the discovery of large petroleum reserves in Texas dramatically reduced the price of gasoline. The steam engine, on the other hand, took far too long to warm up – no one wanted to wait forty-five minutes before hitting the road. And although electric cars were cleaner, quieter, and initially easier to start than cars with internal combustion engines, they “disappeared by 1935,” as did the kind of investment and research that would have improved their performance and affordability. They disappeared even though scientists knew, as early as the 1850s, that pumping CO2 into the atmosphere would affect the earth’s temperature.Footnote 2
Knowing what we now know about the devastating effects of climate change and other negative externalities of crude oil extraction, one can’t help but wonder about what the counterfactual world of a century powered by electric vehicles would have looked like. The question is for the most part unanswerable – perhaps we would have had a century of lithium wars rather than oil wars. But it is hard to escape the feeling that almost any outcome would be preferable to the world of irreversible climate change in which we now live.
I raise the forgotten story of electric vehicles because it draws into relief an important feature of transformative innovation that is all too often obscured once path dependence and the leverage of market dominance set in. In the moment, there are often many design options, not just one obviously superior alternative, and the full cost (including negative externalities) of any one option can be difficult to calculate. But it is gravely irresponsible not to inquire what that cost might be when design options are still open and policy affecting incentives is being determined. Whatever one might have thought about this responsibility in a time before our own – eras preceding climate change, nuclear holocaust, and other apocalyptic consequences of technological innovation – we do not have the luxury of failing to inquire now.
The tech evangelism that reigns in Silicon Valley inhibits precisely this inquiry.Footnote 3 Uber’s proponents spoke with religious fervor about the corrupt monopoly of cab companies and how that arbitrarily hindered both the mobility of customers and the autonomy of cab drivers. Very little was said about whether disrupting the industry would, on balance, be socially beneficial. But as it turns out, one of the negative externalities is likely diminished use of public transportation,Footnote 4 with attendant effects on climate change and declining resources for innovation in public transportation. Moreover, Uber increasingly behaves toward consumers and drivers like the very monopoly it “disrupted.”Footnote 5 In other ways, the company’s conduct is even more ominous. As one commentator has observed, “if you are one of its regular customers, Uber knows more about you than your own mother does.”Footnote 6 And there are grounds to worry that the company’s use of these data is not entirely benign.Footnote 7 Remarkably, despite its market valuation, the company has yet to prove it can turn a profit. It is far more convenient to hail a ride, and in this sense “access” to a form of transportation has increased. But the cost side of the ledger, as with other forms of “surveillance capitalism,”Footnote 8 is daunting.
A third cautionary tale closer to the innovations in legal automation and access to justice I address in this chapter is TurboTax, software that has arguably revolutionized tax filing. As with taxi drivers and Uber, many accountants have either been displaced by Intuit’s product or incorporated it into the services they provide. For consumers, a process fraught with uncertainty has been made more accessible and efficient. On the other hand, Intuit has repeatedly used its market leverage and lobbying power to prevent states and the federal government from adopting legislation that would make free “prefilled” tax forms automatically available to taxpayers.Footnote 9 The so-called disruptive innovator has become yet another monopoly rent seeker, blocking innovations that would be even more accessible, transparent, and consistent with the public good.
The three examples are commonsense reminders that some forms of innovation are transformative in ways we may (predictably) regret, that the rhetoric and fervor surrounding disruptive innovation can obscure sober assessment of the cost side of the ledger, that some transformations saturated with negative externalities become irreversible, and that the power reallocated by innovation – even innovation that increases access to a good or service – can be used to obstruct both broader public access and regulation.
Other examples could of course be given, enough to make the Panglossian enthusiasm surrounding artificial intelligence, legal tech, and online dispute resolution (ODR) smack of dangerously irrational exuberance.Footnote 10 ODR’s moment, we are told, has arrived, “offer[ing] the promise of robust yet radically less expensive dispute resolution.”Footnote 11 Dispute resolution, we are assured, is no different from other sectors in which digital, online systems optimize information processing:
All forms of dispute resolution revolve around communication and information processing. For the Internet to be adapted to serve the needs of dispute resolution is no different conceptually from adapting the Internet to serve the needs of any other information intensive process, such as online banking, online auctions, online education, etc. Indeed, [these industries] often provide links on their home pages to dispute resolution systems.Footnote 12
The excessive cost, delay, complexity, and confrontational culture of the adversary system, we are told, will be displaced by apps that resolve formal legal disputes as efficiently as eBay resolves auction disputes. In the most ambitious forms of ODR, there will be no more third-party mediators, arbitrators, or judges, and therefore no more conference rooms or courthouses.Footnote 13
Justice will roll down in strings of code. Disputes suitable for ODR will not only be resolved cheaply and quickly; the conjunction of data mining, predictive analytics, and dispute systems design will help prevent disputes from arising in the first place.Footnote 14 Academic conferences populated by scholars who are funded by or work in the very industry they write about are conducted with all the sobriety of an Elmer Gentry revival meeting. Skeptics are dismissed as unrepentant sinners – elitists too attached to the ancien regime to appreciate the miracles of the “internet society,” heartlessly indifferent to the crisis in access to justice, captive to self-serving, anachronistic ideas about the administration of justice. The capital offenders are judges and lawyers whose skepticism is dismissed as rationalization covering the prestige and monopoly rents they derive from the status quo. Fundamentally, law is not thought to be different from transportation or any other target of disruptive innovation: If you want to improve access, deregulate, disregard regulation that can’t be set aside, and expand competition by letting the information economy and disruptive innovation work their magic.
Anyone familiar with the history of professions knows that power struggles between professionals often involve characterizing an entrenched group of experts as corrupt and the newcomers (here, software engineers) as avenging angels whose primary care and concern is the welfare of others.Footnote 15 New assertions of power are in this way masked by the discourses of progress and reform. At the same time, anyone familiar with the history of the Anglo-American legal tradition knows that legal elites have commonly and, it must be said, deservedly been targeted for many of the perceived flaws of the adversary system. The first movement to reduce law to code in America occurred in the early-nineteenth century, when Jacksonian populists launched an all-out assault on the complexity of common law pleading and the power of judges and lawyers.Footnote 16 Drawing on the Napoleonic Code and the famous anti-lawyer tracts of Jeremy Bentham, they sought to replace the common law with a democratically enacted code. They simultaneously sought to expand access to the practice of law by eliminating standards for entry to the bar. Finally, they sought to make judges democratically accountable through popular election and recall – procedures that endure to this day in many states. This was the most pronounced anti-lawyer movement in American history, animated by a desire to make the law simple and more affordable. But it is scarcely the only one. Shays’ Rebellion pitted agrarian debtors against lawyers and judges who enforced the claims of elite creditors. The New Deal pitted progressive reformers and proponents of administrative agencies against conservative courts and the adversary system.Footnote 17
What the anti-lawyer and anti-adversary system rhetoric of the current movement obscures is that the bar’s protectionist arsenal is weaker than it has been at any point since the Jacksonian populist threat more than a century ago. Indeed, both practically and doctrinally, the bar’s defenses have been decimated over the last fifty years. The network of price controls, minimum fee schedules, and restrictions on advertising and other rules that limited internal competition and “external” lay-lawyer combinations were dismantled by a series of landmark Supreme Court cases in the 1970s.Footnote 18 Competitors in banking, accounting, and other fields renounced the “treaties” that kept them from offering competing services at the same time. The legal-form business expanded under the protection of First Amendment decisions insulating it from unauthorized practice rules.Footnote 19 Mediators and especially private arbitrators now operate with the blessing of the Supreme Court, allowing corporations to use contracts of adhesion to displace millions of disputes from courts to alternative dispute resolution forums every year.Footnote 20
In fact, the status quo in access to justice is as much if not more the product of neoliberal defunding and restriction of legal services for the poor and defunding of state courts.Footnote 21 Tellingly, some of the same players who never lifted a finger to help low-income Americans obtain meaningful access to the adversary system or support funding of the court system are now enthusiastically supporting ODR as a substitute.Footnote 22 They are joined by liberal ethicists, lawyers, judges, and scholars who never much liked the adversary system to begin with – believing that ADR was the way of the future, that the New Deal vision of centralized, rational technocratic agency adjudication is more efficient and suitable to mass-processing of claims, or that, whatever the alternatives, adversary adjudication is morally flawed.Footnote 23 These are strange bedfellows, united by a shared desire to replace the adversary system – at least for people who cannot already afford it.
Once we understand the neoliberal aspects of the status quo, and liberals’ gradual abandonment of the goal of providing lawyers to poor people and funding courts, there is reason to question why ODR travels under the banner of access to justice and whether it will serve the people its advocates claim they care so much about. After all, if the interests of poor people were truly motivating these reforms, the law already recognizes the right of non-lawyers to create organizations that fund and coordinate not-for-profit legal services.Footnote 24 The failure to innovate in this space suggests that rent-seeking packaged in Silicon Valley “solutionism,”Footnote 25 not access to justice and the needs of ordinary people who stand before the law, is paramount in the current movement.
In the pages that follow I set the debate about AI and ODR on a different plane by granting that access to some form of “law” will be expanded. AI, data mining, predictive analytics, and the widespread use of mobile computing devices are generally superb tools for reducing the cost of large-scale bureaucratic and logistical tasks. They are already proving genuinely transformative in other sectors of the economy and society. There remain, however, questions about how the architecture of these information systems fits with basic ideas about the structure of due process and the rule of law in a pluralistic, democratic society. Ultimately, these are questions about what version of “law” ODR will increase access to and what kind of justice and what kind of legal subject are produced by these systems.
To begin with, ODR advocates and designers tell us there are many “simple” cases that don’t require adversary resolution, but when we look more closely, what we see is that simplicity is assumed from and defined by the monetary value of the case, not the simplicity of the issues involved or the social, moral, and economic stakes for the litigants and the public. Given that small value cases make up the vast majority of American litigation, ODR actually poses a direct challenge to the courts as sites of adversary adjudication for ordinary people. I take up the “simplicity” hypothesis in Section 11.1.
Section 11.2 describes the architecture of ODR that enables efficient mass processing and resolution of legal claims. Although ODR designers and promoters tell us that it is cheaper and faster than litigation and that it helps ordinary people with “simple” disputes, current ODR systems are cheaper and faster mainly because they replace decision on the merits with easier-to-code “interest-based” negotiated resolutions. Equally situated parties may reach mutually advantageous resolution on such platforms, but where the state or powerful creditors are pitted against unrepresented individuals, as is true in most small value claims, “interest-based” resolution may simply enhance collection compliance with respect to debts and other legal claims that the defendant has a right to resist. Far from expanding access to justice, ODR may rather ominously accelerate unwarranted compliance and legal subordination. When the government itself is a creditor, and court systems therefore profit from collection, the risk of conflicts of interest in adoption of ODR systems that the public cannot evaluate is especially acute. In other cases, even though ODR may be cheaper than litigation, fee-for-service structures appear designed to induce early resolution rather than merits resolution for people of limited means.
Finally, the conjunction of AI’s predictive analytic power and big data allows both public and private ODR system to achieve something dispute systems designers have long dreamed of: using data about existing disputes to prevent conflict in the future. As alluring as dispute prevention may seem in a conflict-ridden society, I argue in Section 11.3 that embedding dispute prevention and compliance into the architecture of the administration of justice through an automated system of surveillance and information control is inconsistent with human freedom. Asimov’s second “law of robotics” states that “a robot must obey the orders given to it by human beings,” not that humans must obey the orders of robots. Compliance-oriented and preventive ODR reverse this law, turning right holders into cogs in a machinery of compliance. I discuss alternative paths of ODR development and associated regulatory reforms to protect the due process rights of litigants, basic rule of law values, and the integrity of the administration of justice.
11.1 The Domain of ODR: “Simple” Cases?
They have no lawyers among them, for they consider them as a sort of people whose “profession it is to disguise matters; and therefore they think it much better that every man” should plead his own cause …. [T]he plainest meaning of which words are capable is always the sense of their laws. And they argue thus.
A foundational premise of current ODR systems is that there is a class of “simple” cases for which adversarial resolution is inappropriate, not least of which because the costs of the process exceed the value of these cases as measured by the size of the claim or judgment. The simplicity hypothesis has deep intuitive appeal. There are indeed many cases in which the costs of adversarial process exceed case value.Footnote 26 If these cases can be disposed of through an ODR system that does not require physical appearance in court, or even a judge, we are told, costs of adjudication are reduced for the state and perhaps for the parties. Resolution is faster and cheaper, giving the parties peace of mind sooner and the ability to move on with their lives. More complicated disputes might require the formalities of adversary procedure but due process can sometimes be provided without full judicial process. Finally, ODR will promote access to justice because it is precisely these simple cases in which ordinary people do not have access to counsel or avoid litigation altogether, believing that the game isn’t worth the candle. Ordinary people will thus be better off as ODR expands.Footnote 27
This simplicity/access nexus is pervasive in the ODR literature, the promotional materials of ODR vendors, and the webpages of court systems that have adopted ODR systems. One commentator observes that “the use of ODR to settle small claims and therefore free up judges and courtrooms for more complex cases is a given: ‘Small claims courts, with smaller dollar amounts and less complex issues, are ideally situated to transition their operations online.’”Footnote 28 ODR, the author continues, “has proven to significantly reduce the delays and costs normally associated with a court case by eliminating the need for travel and synchronous communications.”Footnote 29 Until the development of ODR, another advocate writes, “the average experience of a litigant ‘going to court’ amounts to … waiting in long lines” – if and when “a hearing actually begins, it is over almost at once. The outcome is generally predictable … as the decision is determined by standard pieces of information contained in the case file or provided by answers the litigant supplies.”Footnote 30 This is not only hugely inefficient, “access to justice is subverted by the fact that courts continue to operate on the age-old model.”Footnote 31 The adversary system
makes much more sense for complex litigation in which credibility determinations … and diverse forms of evidence are standard fare. For disputes of this character, the costs of physically using a courthouse (even day in and day out) are relatively modest, if not negligible, given the stakes of the lawsuit .… But for minor disputes in state court, in which the stakes are at least initially fairly low and decisions can be made on the basis of papers and are usually straightforward, the tradeoff cuts deeply the other way.Footnote 32
Supposedly “[m]inor legal disputes” account for the majority of state trial court caseloads in the United States.Footnote 33 These include not only small claims cases between private litigants and landlord/tenant disputes, but “lesser misdemeanors and civil infractions” where the state is involved as the complainant or prosecutor.Footnote 34
The first problem with the simplicity/access nexus is the assumption that low value cases are in fact simple. They can be made to seem simple by comparison to more sophisticated causes of action, but for the parties involved in “simple” cases, this is an irrelevant comparator. For parties to a case, the judgment whether their case is simple rests on factors such as
• whether they have dealt with similar disputes before and are familiar with the applicable law and procedure,
• the relative value of the case as compared to their other assets and debts,
• dignitary considerations linked to the harm they have suffered or are accused of causing as compared to other wrongs and emotionally charged problems they have dealt with,
• dignitary considerations linked to the degree of participation and understanding litigants have about the process that results in resolution of a dispute
• the capacity of designers of ODR systems to capture the interests of the parties, and
• the actual merits.
With respect to the first consideration, an experienced landlord likely has dealt with tenant disputes in the past and can therefore navigate legal issues in a new dispute with some degree of comfort even if the matter is not free of frustration. By contrast, a tenant who has never had an abusive landlord will not likely regard even a small money value dispute about a repair, return of a deposit, or penalty for late rent as “simple.” With respect to relative value, the second consideration, even a seemingly “minor” fine or civil liability can loom large for a person of modest means, forcing painful choices about whether to put food on the table or pay up to avoid mounting fines and fees and continuing intervention on the part of the state. The stakes won’t seem “minor” to such a party.Footnote 35
Dignitary considerations are indeterminate and subjective, but an extensive, well-established body of social scientific and cognitive research makes clear that these considerations are central to the perceived legitimacy of dispute resolution systems. Quick resolution of emotionally charged cases in which people do not feel heard can have enduring negative repercussions not only for the parties, but for the administration of justice. As Alan Lind and Tom Tyler summarize, “people usually feel more fairly treated when they have had an opportunity to express their point of view about their situation.”Footnote 36 This is just as true, they note, in “simple” cases as it is for “complex” ones: “In small claims cases … all parties to a case would like to have an opportunity to tell their story, taking as much time as they feel they need to articulate the issues that matter to them.”Footnote 37 Unlimited participation obviously is not possible in any dispute resolution system. Finality matters to fairness. Judges are obliged to impose other limits as well, including legal relevance and consideration for the time that must be spent on other cases.Footnote 38 Nevertheless, Lind and Tyler emphasize, the tension between “objective justice and legal efficiency,” on the one hand, and “the experience of subjective justice on the part of the litigant,” on the other, doesn’t evaporate just because a case is a relatively “simple” one from the perspective of judges and other dispute systems designers.Footnote 39
They offer, as example, proceedings in a traffic court in Chicago:
Judges in that court often take the view that showing up for court and losing a day’s pay at work is punishment enough for a traffic offense. As a result, those who arrive in court often have their case dismissed without any hearing .… However, interviews with traffic court defendants suggest that despite these favorable outcomes they often leave the court dissatisfied. For example, one woman showed up for court with photographs that she felt showed that a sign warning her not to make an illegal turn was not clearly visible. After her case was dismissed (a victory!) she was angry and expressed considerable dissatisfaction with the court. … Outcome-based models might find the woman’s dissatisfaction difficult to explain, but process-based models would have little trouble in accounting for her reaction.Footnote 40
The feeling of not being given the time of day, of not being heard, can ramify in other forms of civic engagement, and in other dealings with the state. If they are tied to deeper mistreatment at the hands of the state, the consequences can be grave.Footnote 41
Even a simple case can also prove quite complex to code in an ODR system. Complexity in coding arises not only from the costs of good ODR design but from limits in natural language processing capacities in even the most advanced AI systems and other algorithms.Footnote 42 These systems are still capable of fairly comical errors in the interpretation of human language.Footnote 43 In some settings these errors present mere inconveniences – your Uber driver arrives late or drives to the wrong location. In law, the consequences can be catastrophic – including erroneous arrest and separation from one’s family, destroyed credit, even the use of deadly force in executing a bad warrant. Natural language processing capacities will improve gradually. However, as with any symbolic system that attempts to reproduce, measure, and operationalize content from another symbolic system (here, human language), there will always be gaps. All symbolic systems err, and all are to one degree or another ineluctably “leaky.”Footnote 44 This derives from the nature of representation itself – the fact that representation depends upon reduction of signified content to signs. Automated systems that displace human judgment offer advantages, but they also remove the possibility of real-time commonsense reconsideration.
This raises the question whether even truly “simple” cases are simple enough for ODR systems to handle.Footnote 45 The answer to that question cannot be found by comparing ODR to the cost of adversary resolution or to cases assumed to be more complex and therefore more suitable to adversary resolution. It can only be answered by taking stock of the merits.
Much of the adversary system is designed not only to maximize participation and party control (albeit for those who can afford it), but to avoid the problem of prejudging the merits – the temptation to decide a case based on reductive first impressions. This temptation is strong, reinforced by powerful cognitive biases (including confirmation bias, halo effects, etc.). A prominent nineteenth-century lawyer and judge famously admonished in defense of the adversary system that “the affairs of mankind are not so nicely adjusted as that one party in a law-suit should be entirely right and the other entirely wrong .… [T]ruth cannot be elicited and justice awarded unless both sides of a case are fairly represented.”Footnote 46 This is so not only because of the “intricacies” of “commercial relations,” the moral complexity of human action subject to legal regulation, or the “nice distinctions to be made in determining the degree of criminality,” but because long experience shows that “[m]any cases which at first seemed to be bad have on examination proved to be good.”Footnote 47 The adversary system thus rests on what one might call procedural skepticism – rules of procedure that reduce the risk of prejudgment.
The idea that there is a general category of “simple” cases and that the merits of these cases reveal themselves on first impression or on the initial pleadings is, from this perspective, a seductive fiction. It rests on a self-serving value judgment about small-dollar-figure claims – that they matter less and contain less complexity – and the triage imperatives of mass-processing.
As importantly, evidence from the last decade provides sobering proof that supposedly simple cases (small claims, landlord/tenant, and traffic and other misdemeanors) are not actually simple and that they have profound ramifications for the administration of justice. Firstly, we know that these cases dominate state trial court dockets, and that, in the vast majority, the individuals involved do not have counsel. So ODR is staking a claim to displace most state court adjudication, not experiment with a small subset of claims. We also know that, regardless of specific subject matter, these are mainly debt collection proceedings in which the plaintiff is either the state seeking to recover “legal financial obligations” (fines, fees, and other penalties imposed by the court) or a creditor (a lender, collection agency, or landlord) represented by counsel. There is thus a powerful repeat player (the state or a private creditor) on one side, and an individual defendant/debtor on the other. This contrasts sharply with the image of the facilitation of cooperative resolution between individuals painted by the ODR literature.Footnote 48
11.1.1 Public Enforcement in “Simple” Cases
With respect to public debts, the Department of Justice Investigation of the Ferguson Police Department and follow-on litigation against municipal courts around the country show that there has been widespread abuse of legal financial obligations as cash-strapped municipalities deprived of general funds by their states have converted courts into fee-for-service systems parasitic on the most financially vulnerable populations within their jurisdictions. In Ferguson, excessive fines and fees were imposed disproportionately on the African American population of the city. The Report’s section on the court system found that enforcement actions involved shocking deviations from basic principles of procedural due process, including failure to make the constitutionally required inquiry into litigants’ ability to pay before using imprisonment as a penalty or inducement for nonpayment.Footnote 49 Litigation in dozens of other jurisdictions reveals that the problems with setting and collecting legal financial obligations are widespread.Footnote 50 One report revealed that the
courts in St. Louis city and the county collected over $60 million in revenue in 2013 … with some cities depending on such fines for more than 40 percent of their general fund. The report found that the cities most dependent on such revenue were majority African-American with large impoverished populations .… In Jennings, which has a population of roughly 14,750, [a] lawsuit found that the city had issued about twice as many warrants as there were households, “mostly in cases involving unpaid debts for [traffic] tickets.” In 2013, a 24-year-old inmate in the Jennings jail who was imprisoned for unpaid tickets hanged himself.Footnote 51
“Tragic incidents such as these draw into vivid relief the human cost of ignoring the connection between procedural due process and human dignity”.
As local courts have faced funding crises and failed to address racial bias, advocates of ODR have been marketing their platforms to court systems around the country and in the pages of law reviews by packaging enhanced efficiency in collecting public debts in the language of “access to justice.” ODR advocates insist that in traffic cases faster, cheaper resolution benefits individual defendants because they get closure and avoid the additional fees associated with failure to appear and default.Footnote 52 But when one examines the use of the platforms marketed to court systems, one finds things like the Fort Collins, Colorado, municipal court’s ODR system for “camera radar/red light tickets.”Footnote 53 The ODR system provides for resolution of traffic camera tickets.Footnote 54 The sophisticated cameras cost $10,000 a month each to operate. Tickets are $75.Footnote 55 The city claims that the cameras are making a “huge difference” in driver education and safety, and it points to data showing reductions in accidents at one of the intersections where the cameras are used.Footnote 56 But the program also reportedly nets the city about $200,000 a year over operating costs, including payments to the private contractor who services the cameras.Footnote 57 Camera citations well exceed officer written traffic tickets in the city, and in other jurisdictions there is evidence that this enforcement tool invites politically corrupt outsourcing to private contractors and generates revenue without improving traffic safety at all.Footnote 58
The Fort Collins ODR system does not appear to follow a pure bargaining model where the resolution results from negotiating what will be paid irrespective of the merits. Nor, however, does it appear to be designed to fully and faithfully ascertain the merits in each case, including educating defendants about how to explore and assert standard defenses. The defendant is assured there is a prosecutor who will review the case and the website’s “About Online Case Review” page indicates that a defendant could “potentially have … fines and fees reduced or in some cases, dismissed altogether.”Footnote 59 The FAQ page, however, characterizes ODR as a process to enter a guilty plea through an “Online Plea” process.Footnote 60 The page indicates that “You can plead guilty, be sentenced, and pay your fines/costs without going to court in person.”Footnote 61
If the system is mainly designed to enter guilty pleas and collect the fines and fees – if, that is, the private “review” conducted by prosecutors and the court is perfunctory, designed to “improve compliance” on the same terms as the designer’s platforms sold to other jurisdictionsFootnote 62 – then the efficiency gains accrue mainly to the city.Footnote 63 In cases where guilt is unclear or revenue generation dominates public safety priorities, defendants are saddled with unwarranted debts and the public with rent-seeking law enforcement. Indeed, a compliance-oriented ODR system that primarily increases the speed of collection for traffic camera fines raises the specter of fully automated law enforcement in derogation of every procedural value other than reduced cost to the state.Footnote 64 Nor does such an ODR enforcement process provide a public forum, as court proceedings do, for airing systemic concerns about whether revenue generation dominates legitimate public safety concerns. Public and media scrutiny of enforcement is literally short-circuited.Footnote 65
Remarkably, ODR advocates commonly ignore or suspend the merits question altogether – either not seeking to measure it, excluding it from the design of ODR systems, or both.Footnote 66 Traffic cases are indeed minor relative to felony cases, but the lesson of Ferguson and broader litigation about excessive fines and fees is that incursions on civil liberty and civil rights in the design of dispute resolution systems for these offenses can be substantial. Recent scholarship on criminal law and procedure reinforces this conclusion, showing that twenty-first-century misdemeanor enforcement has been used to criminalize poverty, to impose onerous systems of regulation and continuing supervision on marginal populations, to feed mass incarceration, and to subordinate racial minorities through biased forms of “order maintenance,” policing, and punishment.Footnote 67 To submerge these distortions in the administration of justice in automatic collection-compliance processes would obviously be inconsistent with the mission of enhancing access to justice for ordinary people.
11.1.2 Private Enforcement in “Simple” Cases
Even when the state is not a party to supposedly “simple” cases, recent empirical studies show that the simplicity hypothesis is untethered from the realities of the administration of justice. As ODR advocates describe small claims, it would be easy to assume not only that private individuals are on both sides of the litigation, but that in many “simple” cases people gin up “highly emotional conflicts over matters with relatively low monetary value.”Footnote 68 This happens, to be sure, but the data show that far more often an individual unrepresented defendant is sued in small claims court by a powerful creditor or landlord represented by counsel. The greater leverage of these plaintiffs is mobilized not to force costly merits adjudication, but rather to accelerate reduction of a claim to a final judgment and proceed with enforcement.
The National Center for State Courts Landscape of Civil Litigation study in 2015 found that three-quarters of all judgments in the state courts were less than $5,200.Footnote 69 Even among non–small claims cases that went to trial, “[t]hree quarters of judgments entered in contract cases following a bench trial were less than half of those in small claims cases ($1,785 versus $3,900). This contradicts assertions that most bench trials involve adjudication over complex, high-stakes cases.”Footnote 70 Whatever the nature of modern trial, most civil cases, the study emphasizes, were “disposed of through an administrative process,” and for cases that reached judgment, the most common resolution was a default judgment.Footnote 71 In these cases, then, there is no meaningful deliberation on the merits, often no hearing whatsoever preceding the entry of judgment.
Most revealingly, the Landscape Study found that
[t]he vast majority of civil cases that remain in state courts are debt collection, landlord/tenant, foreclosure, and small claims cases. State courts are the preferred forum for plaintiffs in these cases for the simple reason that in most jurisdictions state courts hold a monopoly on procedures to enforce judgments. Securing a judgment … is the mandatory first step to being able to initiate garnishment or asset seizure proceedings. The majority of defendants in these cases, however, are self-represented.Footnote 72
The conjunction of self-representation and default judgments in small value debtor-creditor disputes suggests that some state courts are operating as accelerated debt collection forums. In the forty-four states where judges, clerks, magistrates, and justices of the peace are allowed to issue capias warrants for failure to appear at post-judgment asset examination hearings, defendants in civil cases face arrest and incarceration with bond often set equals to the debt owed.Footnote 73 As with misdemeanor criminal cases then, civil litigation for people in financially precarious situations can result in restraints on liberty in order to force payment.
On this evidence, as with the use of ODR for misdemeanor cases and legal financial obligations to courts, the simplicity/access nexus looks quite ominous. The access-to-justice problem is not how to speed things up and substitute automated bargaining over settlement value (or private deliberation on facts adduced through strictly circumscribed online submissions) in place of public inquiry into the merits in small value cases. It is (1) how to slow creditors and other repeat players down in order to ensure attention to the merits and (2) how to address the systemic disparities in power that shape both the debts being collected (click-wrap and other contracts of adhesion; payday lending schemes, etc.) and enforcement procedures (default judgment, capias, etc.). ODR systems oriented toward speed and automated online resolution may be quite attractive to plaintiff creditors and other repeat players, but for individual defendant debtors, the risk is great that ODR systems will not include adequate exploration of defenses available under the relevant contract, lease, or state and federal consumer protection and fair debt collection practices laws.Footnote 74 These defenses are not generally regarded as “simple” by experts. A 2010 FTC report found that even basic affirmative defenses such as state statutes of limitations “on filing actions to recover debt are sometimes variable and complex, and generally not understood by consumers.”Footnote 75 An allegation of identity theft raised by a debtor can “increase the complexity and time required” to resolve a matter.Footnote 76 The Truth in Lending Act’s enforcement structure – which contemplates use of the statute as a counterclaim in a debt collection proceeding – is famously “confusing.”Footnote 77
Enough has been said, I hope, to make clear that small values cases are by no means simple or low stakes either for the parties concerned or the administration of justice. Nor are these cases small in number – they compose a substantial part of court dockets and therefore of the cases handled by the adversary system. The simplicity hypothesis is false. If there is a class of truly simple claims suitable for ODR, the hard question is how to define standards for accurately identifying them without having to adjudicate the merits along the way – the very task the avoidance of which makes ODR so affordable. As matters currently stand, then, the principal effect of using low money-value claims as a proxy for simplicity will be to produce a bifurcated system of justice – one in which low- and middle-income people already priced out of meaningful participation in the adversary systemFootnote 78 will have no alternative but to avail themselves of ODR systems. This bifurcation in the administration of justice will merely formalize, encode and multiply, not mitigate, problems of access to justice.
In the most ambitious ODR systems – those that remove the third party neutral human decision-maker altogether – low- and middle-income people will receive justice defined by software engineers unregulated by standards of judicial ethics except to the extent that courts supervise their outsourcing contracts. And we know that supervision will be limited by the fact that the best AI systems to date are opaque in their operation even to their designers.Footnote 79 So, for instance, a deep learning system used to generate “reasonable” settlement values might not be explainable – even by the engineers who program it. Systems that are more transparent because they rely on expert design rather than deep learning, on the other hand, remain highly reductive.Footnote 80 Absent regulation, this bifurcated system for the administration of justice will flourish on terms that limit assessment and accountability.
11.2 The Architecture of ODR: Access to Justice, Harmony, or “Just Harmony”?
The last section focused on the pervasive assumption that ODR will enhance access to justice because it is suitable for so-called simple cases as measured by money value. In this Section, I shift from the nature of the cases ODR systems regularly handle to the design of the systems themselves in order to examine exactly how the architecture of ODR promotes efficient resolution.
11.2.1 ODR’s Design Features
Although the overall ecosystem is heterogenous and evolving rapidly, most current ODR platforms function by breaking dispute resolution into its component phases, reducing legal forms and rules to plain language questions, instructions, and guidance, and then using asynchronous electronic communication, document storage, retrieval, and review to facilitate the flow of information needed to define and resolve the dispute, as shown in Table 11.1.
Table 11.1 Five phases of ODR
| Phase | I Opt In Phase | II Diagnosis/Intake | III Negotiation | IV Third Party Neutral/ Algorithm | V Closure / Opt Out |
|---|---|---|---|---|---|
| Process | Information to users about process and applicable substantive rules | Information gathered from users |
|
|
|
In the first phase there is typically a webpage with plain language text, videos, and graphical information describing the specific ODR process. In some instances, general information on the law is provided (at varying levels of detail). In others, there is little or no information about the underlying rights and defenses that apply to the dispute or resources for ascertaining what the law is. The emphasis is instead on identifying the parties’ interests. Rechtwijzer, the Dutch online ODR system that, until its demise in 2017, handled divorce, family law, debt, and neighbor disputes, explicitly encouraged the parties to approach the process through the framework of interest-based resolution.Footnote 81 Divorcing parties were initially informed of “rules such as those for dividing property, child support and standard arrangements for visiting rights so that they could agree on the basis of informed consent,” but the software’s negotiation framework prioritized interests, not rights.Footnote 82
British Columbia’s vaunted Civil Resolution Tribunal (CRT) includes a statute-of-limitations information page that defines what a statute of limitations is and advises parties that it is not tolled during the first phases of the CRT process. But it does not answer the question of what statute of limitations applies to the dispute the party has, how to determine either when it started to run, or when it will expire. The page ends with the statement: “We can’t answer these questions for you. You may want to get legal advice. The CRT can’t provide legal advice.”Footnote 83 Indeed, CRT’s detailed downloadable forms to help parties prepare for negotiation generally ignore rights and defenses, instead prompting the parties to identify and articulate their interests.Footnote 84
To generate enthusiasm and induce users to opt in, introductory pages of ODR websites also include substantial promotional content highlighting perceived advantages of ODR relative to adjudication in court. The CRT home page begins: “Save time, money, and stress! The CRT lets you resolve your dispute when and where it’s convenient for you …. The CRT helps you resolve your dispute quickly and affordably.”Footnote 85 Rechtwijzer, the online divorce settlement platform, billed itself as providing a chance to “separate together” without the adversity and acrimony of divorce in court.Footnote 86 ODR websites generally do not counterbalance their self-promotion regarding the advantages of online resolution with sober assessment of potential downsides, clear notice about rights waived, or alternatives.
In Phase II, once users have opted in, questions are directed to the users through a range of graphical interfaces in order to gather basic information about the dispute, classify it, populate relevant forms, and upload relevant documentary evidence. Advanced systems use chatbots or other interfaces modeled off of apps people use daily on their smartphones and other devices. Matterhorn, the ODR system behind the Colorado traffic camera enforcement discussed in Section 11.1 has an online process for “the uploading of statements by parties, law enforcement and court personnel from afar and in lieu of court hearings.”Footnote 87 In this phase, assuming the design for information sharing is not reductive, there are obvious efficiency gains in cost, time, and convenience associated with “asynchronous processes” for fact gathering relative to filing papers in court.Footnote 88
In the negotiation phase, Phase III, there is a wide range of approaches. Some, such as CRT’s, simply provide online portals for direct, unmediated negotiation between parties with some basic ground rules about abusive communication and guidance about how to prepare and conduct a successful interest-based negotiation. Others introduce a human third-party neutral to guide negotiation. Still others are automated and more rigidly structured, channeling the parties into online negotiation in the form of blind bidding with an algorithm that is designed to identify an optimal settlement from the highest one party is willing to offer and the least the other is willing to accept.Footnote 89 This ostensibly allows parties to “overcome tactics often employed in face-to-face negotiations that hinder reaching an agreement despite the existence of a ‘zone of possible agreement.’”Footnote 90 Smartsettle’s software forces parties to “list their interests and assign numerical values to them, thereby creating a weighted spectrum of issues” from which an algorithm “generated various ‘packages’ or combinations of issues that might satisfy both parties” along with “a graph as a visual display of the level of satisfaction each package of issues represented for the parties.”Footnote 91 Overall, both public and private ODR platforms share some of the following algorithmically automated features at varying levels of sophistication: “identifying dispute types; exposing parties’ interests; asking questions about positions; reframing demands; suggesting options for solutions allowing; allowing some venting; matching solutions to problems; and drafting agreements.”Footnote 92
In the phase IV, a third-party neutral human expert may be brought in online to review and approve a deal reached between the parties. This was the case with Rechtwijzer. If online negotiations resulted in an agreement, an independent lawyer would be brought in to review the deal “to ensure that it meets legal requirements and is fair to both parties.”Footnote 93 If no agreement was reached, a mediator or arbitrator would be brought in online to help facilitate resolution. Costs were a fraction of traditional costs for counsel in divorce proceedings in court, and “[e]ven at the higher end,” costs were “lower than the costs to the Legal Aid Board for representing both parties to divorce proceedings in court.”Footnote 94 In the near future algorithms may not only structure negotiation and suggest solutions but displace the third-party neutral altogether by using artificial intelligence to review and validate negotiated solutions.Footnote 95
In the fifth and final phase, an agreement reached online is reduced to an enforceable contract or court order. Generally, ODR negotiations cannot be introduced in court, but there are many exceptions. CRT participants, for instance, lose confidentiality if they use what the system designers determine is “abusive” language in settlement negotiations. ODR websites also vary widely in the promises they make regarding the use of user data (information such as identity, location, negotiation documents and communications, history of participation in ODR or other litigation, etc.), sale of such data to third-party data brokers, and how and on what terms such information is shared with court systems with whom they contract and other government entities. Generally, there is no public access to the ODR process in real time, no public ODR docket, and unlike codes of procedure and evidence in courts and the decisional law interpreting them, the lines of code in ODR systems are proprietary, insulated from public review, and treated as trade secrets by ODR companies.
11.2.2 ODR as Interest-Based Dispute Resolution
What can be inferred from this structure about ODR as a form of dispute resolution? What theory of justice does its architecture embody? The most important thing to notice is that the overwhelming emphasis in the structure of ODR systems is on “integrative negotiation” or interest-based, win-win bargaining, rather than the merits.Footnote 96 The “archetypical ODR approach is to provide an online forum and tools to facilitate the full settlement of claims without any human intervention.”Footnote 97 However, users unfamiliar with law are not in a position to recognize that ODR is oriented toward integrative resolution rather than resolution based on their legal rights and defenses.Footnote 98
Although even the earliest theories of interest-based negotiation insisted that interests must be “legitimate” to warrant consideration, and that “community concerns” are relevant to bilateral negotiation,Footnote 99 some ODR systems eliminate these factors entirely. Blind bidding and other mathematical representations of interests have this effect – resolution arises from overlapping settlement values defined by the parties irrespective of any independent assessment of how the merits relate to the parties’ quantification of their interests. In other ODR systems, such as CRT and Rechtwijzer, attention to the legal merits and third-party effects is backloaded into the review phase (phase IV) when a third-party neutral steps in. But at that stage, rejection of a settlement entails losing the efficiency gains accumulated through the ODR negotiation process. Delayed intervention of the third-party neutral thus virtually guarantees the subordination of concerns about the merits and negative externalities imposed on third parties and the community. The prize of agreement, once reached, is exceedingly difficult to refuse. As importantly, fact development to that point may not illuminate defects even for a third-party neutral deeply comitted exercising independent judgment and disposed to resist the allure of the prize.Footnote 100
Even systems that bring a third-party neutral in earlier may not improve attention to the legal merits. As the orientation of many third-party neutrals is toward efficiency, “harmony and overcoming conflict,” not the legal merits,Footnote 101 their disposition, training, and docket pressure may lead the them to “‘trade justice for harmony.’”Footnote 102 This has long been a criticism of the alternative dispute resolution movement, sharpened by evidence that by privileging “harmony” in pluralistic societies (where there are competing conceptions of the good and systematic marginalization of minority groups), interest-based resolution can “suppress … or silence … the voices of those without political power.”Footnote 103 In cases involving public debts to courts, revenue generated from these cases may diminish impartiality, attention to the merits, and community concerns. Some ODR systems are marketed to courts on precisely these terms, touting increased collection rates, decreased defaults, and reduced time to collection. For example, Matterhorn’s contract with the court in Washtenaw County, Michigan, for traffic violations helped the company expand ODR to dozens of other counties on the back of reports that it produced a more than 40 percent increase in fines paid within thirty days.Footnote 104
Although facts can generally be elicited efficiently online relative to live hearings and paper filings, important information can be lost in shifting from a hearing/forms combination to text-only,Footnote 105 writing skill is a built-in advantage for high-literacy litigants,Footnote 106 and without tailored advice about how the law applies to the facts of their case, litigants may omit relevant evidence. Everything hinges on how questions posed to elicit the facts are framed by the ODR platform. On the other hand, the more detailed the questions and guidance, the more expensive the code. Even the most advanced current AI systems thrive in bound environments, not open-ended, indeterminate ones.Footnote 107 The temptation, therefore, is ever to manufacture a bounded environment in coding for dispute resolution – reducing complexity in order to code efficiently.
Modria’s brochure for courts considering adoption of its ODR software, proudly declares that it “helps courts resolve all manner of case types faster without sacrificing accuracy,” but if we take the architecture of ODR systems seriously, the truth is that it is quite difficult to assess the accuracy of ODR systems.Footnote 108 Indeed, the earlier cases settle and the more interest-based bargaining drives resolution, the less certain we can be that any given settlement is a just reflection of the relevant rights and defenses of the parties.Footnote 109
11.2.3 ODR Funding
Funding remains one of the greatest barriers to the development of ODR and, ironically, one of the weakest pillars in its claim to expand access to justice. Many systems have failed because they have not produced a durable financial model despite receiving public subsidies and favorable user reviews.Footnote 110 The systems are costly to develop, and both the law and technology change in ways that make it costly to keep the systems up-to-date. One prominent provider, CRT, therefore relies heavily on a fee-for-service model. The parties using CRT pay not only initial filing fees, but fees for every subsequent phase of the process, making the process more expensive as it unfolds if no negotiated solution is reached.Footnote 111 CRT also charges a substantial fee ($200) to file a notice of objection in a small claims dispute.Footnote 112 This means that for a small claims dispute involving $1,000, challenging a settlement would cost 20 percent of the value of the dispute (on top of $75 to file the complaint, $75 to add a claim against a third party, $10 per records request, and $50 to switch from negotiation to an adjudicator).Footnote 113 Fee structures like this induce people who have financial constraints to resolve disputes quickly and accept a resolution that may be unjust. People of means, on the other hand, can afford to take advantage of every avenue of relief.Footnote 114
Other ODR systems are free to usersFootnote 115 and presumably paid for by courts. Contract terms between courts and private vendors of ODR software are not generally available to the public.Footnote 116 This makes any comparison with the costs of alternatives to ODR such as civil GideonFootnote 117 difficult to assess. It also contrasts sharply with the public budget process of funding courts as well as rules of judicial ethics that strictly regulate financial conflicts of interest in the judiciary.
11.2.4 ODR Party Structure
Although the ODR literature focuses on small value civil law disputes between roughly equally situated, unrepresented private parties, ODR is currently being used for disputes that involve parties of radically unequal status. These include criminal disputes between the government and individuals, where ordinary people are negotiating with prosecutors and judges regarding fines and fees the nonpayment of which could lead to incarceration, disputes between landlords and tenants, property associations and owners, creditors and debtors, contractors, subcontractors and home owners, nonunion employees and employers, insurers and parties harmed in an accident,Footnote 118 and couples who brought vastly different resources into the marriage and are unequally situated.Footnote 119 Some ODR systems permit parties to be represented if they can afford to do so, introducing potentially vast disparities in the level of expertise that one side brings to negotiations. Experience with other ADR systems such as arbitration is sobering on this front, indicating that powerful repeat-players not only fare better but are able over time to shape ADR procedures to maximize their interests.Footnote 120 Nor do the opt-in pages of ODR systems adequately disclose information to support knowing and informed waivers of due process rights and other rights a party would have in court.
In sum, democratically enacted and public rules of procedure and evidence are displaced by the proprietary technology of ODR systems.Footnote 121 Lawyers who work in principal-agent relationships for clients, and who have expertise in the rules of procedure and evidence, are replaced by software engineers whose highest loyalty is to investors and, secondarily, to the courts who retain them to help reduce costs. There is no malpractice cause of action for users against ODR providers, and no enforceable ethical code requiring fiduciary duty, competence, diligence, loyalty, communication, or confidentiality. Users of ODR platforms, who are constitutional right holders in the adversary system, become mere tertiary beneficiaries of contracts entered between courts and ODR providers. And as with other forms of alternative dispute resolution, ODR replaces the public appearance and reason-giving function of the judge either entirely (in systems that displace these decision makers with AI) or partially by changing the scene of adjudication from a public courtroom to a private online forum where third-party neutrals facilitate or impose resolution behind encrypted walls of code. We know that the reasons for conducting adjudication publicly in adversarial space are not exclusively performative.Footnote 122 Doing so helps prevent corruption, bias, and arbitrary decision. Justice, the saying goes, must not only be done, but must be “seen to be done.”Footnote 123
No one should therefore be surprised to see ODR systems that serve principally to induce settlement and improve compliance with judgments on behalf of the state, powerful creditors, and other well-resourced, represented parties, while failing either to establish the rights and defenses of unrepresented parties or make transparent potentially systemic abuses of the laws being enforced.
11.3 ODR and Preventive Justice
This chapter has so far focused on pragmatic concerns about current ODR systems. In closing I want to raise a more fundamental challenge: ODR systems rest on a theory of justice at odds with liberal democratic principles of the rule of law. For decades the dream of alternative dispute resolution advocates has been to design systems that not only resolve disputes when they arise, but to use information about such disputes to detect patterns of conflict and prevent them from occurring in the first place.Footnote 124 The ultimate goal is preventive justice, a culture of seamless compliance. ODR offers the means to realize this goal because it relies on the technology of algorithmic governance – coding that relies on big data to predict and steer human decision-making.Footnote 125 Whatever the appeal of preventive justice within a corporation for its ability to induce compliance with internal corporate norms and external regulations, or in other domains such as public health, as a general theory for the administration of justice it is destructive of human freedom.
Outside authoritarian regimes, the default rule for the administration of justice is that the law intervenes in the lives of ordinary people (1) after wrongdoing, not before, and (2) in response to specific instances of wrongdoing, not the broader collective conditions giving rise to them. Strict standards of substantive liability generally apply to inchoate offenses. Ex ante intervention and regulation are possible, as is structural relief addressed to conditions that repeatedly cause serious harm, but they are both exceptional – higher standards have to be met to authorize these remedies – and they are generally reserved for misconduct on the part of the state and regulated entities, not ordinary people.
These default rules can be found in the law of procedural due process, remedies, the prohibition on prior restraints against free speech, and so forth. Their purpose is to protect the sphere of social action from domination by the state and others who have the means to maximize enforcement of their interests and bend the law to their will. The adversary system embodies these default rules. It grounds the administration of justice not in a substantive concept of justice but in decentralized, participatory, public, ex post adjudication.Footnote 126
ODR systems are oriented toward a very different theory of justice. Their proponents are quite frank about this, drawing on the work of earlier analog dispute systems designers. Ury, Brett, and Goldberg famously argued from the study of wildcat strikes that employee-employer conflict and ex post resolution costs could be avoided by studying patterns in those labor disputes and altering institutional structures to reduce conflict. For dispute system designers, conflicts are unfortunate and avoidable events; rights assertion and adjudication are disfavored, costly projects that amplify adversity. With the correct information, well-calibrated interest-balancing, and appropriate ex ante interventions, both social conflict and conflict in court can be avoided.Footnote 127 The true promise of the ODR is thus not merely to streamline dispute resolution, it is to develop products that realize the potential of dispute systems’ design to prevent conflicts from arising in the first place – to instantiate a culture of compliance. As one commentator puts it, just as the value of Uber is the relevance of its data to a future market of automated cars, “the seeds of an effort to prevent disputes may lie in the technology employed to resolve disputes.”Footnote 128 How? “[T]he use of technology provides ODR with more opportunities to identify systemic contribution to conflict and systemic opportunities to reduce conflict.”Footnote 129 These opportunities arise from the capacity of ODR systems to exploit the information they gather about pending disputes to provide “automatic detection of problems, obviating the need to passively wait for complaints to arrive and allowing proactive remedying of the problem even before a potential complainant has been made aware of its existence.”Footnote 130
If this sounds futuristic, it is not. Private ODR systems are already valued by corporations for their capacity to prevent customer, user, and employee disputes.Footnote 131 It is now “commonplace” for private ODR systems to “captur[e] data and analyz[e] it for insight into the disputing environment of a particular institution to help prevent future disputes.”Footnote 132 The objective is to identify patterns and sources of conflict, anticipate new ones, and snuff them out in advance. It’s one thing of course when the lens is directed inward, to identify policies and practices of a corporation that are producing conflict or misconduct.Footnote 133 But it is quite another to direct the lens outward, combining the information saturation of digital interaction and predictive analytics to continuously monitor and manipulate the preferences, choices, and norms of users.
Private ODR already involves both – internal and external preventive surveillance, intervention, and restructuring of code to optimize compliance.Footnote 134 For companies like Airbnb, the
effortless recording of large amounts of data relating to anyone operating on the site creates a huge database that can be cross-checked with information on problems and resolutions, generating unique insights on how to structure more satisfactory transactions, what problematic patterns need to be dealt with, what rules and practices require clarification or amendment, and which participants require mentoring or instruction.Footnote 135
In order to promote transactions that “are less likely to generate problems,” the company may alter things like “which listing to display first” to a particular user.Footnote 136 There are plenty of nondigital examples of decisions like this inviting or reflecting bias – for example, one renter gets access to “A list” properties from an agency or landlord based on race, ethnicity, or assumptions about credit or past experience with the renter; another does not. The difference in an online platform is that the user may have no idea she is not seeing all the available properties in a location – code presents a smooth surface relative to live interaction that can obscure not only the fact that the user’s information is siloed, but also the basis for the decision to do so, and thus, crucially, the opportunity to challenge it in court or any other forum.
Conflict prevention grounded in unjustified restriction of users rights which cannot be detected and challenged by the user thus creates a double silo – limiting choices as well as information about the limitation that might subject it to legal scrutiny. Conflict may be reduced, but in the manner of a falconry hood. In some settings the stakes may be relatively trivial, but errors in other settings can have significant life consequences.
Preventive justice is actively used in other areas of private ODR such as automated content moderation of speech on social media platforms, employment, and preventive medicine. Automated prescreening systems have become a significant form of content moderation.Footnote 137 They have the capacity to significantly reduce online disputes, but they can easily err by failing to incorporate important “contextual information” such as language and cultural differences that affect semantic meaning.Footnote 138 There is evidence that these preventive systems have overenforced copyright law to the detriment of fair use, and they “raise concerns about the limits of public speech, cultural sensitivities, and individual rights.”Footnote 139 In the domain of medical prevention, genetic data can save lives, but it has also been incorrectly used to exclude children from school to prevent the spread of disease.Footnote 140 In the workplace, dispute prevention efforts can improve safety and performance, but they can also amplify the power of employers, undercut the rights of workers, and inhibit worker mobilization and unionization.Footnote 141
In public ODR systems, the lines between dispute resolution and prevention are already “increasingly being blurred.”Footnote 142 The data gathering and processing technology underlying ODR provides means previously unavailable to extend preventive intervention beyond the boundaries of individual organizations to the administration of justice.Footnote 143 Commentators emphasize that data produced by ODR systems “will enable … the identification of large-scale trends and patterns we have never seen before, often through automated means instead of human analysis …. Those in control of these large data sets will be able to analyze the data and … prevent the occurrence of a dispute.”Footnote 144 Indeed, “[b]y overcoming the need to rely on the aggrieved party’s ability to recognize and pursue a remedy, a larger portion of society’s problems can be addressed and prevented regardless of the aggrieved party’s awareness of his or her injury.”Footnote 145 For example, the data gathered by ODR systems and data being aggregated to inform predictive policing may prove lucrative for tech companies to combine and market, and tempting for prosecutors and law enforcement to exploit.Footnote 146 Or someone who loses the confidentiality protections of CRT’s online mediation process because she uses what the system unilaterally defines as “abusive” language may find herself classified by any number of databases relying on predictive analytics as mentally unfit, a safety threat, a credit risk, unsuitable for hiring or admission into a training or credential program, and so forth. These cross-referencing effects are possible in analog record-keeping, but with big data and the power of the algorithms driving predictive analytics they can disseminate pervasively, instantaneously, and without public transparency.
The “politics of the preventive has a strong element of irresistibility built into it, since it generally appears perverse to argue against a preventive measure. Who could be against the prevention of harm? … [C]ritics … can be portrayed as courting insecurity and jeopardizing public safety” and harmony.Footnote 147
The problem is that the same class of elite software engineers who delight in disruptive innovation and flagrantly disregard regulation is developing ODR systems that would embed compliance in the architecture for the administration of justice for ordinary people. Doing so in the name of access to justice is not just ironic or hypocritical, it masks a potentially transformative struggle over the nature of the rule of law. Preventive compliance by code might be welcome if dispute system designers were right that legal conflict is usually a sign of social malaise, and if software engineers were uniquely adept at distinguishing legitimate from illegitimate conflict and resistance to law.
But neither proposition is true.
Not all resistance to law and noncompliance is a sign of delinquency or a threat to the body politic; in fact, noncompliance can be one of the very highest forms of civic engagement. Our most important progress in racial justice is the product of nonviolent civil disobedience. Indeed, on some accounts of liberal democratic theory under conditions of value pluralism, “justice is conflict.”Footnote 148 Law without resistance is not law – it is domination, simpliciter, a panoptic prison of code. And software engineers have no particular expertise (or incentive) to correctly draw lines between socially desirable and undesirable rights assertion or defiance. No one does. That is precisely why the adversary system vests decisional authority in decentralized, democratically accountable jurors, judges, and regulators subject to appellate review, and, for the latter two, establishes legally enforceable standards of impartiality and professional conduct.Footnote 149
ODR systems alter this structure – placing a single, professionally, and democratically unaccountable group of elites in control of the procedures for dispute resolution. Using these proprietary systems to prevent, not merely resolve, disputes would subvert the structure altogether. What has traditionally been understood as an exceptional form of the administration of justice would become the rule, at least for those priced out of the adversary system.
I hold no crystal ball, but we don’t have to guess about the ominous authoritarian implications of big data in the hands of tech companies and the state.Footnote 150 ODR firms are currently modest in market capitalization. This gives courts and bar regulators some leverage in negotiating terms. Reliance on AI is nascent, reducing some of the problems for court administrators and the public of opacity in assessing how these systems run. But we know from other sectors of innovation that it doesn’t take long for market concentration to emerge, and AI is evolving rapidly in ways that may amplify opacity. We also know that when tech companies consolidate a market, they not only exclude competitors in anticompetitive ways;Footnote 151 they use their position to exclude inexpensive public alternatives. This is precisely what TurboTax has done.
But these are not inevitable outcomes. ODR systems can be improved in a variety of ways if judges, court administrators, and state bar regulators recognize that we have arrived at the legal equivalent of the combustion engine–electric car design decision. On one path, unwarranted faith in innovation will lead to deregulation or lax regulation. Rent-seeking in the administration of justice will masquerade as innovation. By the time the costs of this path become clear, ODR providers may be too powerful to control. Alternatively, ODR could support and preserve human judgment on the part of all participants by exploiting the technology’s efficient information gathering capacity while providing greater transparency. To achieve this ODR systems have to:
• be opt-in;
• include strict disclosure and consent requirements to validate opt-in decisions;
• protects the privacy of user data and require publication of ODR rules and system design;
• develop more precise heuristics for separating cases appropriately eligible for ODR from those that are not;
• elevate procedural values in ODR system design other than compliance and efficiency, especially merits assessment;
• strictly distinguish interest-based negotiation platforms from adjudication platforms;
• require that mediators, judges, and prosecutors document their reasons on record for resolutions they reach in ODR systems;
• gather and report performance data to promote assessment of biases and other distortions in relation to established standards of impartiality;
• reduce the costs of appeal;
• publicly disclose court/ODR firm contracts.
A third path would permit ODR systems to resolve disputes without human mediators or adjudicators subject to heightened scrutiny under the above regulatory standards. A fourth would refocus courts’ access-to-justice initiatives on some combinations of paths two or three and increasing direct funding for access to counsel. Whatever the path, ODR’s costs and benefits must be placed in relation to a tangible innovation baseline. All too often ODR’s proponents compare it to “no justice” – the position of a party who cannot afford to appear in court or cannot afford a lawyer. But just as new medical interventions and drugs are assessed in relation to an evidence-based standard of safety and efficacy, not the condition of the untreated, ODR should be judged by comparison to standards of procedural fairness and the costs of alternatives such as expanding access to counsel.
In sum, innovation worthy of the name should improve the administration of justice for ordinary people, not just impose different and potentially more tragic trade-offs than the adversary system and traditional forms of alternative dispute resolution. ODR proponents contend that without the freedom to experiment, the best designs may never be developed. But just as any automobile must have more than an efficient engine, ODR must do more than end disputes for ordinary people quickly and cheaply.
The promise of ODR has prompted courts and bar associations to rush to deregulate under the banner of exemptions that promote experimentation.Footnote 152 Solemn language about the importance of consumer and public protection and risk assessment can be found in these materials, but all too often without reducing these lofty principles to concrete design parameters. ODR is no longer on the salt flats. It seeks to transform the rule of law. The brave new world will not be a form of code as law in the conventional sense – the way code architecture structures online behavior. It will be a (re)codification of law, of the administration of justice itself. We leave design parameters to unregulated engineers at our peril.
Open access