Making intelligent online dispute resolution tools available to self-represented litigants in the public justice system: towards and ethical use of the AI technology in the administration of justice

Making In elligen Online Dispu e Resolu ion Tools a ailable o Sel -Rep esen ed Li igan s in he Public Jus ice Sys em∗ Towa ds and E hical use o he AI echnology in he adminis a ion o Jus ice Fe nando Es eban de la Rosa† Uni e si y o G anada G anada, Spain es eban@ug .es John Zeleznikow† La T obe Uni e si y Law School Melbou ne, Aus alia [email p o ec ed] ABSTRACT O e he las decade online dispu e esolu ion (ODR) has mo ed om me ely e-comme ce li iga ion o widesp ead use in cou sys- ems. Two phenomena ha e led o his si ua ion: he ise o Sel - Rep esen ed Li igan s and Cou s mo ing beyond hei adi ional ocus, allowing pa ies, o ins ance, o ile a claim, o mula e hei a gumen s, ob ain legal in o ma ion o e en a ecei e a o ecas abou he esolu ion o he case. AI ools ha e mainly been used o enable legal p o essionals (lawye s, media o s) o be e pe o m hei asks. Today some ju isdic ions ha e begun o p o ide jus ice use s wi h uly use ul in elligen - use cen ic ODR sys ems in- co po a ing assessmen and diagnosis AI ools. These ools may p o ide in o ma ion abou a possible ou come. This pape analy- ses he use being made by some ju isdic ions o combined Online Dispu e Resolu ion and A i icial In elligence ools and aims o p omo e he deba e on he e hical go e nance o making hese ools a ailable o un ep esen ed li igan s. The e alua ion ollows a Eu opean pe spec i e on he e hical go e nance o he use o AI in he Jus ice Sys em. KEYWORDS Jus ice Sys ems, In elligen -use cen ic Online Dispu e Resolu ion Sys ems, Sel -Rep esen ed Li igan s, E hics ACM Re e ence Fo ma : Fe nando Es eban de la Rosa and John Zeleznikow. 2021. Making In elligen Online Dispu e Resolu ion Tools a ailable o Sel -Rep esen ed Li igan s in he Public Jus ice Sys em: Towa ds and E hical use o he AI echnology in he adminis a ion o Jus ice. In Eigh een h In e na ional Con e ence o A i icial In elligence and Law (ICAIL’21), June 21–25, 2021, São Paulo, B azil. ACM, New Yo k, NY, USA, 5 pages. h ps://doi.o g/10.1145/3462757.3466077 ∗ Lodde and Zeleznikow [8] indica e, whils he e is no gene ally accep ed de ini ion o ODR, we can hink o i as using he In e ne o pe o m Al e na i e Dispu e Resolu ion (ADR). †Bo h au ho s con ibu ed equally o his esea ch. Pe mission o make digi al o ha d copies o all o pa o his wo k o pe sonal o class oom use is g an ed wi hou ee p o ided ha copies a e no made o dis ibu ed o p o i o comme cial ad an age and ha copies bea his no ice and he ull ci a ion on he i s page. Copy igh s o componen s o his wo k owned by o he s han ACM mus be hono ed. Abs ac ing wi h c edi is pe mi ed. To copy o he wise, o epublish, o pos on se e s o o edis ibu e o lis s, equi es p io speci ic pe mission and/o a ee. Reques pe missions om [email p o ec ed]. ICAIL’21, June 21–25, 2021, São Paulo, B azil ©2021 Associa ion o Compu ing Machine y. ACM ISBN 978-1-4503-8526-8/21/06...$15.00 h ps://doi.o g/10.1145/3462757.3466077 1 INTRODUCTION: THE NEW COMBINED USE OF ONLINE DISPUTE RESOLUTION AND ARTIFICIAL INTELLIGENCE TOOLS IN THE PUBLIC JUSTICE SYSTEM Ad ances in Compu ing and In o ma ion and Communica ion Tech- nology ha e opened new possibili ies o implemen ing adi ional models o jus ice sys ems. The de elopmen o he COVID19 pan- demic has u he enhanced he ise o Online Dispu e Resolu ion (ODR) and led o he inco po a ion o a wide ange o echnological ools in o dispu e esolu ion. De elopmen s in A i icial In elligence (AI) hold p omises o imp o ing e iciency and quali y in he p o- ision o access o jus ice, leading o imp o ed anspa ency and s anda disa ion o case-law. Tania Sou din [ 12 ] has sugges ed ha he e a e h ee p ima y ways in which echnology has al eady es uc u ed he jus ice sys em. Fi s , and a he mos udimen a y le el, a e “suppo i e” echnologies – hese echnologies aim o in o m, suppo and ad- ise indi iduals in ol ed in he jus ice sys em and include, o example, online legal applica ions (apps). A he second le el a e “ eplacemen ” echnologies – hese echnologies eplace he oles and ac i i ies adi ionally conduc ed by humans and include, in e alia, e- iling p ocesses and online media ion se ices. Finally, and a he mos ad anced le el, a e “dis up i e” echnologies – hese ech- nologies undamen ally al e he way in which legal p o essionals wo k and include, o example, AI judges o o he algo i hm-based decision-making p og ams ha may eshape he judicial ole. Whils he e a e nume ous e-cou s ( o ins ance in he Uni ed Kingdom, USA and Aus alia), no many ju isdic ions ha e aken he s ep o wa d o p oposing a combined use o ODR ools designed o p o ide in o ma ion in such a way ha he pa ies may ob ain a diagnosis o a p edic ion o he ou come so ha dispu an s can be be e p epa ed o deal wi h a di ec nego ia ions conduc ed online. Mani es a ions o his combina ion o ools a e ound oday in he Du ch Pla o m Rech wijze , he new In e ne Cou s in China and he Ci il Resolu ion T ibunal (CRT) in B i ish Columbia (Canada). The e a e also p ojec s wi h he same aim in Singapo e and in Es onia, in he la e case his includes he c ea ion o a “ obo judge”. This pape analyses he combined use o ODR and AI ools in cou s, wi h a special ocus on cases whe e AI ools aim o suppo Sel Rep esen ed Li igan s (SRL)s in eaching se lemen s. This combina ion o ools makes use o he no ion o Ba gaining in he Shadow o he Law de eloped by Mnookin and Ko nhause [9] . The 195 ICAIL’21, June 21–25, 2021, São Paulo, B azil F. Es eban de la Rosa and J. Zeleznikow pape desc ibes he unde lying easons ha ha e igge ed he ise o SRLs and includes p oposals o ensu e hei ai ea men . 2 EXAMPLES INCORPORATING ARTIFICIAL INTELLIGENCE INTO ONLINE DISPUTE RESOLUTION SYSTEMS 2.1 The Du ch pla o m Rech wijze Rech wijze 1(Roadmap o Jus ice) was designed o couples, wi h child en who a e sepa a ing. The aim o Rech wijze was ‘ o em- powe ci izens o sol e hei p oblems by hemsel es o oge he wi h his o he pa ne . I necessa y, i e e s people o he assis ance o expe s.’ Couples pay €100 o access o Rech wijze , which s a s by asking each pa ne o in o ma ion such as hei age, income, educa ion, whe he hey wan he child en o li e wi h only one pa en o pa ime wi h each, hen guides hem h ough ques ions abou hei p e e ences. The pla o m had a diagnosis phase; an in ake phase o he ini ia ing pa y; and hen in i ed he o he o join and unde ake he same in ake p ocess. Once in ake is comple ed, he pa ies s a wo king on ag eemen s. The dispu e esolu ion model is ha o in eg a i e (p incipled) nego ia ion[ 5 ]. The pa ies a e in o med o ules such as hose o di iding p ope y, child suppo and s an- da d a angemen s o isi ing igh s so ha hey could ag ee on he basis o in o med consen . Ag eed ag eemen s a e e iewed by a neu al lawye . I he p oposed solu ions a e no accep ed, hen couples can employ he sys em o eques a media o o an addi- ional €360, o a binding decision by an adjudica o . Rech wijze is olun a y and non-binding up un il he poin whe e he pa ies seek adjudica ion. Rech wijze had aimed o be sel - inancing h ough use con ibu ions. This has no occu ed. 2.2 The B i ish Columbia Ci il Resolu ion T ibunal The B i ish Columbia Ci il Resolu ion T ibunal [ 11 ] is he mos signi ican cu en widely a ailable ODR sys em ha comes closes o p o iding a ull sui e o dispu e esolu ion se ices. I commences by diagnosing he dispu e, by p o iding a decision ee, and p o ides legal in o ma ion and ools such as cus omized le e empla es. I his ac ion does no esol e he dispu e, one can hen apply o he Ci il Resolu ion T ibunal o dispu e esolu ion. The sys- em di ec s he use o he app op ia e applica ion o ms. Once he applica ion is accep ed, he use en e s a secu e and con iden ial nego ia ion pla o m, whe e he dispu an s can a emp o esol e hei dispu e. I he pa ies canno esol e he dispu e, a acili a o will assis . Ag eemen s can be u ned in o en o ceable o de s. I nego ia ion o acili a ion does no lead o a esolu ion, an indepen- den membe will make a de e mina ion abou he dispu e. Cu en ly, he Ci il Resolu ion T ibunal deals wi h Mo o ehi- cle inju y dispu es, Small claims dispu es, S a a p ope y dispu es, Socie ies and coope a i e associa ions dispu es and Sha ed accom- moda ion and some housing dispu es. Fo some o hese domains po en ial li igan s can only use he Ci il Resolu ion T ibunal. To assis digi ally disad an aged li igan s, echnical suppo is p o ided in accessing he In e ne . One o he majo easons ha 1h ps:// ech wijze .nl/ las iewed 5 Feb ua y 2021. he Ci il Resolu ion T ibunal has been so success ul, is ha B i ish Columbia esiden s a e manda ed o use he sys em when dealing wi h issues lis ed abo e. Whils such an app oach may be seen no el and disc imina o y, i does ensu e ha he sys em is used, wi h ela i e ease, quickly and a minimal cos . In mos cases pa ies a e o ep esen hemsel es, e en i ep esen a ion and legal assis ance is allowed. 2.3 The In e ne Cou s in China Be ween 2017 and 2018 China c ea ed h ee new Cou s: he Hangzhou In e ne Cou , he Beijing In e ne Cou and he Guangzhou In- e ne Cou . These cou s only ha e ma e ial ju isdic ion o e in e ne - ela ed cases. The online pla o m makes an in elligen li iga ion isk assessmen sys em a ailable o he use and can p o- ide a epo syn hesising he li igan s’ case and he co esponding isk based on he analysis o cou da a and simila cases. Li iga ion isk assessmen aims o help he pa y wi hou legal knowledge o iden i y and exclude common li iga ion isks, he eby educ- ing unnecessa y losses. Meanwhile, he assessmen can make he pa y awa e ha li iga ion is isky and cos ly and guide he pa ies o choose ADR o di e si ied dispu e esolu ion. The sys em can au oma ically gene a e a complain le e by simply selec ing he sui able esponse op ions. [4] 2.4 P ojec s in Es onia and Singapo e In July 2019 he Es onian Minis y o Jus ice launched a p ojec de eloping AI so wa e o hea and esol e small economic dis- pu es by elimina ing human in e en ion [ 10 ]. The “ obo judge” is con igu ed o decide dispu es o up o 7,000 eu os. Acco ding o he p ojec , he dispu ing pa ies would ha e o upload hei documen s and ele an in o ma ion o a judicial pla o m. The AI machine ende s a decision ha can be appealed o a human judge. The p ojec limi s i s scope o con ac ual dispu es. Singapo e has been commi ed o digi al jus ice since 2000. In ecen yea s i has been de eloping a mo e ambi ious online sys em, ini ially only o inju ies a ising om mo o ehicle acciden s. An ou come simula o will p o ide guidance o po en ial claiman s, p io o he commencemen o p oceedings, helping hem o de- cide on o e s om insu ance companies. The aim is o pa ies o i s use he echnology o each amicable se lemen s wi hou p o essional legal ad iso s [14]. 3 A FRAMEWORK FOR BUILDING ONLINE DISPUTE RESOLUTION TOOLS FOR SELF-REPRESENTED LITIGANTS An inc easing phenomenon in Common Law coun ies is he g ow- ing numbe o p o se (o sel - ep esen ed) li igan s. Landsman [7] a gues ha p o se cases pose inhe en p oblems: hey can cause de- lays, inc ease adminis a i e cos s, unde mine he judges’ abili y o main ain impa iali y and can lea e he o en-unsuccess ul li igan eeling as hough she has been ea ed un ai ly. Resea ch conduc ed in he Family Cou o Aus alia shows ha he e a e a ange o easons why people ep esen hemsel es, such as unding cu s and changes in eligibili y o legal aid [ 3 ]. O he con ibu ing ac o s include changes in echnology, cul u al shi s owa ds sel -help and sel - ep esen a ion, and changes in 196 Making In elligen Online Dispu e Resolu ion Tools a ailable o Sel -Rep esen ed Li igan s in he Public Jus ice Sys em ICAIL’21, June 21–25, 2021, São Paulo, B azil legisla ion. The expe iences o sel - ep esen a ion in Aus alian law has gene ally accep ed ha SRLs a e a a disad an age in legal p oceedings and hei expe ience o he legal sys em may indeed be nega i e. The lack o knowledge o skills o SRLs means ha some a e no able o access ai and equal jus ice in a sys em o en gea ed owa ds legal ep esen a ion. In England li igan s can go o cou wi hou legal aid, in p ac ice he echnical and o mal na u e o p oceedings, wi h he excep ion o he small claims p ocedu e ( o claims up o £10,000), makes legal aid necessa y. I s lack has led o public dissa is ac ion bu also us a ion among judges, aced wi h he need o in o m lay li igan s abou he echnicali ies o he p ocess wi hou being able o c oss he line be ween in o ma ion and legal aid. This si ua ion led o a conside able inc ease in he ime and cos spen on each judicial decision, e en doubling i 2 . While some SRLs can p esen hei case compe en ly, mos esea ch sugges s ha SRLs s uggle wi h subs an i e law and p ocedu e [6]. Recen expe iences, such as he online cou es ablished in U ah, a e demons a ing ha ODR has he po en ial o ans o m he way he Ame ican legal sys em deals wi h p o se li igan s and access o jus ice issues a la ge. Al hough i may seem coun e in ui i e o b idge he jus ice gap by p ecluding people om appea ing in cou , equi ing ce ain ypes o claims o begin online will ac ually p o ide quicke and mo e accessible legal solu ions. As long as he p og amming and adminis a ion o ODR echnology a e conduc ed wi h a en ion o legal and e hical conce ns, p o se li igan s will bene i om ha ing hei claims esol ed online [ 2 ]. Fo his aim access o jus ice is helped by he use o in elligen -use cen ic ODR sys ems inco po a ing assessmen and diagnosis AI ools [15]. S anie i e al . [13] app oach o p o iding ad ice abou he dis- ibu ion o ma i al p ope y ollowing di o ce in Aus alia was o use machine lea ning o p o ide ad ice abou BATNAs (a BATNA is used o in o m dispu an s o he likely ou come i he dispu e we e o be decided by decision-make e.g. judge, a bi a o o ombuds- man). Despi e using Machine Lea ning, i in ol ed he de elopmen o 94 Toulmin a gumen s uc u es [ 16 ] o model he domain as i exis ed in 1995. Twen y- i e yea s la e , he heo e ical p inciples behind machine lea ning so wa e ha e no changed. Bu compu e ha dwa e is now much cheape and da a can be much mo e easily s o ed. This has led o he de elopmen o ‘quicke ; sys ems’, which he communi y has seen as ‘mo e in elligen ’3. Whils he Spli -Up sys em p o ides ad ice abou BATNAs, he Family Winne Sys em [ 1 ] p o ided ad ice o dispu ing pa en s on how hey could bes nego ia e ade-o s. The dispu ing pa ies we e asked o indica e how much hey alued each i em in dispu e. Using log olling, pa ies ob ained wha hey mos desi ed. Zeleznikow [20] discusses how i is possible o build ODR sys- ems ha suppo sel - ep esen ed li igan s and wha skills do sel - ep esen ed li igan s equi e o use such sys ems. Zeleznikow [21] conside s how we can cons uc such sys ems wi h use cen ic compu ing. So, wha a e he a ious ypes o ODR sys ems and how can sel - ep esen ed li igan s use hem? Ha ing ega d o he 2 JUSTICE, “Deli e ing Jus ice in an Age o Aus e i y” (Ap il 2015). A ailable in h ps: //jus ice.o g.uk/jus ice-age-aus e i y-2/ las iewed 19 Ap il 2021 3 See o example amica.go .au which uses machine lea ning o ad ise upon p ope y dis ibu ion amongs sepa a ing couples in Aus alia. ulne able posi ion o he un ep esen ed li igan a uly help ul ODR sys em should p o ide he ollowing acili ies: (1) Case managemen : he sys em should allow use s o en e in o ma ion, ask hem o app op ia e da a and p o ide o empla es o ini ia e he dispu e. Sel - ep esen ed li igan s should be able o ini ia e he dispu e, en e hei pe inen da a and also ack wha is happening du ing he dispu e as well as being awa e o wha documen s a e equi ed a speci ic imes; (2) T iaging: he sys em should p o ide in o ma ion on how impo an i is o ac in a imely manne and whe e o send he dispu e. This may be pa icula ly impo an in cases o domes ic abuse o whe e he e is a po en ial o child en o be kidnapped. T iaging sys ems a e i al o expedi ing ac ion in high isk cases; (3) Ad iso y ools: he sys em should p o ide ools o eali y es ing: hese could include, books, a icles, epo s o cases, copies o legisla ion and ideos; he e would also be calcu- la o s (such as o ad ise upon child suppo ) and BATNA ad iso y; sys ems ( o in o m dispu an s o he likely ou - come i he dispu e we e o be decided by decision-make (e.g. judge, a bi a o o ombudsman). Ad iso y ools, as sugges ed by Zeleznikow [19] a e a i al cog in suppo ing sel - ep esen ed li igan s. An impo an associa ed ques ion is how can we design ad iso y ools ha sel - ep esen ed li igan s can gain ully use? A e he legal concep s behind hese ools oo di icul o ama eu s o unde s and? How do we cons uc sui able use in e aces? (4) Communica ion ools – o nego ia ion, media ion, con- cilia ion o acili a ion. This could in ol e shu le media ion i equi ed. Fo many ODR p o ide s, he p o ision o com- munica ion ools is hei main goal; (5) Decision Suppo Tools – i he dispu an s canno esol e hei con lic , so wa e using game heo y o AI can be used o acili a e ade-o s. P o essionals (such as lawye s) can p o ide use ul ad ice e ade-o s. In hei absence, sui able decision suppo ools a e i al; (6) D a ing so wa e: i and once a nego ia ion is eached, so wa e can be used o d a sui able ag eemen s. D a ing plans (such as pa en ing plans) once he e is an in-p inciple ag eemen o a esolu ion o a dispu e, is a non- i ial ask. No single dispu e is likely o equi e all six p ocesses. Howe e , he de elopmen o such a hyb id ODR sys em would be e y signi i- can . A o al sys em would equi e us o cons uc he app op ia e sys ems 1 o 6, and he ul ima e solu ion is o make su e ha all he sys ems a e capable o communica ing wi h each o he . 4 ETHICAL ISSUES RELATING TO THE PROVISION OF ARTIFICIAL INTELLIGENCE-BASED TOOLS TO SELF-REPRESENTED LITIGANTS BY THE PUBLIC JUSTICE SYSTEM: A EUROPEAN PERSPECTIVE Nei he he ecen o icial documen s o he Eu opean Union de e - mining how AI should be used in he ield o he adminis a ion o 197 ICAIL’21, June 21–25, 2021, São Paulo, B azil F. Es eban de la Rosa and J. Zeleznikow jus ice 4 no he Eu opean E hical Cha e (EEC) on he use o AI in judicial Sys ems and hei en i onmen adop ed in 2018 by he Eu opean Commission o he E iciency o Jus ice o he Council o Eu ope deal di ec ly wi h he admission o AI ools aimed a enabling he pa ies o assess hei legal posi ion. Because SRLs gene ally lack legal skills and in iew o he objec i e o encou - age nego ia ion we submi ha his use o echnology o hese pu poses should be conside ed high- isk. The EEC poin s ou he inhe en isks in hese echnologies may e en anscend he ac o judging and a ec essen ial unc ioning elemen s o he ule o law and judicial sys ems. These include p inciples such as he p imacy o law. These ools could c ea e a new o m o no ma i i y, which could supplemen he law by egula ing he so e eign disc e ion o he judge, and po en ially leading, in he long e m, o a s anda disa ion o judicial decision based no longe on case-by-case easoning by he cou s, bu on a pu e s a is ical calcula ion linked o he a e age compensa ion p e iously awa ded by o he cou s. Tha is why he epo submi s a need o conside whe he hese solu ions a e compa ible wi h he indi idual igh s ensh ined in he Eu opean Con en ion on Human Righ s (ECHR). These would include he igh s o a ai ial (pa icula ly he igh o a na u al judge es ablished by law, he igh o an independen and impa ial ibunal and equali y o a ms in judicial p oceedings) and, whe e insu icien ca e has been aken o p o ec da a communica ed in open da a, he igh o espec o p i a e and amily li e. Thus he EEC conside s ha applica ions o p edic i e jus ice should be assigned o he ield o esea ch and u he de elopmen in o de o ensu e ha hey ully ie in wi h ac ual needs be o e con empla ing use on a signi ican scale in he public sphe e. The Eu opean Commission (EC) ecognises ha he use o AI applica ions can b ing many bene i s, such as making use o in o - ma ion in new and highly e icien ways, and imp o e access o jus ice, including by educing he du a ion o judicial p oceedings. A he same ime i is awa e ha he opaci y o biases embedded in ce ain AI applica ions can also lead o isks and challenges o he espec o and e ec i e en o cemen o undamen al igh s, includ- ing in pa icula he igh o an e ec i e emedy and a ai ial. The EC ecognises as a possible high- isk a use case using he ech- nology as pa o decision-making p ocesses wi h signi ican e ec s on he igh s o people. Howe e , i also conside s ha he p o- posed equi emen s in he Whi e Pape on inc eased anspa ency, human o e sigh , accu acy and obus ness o hese sys ems aim o acili a e hei bene icial use, while ensu ing ha undamen al igh s including non-disc imina ion based on sex, acial o e hnic 4 Among he las o icial documen s a e he P oposal o a Regula ion Laying down Ha monised Rules on A i icial In elligence (A i icial In elligence Ac ) and Amending Ce ain Union Legisla i e Ac o 21.4.2021 COM (2021) 206 inal; he Communica ion om he Commission o he Eu opean Pa liamen , he Council, he Eu opean Economic and Social Commi ee and he Commi ee o he egions called “Digi alisa ion o jus ice in he Eu opean Union. A oolbox o oppo uni ies”, COM (2020) 710 inal, o 2.12.2020; he Eu opean Pa liamen Resolu ion o 20 Oc obe 2020 wi h ecommenda ions o he Commission on a amewo k o e hical aspec s o a i icial in elligence, obo ics and ela ed echnologies (2020/2012 (INL); Whi e Pape on A i icial In elligence - A Eu opean app oach o excellence and us , COM(2020) 65 inal o B ussels, 19.2.2020; he Eu opean e-Jus ice S a egy 2019-2023 o 13 Ma ch 2019 (2019/C 96/04) Council 2019-2023; he Digi al Re olu ion in iew o Ci izens’ Needs and Righ s. Opinion o he Eu opean Economic and Social Commi ee o 20.02.2019. o igin, eligion o belie , disabili y, age o sexual o ien a ion) a e espec ed and ule o law and due p ocess p inciples upheld. In o de o unde s and he Eu opean posi ion i is also ele an o know he c i e ion ollowed by he new p oposal o a Regula ion o Ap il 2021. AI sys ems in ended o he adminis a ion o jus ice a e no lis ed among he p ohibi ed p ac ices (a . 5) bu among he high- isk AI sys ems (poin 40 o he p eamble). The new p oposal o a Regula ion sepa a es wo kinds o judicial ac i i ies: i is conside ed as high- isk he sys ems in ended o assis judicial au ho i ies in esea ching and in e p e ing ac s and he law and in applying he law o a conc e e se o ac s. Such quali ica ion is no ex ended o AI sys ems in ended o pu ely ancilla y adminis a i e ac i i ies ha do no a ec he ac ual adminis a ion o jus ice in indi idual cases. The p oposed Regula ion does no es ablish he de ini i e answe as any use o AI mus con inue o occu solely in acco dance wi h he applicable equi emen s esul ing om he Eu opean Cha e o Fundamen al Righ s, he es o Eu opean Law and he na ional law. We submi ha in iew o he bene icial impac i may ha e on he unc ioning o he judicial sys em, i is necessa y o iden i y he eal possibili ies, echnical limi s and sa egua ds o be me by he machines o e ed by he public jus ice sys em o SRLs. Fo speci ic a eas o adminis a i e law i is possible o de elop legal ules as code p o iding use ul in o ma ion and suppo o SRLs. The use o code as ules in combina ion wi h Use Cen ic ODR Tools using decision ees, may ha e success p omo ing ac- cess o jus ice o SRLs. The CRT in he B i ish Columbia is an example o success. The design o AI ule-based sys ems does no exhibi he di icul ies a ising om he lack o anspa ency and he c ea ion o biases ha may a ise employing ML induc ion al- go i hms. Deduc i e AI ools ( he so called Expe s Sys ems) allow anspa ency and he moni o ing o he machine ou pu is acili- a ed o be able o ec i y wha is necessa y in case any e o s in he p og amming a e disco e ed. P og amming is, howe e , a delica e p ocess and i no done well can lead o un ai ea men when he algo i hm doesn’ ma ch eali y. This can occu when a one-size- i s-all ule is implemen ed in a complex en i onmen . A ecen example is Aus alia’s Cen elink “ obodeb ” debacle 5 . In ha case, wel a e paymen s made on he basis o sel - epo ed o nigh ly income we e c oss- e e enced agains an es ima ed o nigh ly in- come, aken as a simple a e age o annual ea nings epo ed o he Aus alian Tax O ice, and used o au o-gene a e deb no ices wi hou any u he human sc u iny o explana ion. This assump- ion is a odds wi h how Aus alia’s highly casualised wo k o ce is ac ually paid. Fo example, a g aphic designe who was unable o ind wo k o nine mon hs o he inancial yea bu ea ned A$12,000 in he h ee mon hs be o e June would ha e had an au oma ed deb aised agains he . This is despi e no aud ha ing occu ed, and his scena io cons i u ing exac ly he kind o ha dship Cen elink is designed o add ess. Rules as codes equi es al e a ions o be in oduced in case o legisla i e changes. Al hough i will no be possible o a ain he quali y o ad ice o e ed by a legal expe , we submi ha he in o - ma ion p o ided o SRLs h ough machines makes a con ibu ion o 5See h ps:// inyu l.com/y3dqe6mg las iewed 19 Ap il 2019 198 Making In elligen Online Dispu e Resolu ion Tools a ailable o Sel -Rep esen ed Li igan s in he Public Jus ice Sys em ICAIL’21, June 21–25, 2021, São Paulo, B azil imp o ing access o jus ice o hose who canno a o d legal assis- ance. Rega ding he quali y o ad ice p o ided by hese machines, i seems easonable ha he p oposals o he Eu opean Commission abou equi emen s conce ning possible es ing o applica ions and he need o p o ide ele an documen a ion on hei pu poses and unc ionali ies. I seems also easonable o equi e main aining he possibili y o co ec e o s and p o iding in o ma ion o he use ha he answe gi en by he machine may no necessa ily ma ch he answe ha would be gi en by a judge hea ing he case. Two o he disad an ages o he use o Machine Lea ning sys ems a e ha hey a e no anspa en , and he da a and he so wa e on which hey a e based may be manipula ed. The e is also a conce n ha he use o Machine Lea ning in he legal sys em will wo sen biases agains mino i ies o deepen he di ide be ween hose who can a o d quali y legal assis ance and hose who canno [ 17 ]. Algo- i hms will con inue o pe o m exis ing biases agains ulne able g oups because he algo i hms a e la gely copying and ampli ying he decision-making ends embedded in he legal sys em. The e is al eady a class di ide in legal access – hose who can a o d high quali y legal p o essionals will always ha e an ad an age. The de- elopmen o in elligen suppo sys ems can pa ially ed ess his powe imbalance by p o iding use s wi h impo an legal in o - ma ion ha was p e iously una ailable o hem. Di icul ies may s em om biases. One example is COMPAS, a decision suppo sys em designed o help pa ole boa ds in he Uni ed S a es [ 18 ] decide which p isone s o elease ea ly, by p o iding a p obabili y sco e o hei likelihood o eo ending. Ra he han ely on a simple decision ule, he algo i hm used a ange o inpu s, including demo- g aphic and su ey in o ma ion, o de i e a sco e. The algo i hm did no use ace as an explici a iable, bu i did embed sys emic acism by using a iables ha we e shaped by police and judicial biases. Wha can be done is o ensu e he aceabili y and cleanliness o he da a wi h which he machine ope a es, and o in oduce elemen s o weigh ing. Bu as Richa d Susskind conside s, an e hical p og amming is no easible. I is no a all clea , ei he echnically o philosophically, wha is mean when i is p oposed ha e hics should be embedded in Machine Lea ning. No i is clea wha is mean when i is demanded ha so wa e enginee s p og am Machine Lea ning sys ems o p o ide in elligen explana ions. To hink so is o misunde s and he di e ence be ween he induc i e p ocesses inhe en in Machine Lea ning and he kind o a gumen we expec when we ask o an explana ion [14]. A di e en issue is he use o AI ools by judges o decide a case. We sha e he Eu opean Commission’s iew ha i is impo an ha judgmen s a e deli e ed by judges who ully unde s and he AI applica ions and all in o ma ion aken in o accoun he ein ha hey migh use in hei wo k (AI no o eplace bu as Augmen ed In elligence), on he unde s anding ha he use o AI applica ions mus no p e en any public body om gi ing explana ions o i s decisions. As o he machine being able o decide he case on i s own, as he Es onian p ojec poses, his should no be comple ely uled ou . Howe e , we a e no a ha s age ye ! In he cu en s a e o he a , machines can nei he mo i a e no explain he decisions and p edic ions hey make [ 14 ]. Legal a gumen s equi e pe suasion ha does no depend on p edic able a iables. 5 CONCLUSION One o he la es ends in he inco po a ion o echnology in he adminis a ion o jus ice is he p o ision by public jus ice sys ems o suppo SRLs by he use o a combina ion o AI and ODR ools. These allow SRLS o ha e a diagnosis o he case, which in luences he pa ies ei he o de e mine a dismissal o he ac ion o how o nego ia e. This combina ion o ools shows g ea po en ial in educing he le el and du a ion o li iga ion. The pape submi s ha his use o he echnology mus be conside ed as high isk as i may unc ion as a eplacemen o judicial ac i i ies. Howe e , i is s ill possible o ob ain posi i e esul s om his echnology by inse ing some sa egua ds, as is beginning o eme ge om he Eu opean legal sphe e. The deba e is now abou wha sa egua ds a e necessa y o ensu e ha he use o high- isk a i icial in elligence ools in he ield o jus ice is ully compa ible wi h he ule o law. The implemen a ion and use o his echnology should be p eceded by he de ec ion and diagnosis o he unc ioning o jus ice in speci ic sec o s, so ha he e o s a e made in he a eas wi h mos p essing needs. REFERENCES [1] Emilia Bellucci and John Zeleznikow. 2006. De eloping Nego ia ion Decision Suppo Sys ems ha suppo media o s: a case s udy o he Family_Winne sys em. Jou nal o A i icial In elligence and Law 13, 2 (2006), 233–271. [2] Julianne Da danes. 2021. When Accessing Jus ice Requi es Absence om he Cou house: U ah’s Online Dispu e Resolu ion P og am and he Impac i Will Ha e on P o Se Li igan s. Peppe dine Dispu e Resolu ion Law Jou nal 21, 1 (2021). [3] John Dewa , Ba y W. Smi h, and Ca e Banks. 2000. Li igan s in Pe son in he Family Cou o Aus alia – Resea ch Repo No 20, Family Cou o Aus alia. Vol. Resea ch Repo No 20. Family Cou o Aus alia Canbe a. [4] Xuhui Fang. 2018. Recen De elopmen o In e ne Cou s in China. In e na ional Jou nal on Online Dispu e Resolu ion 5 (2018), 1–2, 49–55. [5] Roge Fishe and William U y. 1981. Ge ing o yes. Penguin G oup, New Yo k. [6] Hazel Genn and Y e e Genn. 1989. The e ec i eness o ep esen a ion a ibunals. Lo d Chancello ’s Depa men . [7] S ephan Landsman. 2009. The g owing challenge o p o se li iga ion. 13 Lewis & Cla k L. Re . 439 (2009). [8] A no Lodde and John Zeleznikow. 2010. Enhanced dispu e esolu ion h ough he use o in o ma ion echnology. Camb idge Uni e si y P ess. [9] Robe H. Mnookin and Lewis Ko nhause . 1979. Ba gaining in he shadow o he law: The case o di o ce. The Yale Law Jou nal 88, 5 (1979), 950. [10] E ic Niile . 2019. Can AI be a Fai Judge in Cou ? Es onia Thinks So. h ps: //www.wi ed.com/s o y/can-ai-be- ai -judge-cou -es onia- hinks-so/ [11] Shannon Sal e and Da in Thompson. 2016. Public-Cen ed Ci il Jus ice Redesign: A Case S udy o he B i ish Columbia Ci il Resolu ion T ibunal”. McGill Jou nal o Dispu e Resolu ion 3 (2016), 113. [12] Tania Sou din. 2018. Judge . Robo : A i icial In elligence and Judicial Decision- Making. UNSWLJ 41 (2018), 1114. [13] And ew S anie i, John Zeleznikow, Ma k Gawle , and B yn Lewis. 1999. A hyb id ule–neu al app oach o he au oma ion o legal easoning in he disc e iona y domain o amily law in Aus alia. A i icial in elligence and Law 7, 2-3 (1999). [14] Richa d E. Susskind. 2019. Online cou s and he u u e o jus ice. Ox o d Uni e si y P ess. [15] Da in Thompson. 2015. C ea ing New Pa hways o Jus ice Using Simple A i icial In elligence and Online Dispu e Resolu ion. In e na ional Jou nal o Online Dispu e Resolu ion 4 (2015). [16] S. Toulmin. 1958. The Uses o A gumen . Camb idge Uni e si y P ess, Camb idge. [17] Pe e K. Yu. 2020. The Algo i hmic Di ide and Equali y in he Age o A i icial In elligence, 72 FLA. L. REV 331 (2020). [18] Monika Zalnie iu e, Ly ia Benne Moses, and Geo ge Williams. 2019. The ule o law and au oma ion o go e nmen decision-making. The Mode n Law Re iew 82, 3 (2019), 425–455. [19] J. Zeleznikow. 2002. Using Web-based Legal Decision Suppo Sys ems o Imp o e Access o. Jus ice In o ma ion and Communica ions Technology Law 11, 1 (2002). [20] John Zeleznikow. 2020. The challenges o using Online Dispu e Resolu ion o suppo Sel Rep esen ed Li igan s. Jou nal o In e ne Law 3 23, 7 (2020). [21] John Zeleznikow. 2021. Using A i icial In elligence o p o ide In elligen Dispu e Resolu ion Suppo . G oup Decision and Nego ia ion (2021). h ps://doi.o g/10. 1007/s10726-021-09734-1 199