An AI-Based Future: Diverse, Unbiased and Just?

An AI-Based Future: Diverse, Unbiased and Just?

[This piece has been authored by Parul Anand, a student at the National Law University, Jodhpur.]

The fourth industrial revolution has long begun, and the legal profession is no stranger to it. Artificial intelligence is here to stay. Often referred to as ‘disruptive technology’, could these tools make the legal system more diverse, and by extension more just? The introduction of artificial intelligence tools in the legal profession could potentially mitigate bias, increase access to justice and promote diversity. This piece seeks to illustrate the same with contemporary applications and potential actualities.

With pending cases touching 4,04,56,940– the Indian legal apparatus is undeniably struggling with providing access to justice. That gatekeeping exists in the legal profession, making it one of the least diverse, is not news.  As one moves up the tiers of the lower judiciary, a near-uniform trend is noticeable in the proportion of women judges. Furthermore, being humans, legal practitioners are not untainted by bias. Research from Danziger et al. (2011)  demonstrated how legally irrelevant situational determinants, such as taking a food break, may lead a judge to rule differently in cases with similar legal characteristics.

It is in this context that the introduction of AI becomes imperative as the current judicial system can do with all the feasible help that it can obtain. AI makes justice more reachable with its supplemental digital advantage, precise algorithmic approach and de facto 24/7 availability. In addition, most significantly— it allays bias by indicating inconsistencies in the judicial decision-making process, divulging biases, and substituting possible bias inducing indicators with neutral alternatives. It also helps in making the process of justice truly blind.

Using AI for translating documents like FIRs into regional languages by ICJS can improve the accessibility of law, mitigate bias against certain linguistic groups, and move us towards a more democratic society. Justice D.Y. Chandrachud sustains this prospect by remarking how “[the] language of the courts [would now be] available at the doorsteps of the litigant”. To make justice truly more accessible, an AI-based system can be employed to handle traffic challans and other summary violations by automating tasks such as sending notices and monitoring feedback. To assist its application, the extant eCourts project and digital judicial database can be synchronized with it.

It can also be constructively employed to answer legal questions where a straightforwardly algorithmic approach is required. exemplifies one such initiative providing critical information about wages, layoff, and termination compensation amongst other things in an accessible simplified manner. Another model is UK-based DoNotPay which offers services ranging from challenging parking tickets to helping with asylum applications, maternity leave, checking on landlord contract violations and claims against Equifax. Both cases illustrate how the use of AI has made justice substantially more accessible. Encouragingly, DoNotPay has a success rate of 64% and has won 1,60,000 cases so far.

Markedly, AI can also help counter the ever-mounting list of judicial vacancies. Former Supreme Court Judge BN Srikrishna in an article on LiveMint suggests “draw[ing] up a reserve list of judicial officers that can be kept updated so that proper person can be identified and promptly placed in [an] appropriate vacancy without [the] loss of time’ with the help of artificial intelligence systems.” Such an apparatus can model the appointment process as more objective, consequently opening channels for a more diverse legal force.

Solely run on electricity, the computing machine need not ever recline as it winnows through details from intricate cases with dozens or hundreds of volumes. A legal AI currently thought  of being implemented, with a speed of about ten lakh words per second, would have been immensely advantageous in a documentation-heavy case like the Ayodhya dispute. In research from LawGeex (2018), AI was not only more exacting but also significantly diminished the time taken to scrutinize a standard business contract to identify potential issues by approximately 200%. This demonstrates the optimistic possibility that besides providing increased productivity, AI could free up swathes of time for lawyers and enable them to undertake a larger amount of pro-bono cases. It would also lower the gross time taken for a legal proceeding, and aid in actively spinning the wheels of justice, and more crucially, arousing the judiciary from its contemporary slowdown.

Furthermore, a Strategy paper by Vidhi Legal also recommends AI interventions “at the level of smart e-filing, intelligent filtering/prioritization of cases or notifications and tracking of cases” which could further streamline judicial administrative functions. (Jauhar et al., 2021).  In addition to that, AI filters can also be used to isolate documentation sent to the court which does not comply with the appropriate format or requires more information—reducing time spent on trivial jobs. Articulated prediction of court outcomes using Predictive AI in civil lawsuits can preserve litigants’ resources and time, thus making justice timelier and supplanting the already inundated judiciary.

Possibly the most significant area where AI can display its capacity and potential is in mitigation of bias. Machine learning algorithms discover inconsistencies that escape the trained eye with relative ease and speed, this “dynamically analysed data could call into question whether certain legal outcomes were driven by factors different from those that were expressed in the language of an opinion.”  A study by Theodore W. Ruger et al. (2004) illustrated precisely the same phenomenon while predicting Supreme Court judgments more accurately and also revealing previously hidden data. In such an arrangement, AI can proffer an added layer of reflection which could make the system more detached and equitable. Additionally, with AI, cherry-picking of data to arrive at a specific conclusion is not feasible; it cannot help but examine whole data. Therefore, through the merit of its manner, it can effortlessly point out discrepancies in multiple legal decisions where the conclusion appears out of reach from the data and metrics to be recognised and where with all the supplied data, the conclusion to be reached, has not been reached.

AI can also recognise and decrease bias in “client intake and initial consultations, it can assess the uniformity of criminal charging decisions made by prosecutors, and it can help to diversify law firm ranks, judicial ranks, and even juror pools.” In addition, it can eliminate the traits that lead to prejudices and can learn how to recognise likely bents.

It can further be “train[ed]” to be more inclusive of “other races and genders, despite statistics  to the contrary in the data set” which proceeds further from sole alleviation of bias to the active introduction of diversity. Additionally, AI systems can support the Chief Justice in allocating cases as it can promptly examine multiple variables like existing cases a judge holds, concerning dates, natural justice, specialisation, and availability. AI can time and again expose discrepancies, point out biases, automate mundane algorithmic decision-making and go over large swathes of data, whereas humans can employ the values of justice, equity, clemency, compassion, and weigh factors that the AI cannot. The judicious corollary would be to have both operating in tandem. Professionals can administer AI with continual feedback and communication, making it extensively more suitable. With a few basic caveats, the introduction of AI in the legal context and the unification of AI with legal practitioners can undoubtedly transform the legal system and catalyse its much-needed progression towards diversity, impartiality and accessibility.

Leave a Reply

Your email address will not be published. Required fields are marked *