12.1.1 Where the Story Begins: Courts, ODR, and Access to Justice

Less than ten years ago, state courts began implementing ODR platforms for high-volume, small-stakes cases, like alleged traffic violations and other cases where physical resolution of the matter could be very costly to one of the litigants (e.g., outstanding failure-to-pay warrants), making remote resolution particularly attractive. One of the very first ODR platforms, if not the first, to be widely adopted in US courts was Matterhorn, which Michigan adopted starting in 2014.Footnote ²²

The earliest Matterhorn ODR implementations sought to resolve disputes between individuals and the government: traffic, civil infractions, warrants, and a couple of misdemeanors. These ODR platforms were in many ways designed to “mimic” traditional proceedings, with a few key differences: ODR was made available at any time of day, information was exchanged asynchronously by text, and litigants could access the platform with any online remote device. But in most respects, litigants could expect to experience a process similar to going to court on their own. These systems inform the litigant of the charge or issue, “ask” the litigant whether they would like the court to review their case, and then ask the litigant to answer questions, including an open-ended invitation to “let the court know” anything else they felt was potentially relevant. These implementations are easy to use and make the process much easier to follow than what a litigant walking into a courthouse should expect to face, but at their core, they are an online version of a pro se courthouse experience.Footnote ²³ The individual locates their case, makes a request for relief, and answers questions from the court.Footnote ²⁴

Matterhorn ODR soon grew to address disputes between private individuals. Early examples include ongoing family court cases, where parties negotiate in the online presence of a case manager without the need for in-person hearings,Footnote ²⁵ and small claims disputes, where parties use an online text-based negotiating (chat) space (and perhaps online mediators) to engage in informal discovery or settle their case outright. These ODR implementations sought to reduce default by making it easier for everyone to come together outside of court in an informal way before being compelled to come together in court. Parties usually negotiate a resolution, and by making online negotiation and mediation free, easy to use, and court-sanctioned, courts were soon playing an important role via ODR in a significant fraction of cases. What’s more, some tentative evidence hints that early engagement with court-connected ODR promotes traditional courtroom engagement later, which might have social benefits,Footnote ²⁶ even if a case ultimately ends in face-to-face litigation.

Regardless, first-gen ODR works by moving courtroom or court hallway activity to a more amenable online venue that can be streamlined/individualized for the type of case. A pro se litigant may find it easier to negotiate with a counterparty or contest a civil infraction through an ODR platform, but such litigants are just as “pro se” during an online interaction as they would have been during an in-person tête-à-tête.Footnote ²⁷ Whether first-gen ODR moves into asynchronous text-based exchanges or real-time video-based hearings (or any number of other ways parties can communicate, make offers and demands, share/present information, and work through outcome-determinative decisions), existing ODR platforms can make dispute resolution easier, faster, cheaper, and therefore more accessible, but the value of legal advice is likely to matter just as much as it does in traditional settings. True, more “have-nots” will be able to use courts and access the law. But the “haves” (with their lawyers, experience, and mastery of the rules of the game) will remain just as privileged (if not more so, given digital divide concerns) relative to their “have-not” counterparts as before.

Empirical evidence on ODR outcomes is consistent with this story. This expanding literature speaks to the improvements enjoyed by pro se litigants (as opposed to represented parties): Cases proceed more quickly, default rates plummet, and communication between parties, court staff, and judges increases.Footnote ²⁸ This accessibility leads to more accurate outcomes (primarily by reducing default), giving more muscle to substantive law and reducing the role litigation costs play in determining legal outcomes.Footnote ²⁹ In the case of Matterhorn, at least, litigants also report that they find ODR proceedings fair, the online process easy to understand, and the software straightforward to use.Footnote ³⁰ On the whole, ODR-based improvements in pro se litigant outcomes (both quantitative and qualitative) appear significant, although these advances are relative to a pro se baseline where the relevant control groups of litigants are also unrepresented.

12.1.2 From Courtrooms to ODR: Responding to Critics

Moving from courtrooms to ODR is no panacea, of course. Critics raise a number of concerns, including the importance of the courtroom and human interaction in how litigants experience justice, the ability of efficiency-oriented designers to reduce choice and induce compliance, the lack of transparency in ODR hearings, the rigidity and inflexibility of procedures, and many others.Footnote ³¹ Most of these complaints, however, focus less on ODR’s basic principles and more on the weaknesses critics observe in specific first-gen systems.Footnote ³² Like traditional procedures, ODR will only be as good as its design. Yet ODR’s inherent flexibility gives courts the means to resolve many worries raised about existing systems, if they have the will: for example, all ODR-related policies, choices, and outcomes can be made public, and the workflow can be designed to maximize litigant voice, even if current instances of ODR score poorly on these measures. Furthermore, ODR is almost always an option, not a requirement, so those who object to ODR must defend disallowing low-cost access alternatives for litigants who retain more expensive “Cadillac” access to justice and yet prefer ODR when given a choice.Footnote ³³

More importantly, most criticisms fail to reckon with the very real problems of traditional adjudication, especially in low-stakes cases. Detractors compare ODR processes to a stylized and idealized version of traditional adjudication.Footnote ³⁴ But this perspective ignores the existing brick-and-mortar system’s huge access-to-justice issues – including regressive costs, biases, disparities, confusion, and intimidation – as well as the paper-thin procedures that many if not most litigants experience in courtrooms. These critics compare the best of courts to the worst of ODR.

Some criticisms rightly focus on ODR being peddled by the private sector and the implications of long-term profit motives.Footnote ³⁵ But there is nothing inherent in ODR as a concept that requires deferring to a private vendor; courts can and do build their own ODR systems (albeit, often via work-for-hire, which is par for the course when it comes to government software adoption). There are potentially negative, more indirect consequences to making courts and the law easier to access, too. Once lawsuits are extremely easy to file, answer, and navigate, more disagreements will become lawsuits. While ODR can be designed to balance access costs for plaintiffs and defendants, for instance, it may actually increase litigation rates. But the argument that making courts more efficient and accessible in a party-neutral way might be socially problematic simply proves too much; a corollary would be that courts ought to move back to pen-and-paper orders and file cabinets.

12.1.3 Auguries: First-Gen ODR and the Representation Gap

With respect to pro se litigants and the “representation gap,” at least some evidence suggests that ODR systems, even first-generation ones, can level the playing field just as retaining legal representation does. For instance, ODR proceedings in the traffic context are associated with fewer race- and age-based disparities, either because switching to online proceedings (for example, text-based, asynchronous communication or at-home, video-based hearings) differentially benefits some groups over others or because ODR platforms reduce implicit bias by limiting decision-maker exposure to legally irrelevant identity traits.Footnote ³⁶ The precise mechanism here is unclear, but this reversal is important and ties into questions about the benefits of corporeal representation in court. Lawyers “represent” their clients and so may disrupt implicit biases or other cognitive/behavioral distortions that attach to litigants who appear in person on their own. ODR may do this, too. At the same time, by offering litigants an alternative to a single, one-size-fits-all track, any single limitation a litigant might have in, say, communication (e.g., difficulty speaking in open court) is less likely to matter. In this sense, the addition of ODR makes our legal system more robust in its ability to accommodate pro se litigants.

Existing ODR platforms also implicitly provide guidance on the law and litigation process to litigants. For instance, ODR hearings are configured for the litigant and legal issue in question. All content is targeted and laid out in a single dashboard, and material that speaks to irrelevant contingencies can be omitted. Now consider a litigant’s experience walking into a state courthouse, which is, by design or happenstance, an all-purpose space we use to resolve many types of legal matters. A courtroom is not hung with customized signs. The litigant must digest, sort, and navigate. This is costly and often confusing and scary. By contrast, ODR is akin to being met at the door by an usher who helps you to where you need to be, reminds you of your to-dos, and tsk-tsks when you cut a corner or appear distracted. ODR makes the experience less stressful and reduces error. Lawyers also play this role. Thus, even first-gen ODR mitigates the effect of pro se status, but ODR today is better analogized to an usher than to a counselor. A lawyer can help you decide whether going to your seat is a good idea; an usher can just tell you how to get there.

12.2.1 Experience and Predicting Outcomes

One of the most important services a lawyer provides a litigant is an informed, experienced perspective on the likely outcome of a case, with some overall take on whether the outlook is “good.”Footnote ⁴⁴ Typically, during intake, the prospective litigant describes the nature of the dispute, key aspects of the case, and the like, and the lawyer may even offer some preliminary advice. Once retained, however, the lawyer will engage in more intense prediction-oriented activity on behalf of the client, perhaps after researching the law, collecting more information, and weighing the necessary investment, probable venue, expected legal rulings, possible judges, and even likely jurors. These predictions are necessarily fuzzy; there is considerable uncertainty, but as time resolves many contingencies (e.g., the identity of the opposing attorney), predictions will become more accurate. Consequently, the lawyer is likely to revisit the prediction task repeatedly, presumably at each important decision point, when the litigant might seek advice on how to proceed. In a traditional litigation setting, the prediction problem is a complicated one; the environment is relatively unstructured, and the variables are nearly infinite. There are also conflicts of interest and cognitive biases tugging on a lawyer’s predictions.

A lawyer’s ability to “predict” a dispute’s outcome builds on the human equivalent of supervised learning.Footnote ⁴⁵ In most cases, the client defines a “good” outcome for the lawyer.Footnote ⁴⁶ But even if the target is unclear or moving, the client and the lawyer are normally decided on the type of outcome that matters (e.g., damages, liability, share of custody, opportunity to be heard, etc.). The lawyer’s task is to take the client’s account, the law, the client’s set of plausible “moves,” the presiding judge, and anything else observable to the lawyer and then predict an outcome using the agreed-upon metric of success. This prediction may come in the form of an expected outcome alone and possibly with some measure of uncertainty that implicitly conveys the expected accuracy of the prediction (e.g., “I think you’ll get $75K, give or take $10K”). If the client defines a minimally acceptable outcome, the lawyer will review the likely consequences for each set of potential moves and either report that this outcome is unrealistic or that taking a particular set of actions will be necessary (under current circumstances).

How does a lawyer do this? The lawyer effectively cognitively “models” prior cases to understand the complicated relationships between facts, choices, and outcomes. Put another way, lawyers detect patterns between predictors and outcomes.Footnote ⁴⁷ The greater the number and variety of cases and types of predictors that lawyers incorporate into their modeling, the more flexible, unbiased, and accurate their predictions can be. A mental model is what we mean when we talk about the value of experience. An expert lawyer has witnessed many cases with many different characteristics proceed under many different circumstances. Delivering accurate predictions is helpful because they can guide a litigant’s behavior toward their preferred outcome, whatever it happens to be. If the litigant’s goal is to maximize their net “return” from a dispute, a schedule of litigation strategies (including their costs) and their likely outcomes (and uncertainties) produced by a lawyer is what a litigant needs to succeed.

ODR platforms can perform this task. In fact, under the circumstances, they may be able to do a better job.Footnote ⁴⁸ First-gen ODR platforms already access an individual litigant’s case, drawing information from the court’s case management system. Once a case resolves, the platform archives the record and moves onto the next case. But there is no reason an enhanced platform cannot build models from all prior cases and even from data outside the court’s case management system. Models can incorporate any plausible predictor, including judge identity, hearing timing, and docket backlog. By contrast, lawyers build models only from their own cases or those about which they hear or read (inevitably with error and uncertainty), and only with what they can observe about those cases and their circumstances. In effect, an enhanced ODR platform could display a distribution of how similar past cases – or, with enough data, only identical ones – fared using the litigant’s preferred measure of success. These “predictions” (tendered as descriptions of what happened in prior cases – not as promises or legal advice!) can be presented at decision points and can be joined with statistics to impart precision.Footnote ⁴⁹

The challenges to operationalizing these ideas are manageable.Footnote ⁵⁰ In many ways, litigation that occurs in an ODR environment may be more amenable to accurate prediction than litigation in a more traditional environment because the platform can easily capture the information about cases, choices, and outcomes needed for prediction. An ODR system may be more structured and less nuanced, either by necessity or design, which might also simplify the prediction problem. Consider an enhanced platform with asynchronous, text-based communication that requires parties to answer the same questions in the same order with the same answer options as part of an online hearing (versus open court with a judge asking different questions in different ways and expecting a narrative response on the spot) and where the system necessarily prevents or obscures certain actions (e.g., a litigant can’t visibly sigh or roll their eyes at a judge).Footnote ⁵¹ The number of predictors would no longer be infinite because the environment eschews legally irrelevant details; in other words, the medium itself might simplify some of the prediction problems a lawyer would normally face in court.Footnote ⁵²

12.2.2 Navigating, Counseling, and Identifying Options

Lawyers often present options. They take raw facts and abstract law and suggest possible legal solutions. They also identify various strategies (or implement them directly), and they help clients weigh the upsides and downsides of those strategies. Lawyers do not necessarily perform this task well. The quality of counseling, when it occurs, is often suspect. Lawyers implicitly build models, but they also suffer from biases (e.g., availability), have limited observational and processing capacity, and must navigate their own conflicts (like economizing on time and avoiding high-risk innovations). In most contexts, lawyers are also busy. Identifying options, discussing them with clients – these activities take time. Consequently, lawyers work from scripts,Footnote ⁵³ stocked with standard “moves” that usually do the trick in the vast majority of situations. Lawyers may deliver these options in rote fashion, with little thought paid to details that are individually unlikely to make a material difference in the long run (even if, collectively, such details can alter a legal dispute’s trajectory).

This account is designed to present a human lawyer as a fairly simple algorithm. The lawyer asks intake questions, processes the answers, identifies the nature of the dispute and a set of typical solutions, explains the most common approaches to the client, and translates the client’s reaction into a set of choices or next steps. Two tasks are occurring here. One is pattern recognition (the main point here). The second is prediction combined with an interpretation of what the client values. Put differently, lawyers determine the sort of dispute they face, collect relevant information for a dispute involving such features, identify an array of potential approaches or strategies, and assist the client in understanding and deciding between them.

This last step – helping a litigant understand and evaluate various options – falls squarely in the camp of outcome prediction. Enhanced ODR can define a large number of potential outcomes. Not all, true, but easily those outcomes that would matter to 99 percent of litigants. Counseling between options can be carried out simply by informing litigants of the likelihood that certain outcomes will manifest if the litigant pursues, say, track 1 or track 2.

With a real lawyer, counseling involves nuance.Footnote ⁵⁴ But much of that nuance is the lawyer attempting simultaneously to ascertain a client’s preferences. An enhanced ODR platform would do less “screening” and instead make much more information available to the client so they may digest with their own priorities in mind instead of relying on a black-box lawyer. There are challenges here: Litigants also face cognitive limits, and while they may have more time to immerse themselves in their own affairs, too much data can be worse than too little.Footnote ⁵⁵ Just as an intimidating courthouse can paralyze, so too can an environment that dumps unstructured information on litigants and then asks them to sift it.Footnote ⁵⁶ Thus, ODR platform structure and design must not only be accessible but also support effective decision-making.Footnote ⁵⁷ We already use technology to help people make important choices about many things.Footnote ⁵⁸ Even in the legal-aid and access-to-justice contexts, existing decision support tools streamline processes, minimize confusion, and improve decision-making.Footnote ⁵⁹ Efforts are imperfect but improving,Footnote ⁶⁰ and one must remember that the relevant benchmark is a pro se world with no support whatsoever.

To counsel, lawyers categorize a dispute and then compile the information necessary to identify options, avoid legal and factual pitfalls, and assemble plausible strategies (matched with predictions). This is complex, and early versions of enhanced ODR will likely be subject-matter specific. But precedents for algorithmic technology of this sort already exist, and courts should avoid letting the perfect be the enemy of the good. Legal aid groups already deploy triage and intake forms that are “smart.”Footnote ⁶¹ An initial goal for next-gen ODR might be to determine a litigant’s eligibility for digital or other assistance by collecting critical information about the litigant’s situation.Footnote ⁶² The law’s many branches can be incorporated directly into questions put to users.Footnote ⁶³ Structure in the law and implicit groupings of fact patterns will lay bare most likely strategies.

There is no limit to how complex any enhanced ODR data collection process might be. But this is also true when a litigant hires an attorney: The litigant explains their situation, and the attorney asks questions in response. The attorney may also look over papers, review video, or examine other evidence. This exchange might take a while, but lawyers ultimately triage and end their information gathering, and ODR software can too. As natural language processing improves, faster intake methods that ask open-ended questions instead of closed-ended questions can reduce the total number of questions.Footnote ⁶⁴ Analysis of narrative answers can trigger relevant follow-up inquiries.Footnote ⁶⁵ But there are reasons doctors don’t only ask “so, what’s bothering you?” at an appointment. Instead, you are first subjected to a long list of questions designed to prompt you to think about every system and part of your body.Footnote ⁶⁶ Any natural language response to an open-ended question like “what’s your legal problem?” may come from a person unfamiliar with the law whose take on what matters may be off-base.Footnote ⁶⁷ Still, natural language responses, perhaps even voice submissions, might be a better, more accessible option for some litigants.Footnote ⁶⁸

A barrier to making data-collection functionality successful in next-gen ODR platforms is simply how long it might take a litigant to get in the front door. So much of what is good about ODR is that it is easy to use and minimally costly to try. Next-gen ODR might dissuade people from using courts all over again if it bogs down with seemingly countless questions and checklists. But consulting with a lawyer takes no small amount of time, and stacks of forms are not new to human institutions. Nevertheless, enhanced ODR platforms might be more attractive to more people if guidance and counseling functions were designated as optional.

12.2.3 Replication/Obfuscation/Translation versus Representation

Lawyers aren’t just analytical input-output machines offering a litigant legal expertise in the background. Lawyers are corporeal and play a physical role on the stage of our justice system.Footnote ⁶⁹ They actually “do” law by going to a courthouse to serve as their client’s mouthpiece. The corporeal representation function of lawyering matters in at least three ways in the context of court-connected ODR: lawyers replicate a litigant’s ability to be physically present, lawyers obfuscate litigant traits, and lawyers translate and distill a litigant’s lay arguments into an “expert’s” legal arguments. Considering these roles separately demonstrates that enhanced ODR can advance these functions through text on a screen or other digital-information display even without a human body in a courtroom.

First, lawyers allow clients to be in two places at once. This function reduces the cost of litigation to clients, courts, and other parties, since schedule conflicts can slow litigation. Representation offers flexibility to meet competing obligations. Often, a client and their attorney come together in a courtroom at critical junctures, but there are many stages when a lawyer can physically “be” the client for purposes of advancing a case. Technically, ODR is unable to fulfill this function because, whatever it is, software does not stand in for a litigant before other parties or decision-makers. Yet, in an ODR environment, there is arguably no need for such a service in the typical case. Indeed, unlike hearings that take place at a particular time and place, ODR can be more flexible, accessible, and efficient, making the use of a lawyer to “cover” for someone less important.

Second, lawyers can “obfuscate” client traits. When judges walk into a courtroom and see a litigant’s race, gender, age, or other identity traits, evidence suggests that unconscious implicit bias, if not explicit bias, affects subsequent decision-making.Footnote ⁷⁰ Furthermore, the anticipation of such bias can alter a litigant’s ability to communicate well in a courtroom. One strategy for a litigant facing the possibility of discrimination is to hire a lawyer with different – perhaps socially “preferred” – characteristics to block, blunt, or attenuate any such bias. Lawyers even advertise on this basis – for example, hinting at the advantages of a female lawyer defending a male defendant accused of sexually assaulting a female victim.Footnote ⁷¹ True, having a lawyer does not necessarily preclude a judge from ascertaining litigant identity traits, but outside of explicit, intentional discrimination, even sharing the stage with an in-group lawyer may disrupt the psychological processes that lead to bias.

For those unable to reduce their exposure to bias by using a lawyer,Footnote ⁷² ODR offers a substitute, perhaps superior solution. ODR can structure online communication to insulate judges and other decision-makers from legally irrelevant information that might trigger implicit biases (or allow purposeful discrimination). At least some first-gen ODR systems, for example, do not include a litigant’s driver’s license picture in the judge’s dashboard. At least one court collects required picture ID information only after the judge decides the case but before the ruling is entered. First-gen ODR systems, by tracking traditional in-person processes, still tend to make a litigant’s name visible (which may reveal gender and ethnicity) and, in some instances, date of birth (which, with some math, reveals a litigant’s age).Footnote ⁷³ ODR platforms can go further by thoughtfully, comprehensively obscuring litigant traits that are legally irrelevant (or even legally relevant when they are likely to bias or confuse decision-making),Footnote ⁷⁴ whenever it is consistent with fair process.

Third, lawyers provide a “translation” function to clients. Lawyers distill what they learn from clients into core substantive facts, and then assemble and present those facts in the highly reticulated way that legal actors expect from legal professionals.Footnote ⁷⁵ This professional repackaging may augment a litigant’s story and substantive arguments by making them more palatable, reliable, and understandable to a judge.Footnote ⁷⁶ Or it may just be a key into the club. If there are lawyers on both sides, any imprimatur may cancel out, yet the dynamics of the prisoner’s dilemma apply, so all who can “lawyer up” still do so. Pro se litigants are left behind. Even if translation and delivery by an expert improves the litigant’s prospects (relative to a counterparty), it remains an open question whether this function enhances the accuracy or efficiency of any adjudication. Regardless, if only one party has access to an attorney’s stamp of approval or can speak “legalese,” a pro se gap might develop.Footnote ⁷⁷

Enhanced ODR can diminish the role that legalese and professional status might play in litigation by reducing their use, making them ambiguous or less salient, or facilitating a pro se litigant’s ability to employ them within the platform. ODR design can encourage (or require) litigants to use plain language, either via the structure of its data collection strategies (e.g., forms) or through ex post screening of user language. Alternatively, one can imagine a more involved ODR process that educates litigants about the meanings of legal words or legal communication norms or that offers something closer to a translation service, much the way smart form procedures operate today, by giving people choices. Either way, ODR platform “rigidity” in the allowable scope and style of communication can be a disparity-reducing positive;Footnote ⁷⁸ such structure can ensure that parties speak the same language while also giving them equal opportunities to be heard.

Importantly, reducing “degrees of freedom” in ODR communications can also close a back-door source of implicit bias. An ODR system that scrubs case materials of identity-trait data after having encouraged a litigant to submit an open-ended statement may be closing the barn door after the horse has bolted. Intentionally or unintentionally, litigants reveal personal traits in their communications:Footnote ⁷⁹ they explicitly indicate their race, gender, occupation, or age,Footnote ⁸⁰ or they implicitly reveal these traits through language that correlates with demographic characteristics.Footnote ⁸¹ In any event, to the extent that legal representation liberates litigants from the need to speak legalese in the courtroom and insulates litigants from implicit bias by insulating judges from triggering face-to-face interactions, ODR systems have the potential to do much the same through careful design and creative solutions.