Automating Fairness? Artificial Intelligence in the Chinese Court

tags: Law and Technology

Notes

difficulties of making a “techno-tatorship” real when local officials are more incentivized to show action than to produce useful software, and when poor quality data are perva- sive.

NOTER_PAGE: (6 . 0.48527004909983634)

centralizing oversight and control of the courts was a political priority that resonated with major themes em- phasized by Communist Party General Secretary Xi Jinping, including global technology primacy.

NOTER_PAGE: (8 . 0.16366612111292964)

Court leaders and rank-and-file judges saw the ways in which technology could boost the power and legitimacy of the courts and also ease their workload.

NOTER_PAGE: (8 . 0.3207855973813421)

Under pressure from widespread media coverage of wrongly-decided cases, the SPC endorsed judicial transparency as a goal in 2009 and started encouraging lower court efforts to pub- lish judicial decisions.

NOTER_PAGE: (8 . 0.4574468085106383)

new rules called on all courts to begin posting most cases online

NOTER_PAGE: (9 . 0.18166939443535188)

China Judgments Online also marked a victory for reformers inside the court system who saw disclosure as a way to strengthen public trust, combat corrup- tion, and help courts resist external pressure.

NOTER_PAGE: (10 . 0.16530278232405893)

NOTER_PAGE: (10 . 0.2479541734860884)

“Smart courts” is a capacious slogan that also sometimes encompasses basic techno- logical efforts to make it easier for lawyers and plaintiffs to file paper- work and follow the progress of a case online.

NOTER_PAGE: (11 . 0.36088379705400986)

better monitor both judges and society

NOTER_PAGE: (12 . 0.21358428805237317)

NOTER_PAGE: (12 . 0.2463175122749591)

NOTER_PAGE: (13 . 0.40589198036006546)

Although the judge would retain discre- tion to disregard the recommended sentence, research at the inter- section of law and psychology suggests that reference points “an- chor” judges’ decisions.

NOTER_PAGE: (14 . 0.2127659574468085)

judges are typically pressed for time and reluctant to take responsibil- ity for a decision that strays from the norm.

NOTER_PAGE: (14 . 0.32160392798690673)

technology could make their jobs easier and lives bet- ter.

NOTER_PAGE: (15 . 0.19639934533551556)

dash- board that sums up the daily activities of courts and judges is an ap- pealing source of information and control.

NOTER_PAGE: (15 . 0.21194762684124388)

prospect of a lighter workload and decreased responsibility for decision-making.

NOTER_PAGE: (15 . 0.2765957446808511)

saving time is a priority for many judges

NOTER_PAGE: (15 . 0.3199672667757774)

allow judges to sidestep responsibility for their decisions

NOTER_PAGE: (15 . 0.3878887070376432)

it is hard to imagine a judge getting into trouble for recommending an outcome suggested by court-approved software

NOTER_PAGE: (16 . 0.1955810147299509)

hmmmm… I think I could imagine it

A credible claim that any out-of-the- ordinary decisions are likely to be detected could also fend off pressure from politically well-connected litigants

NOTER_PAGE: (16 . 0.25859247135842883)

opportunity to position their work as cutting edge

NOTER_PAGE: (16 . 0.3240589198036007)

Xi Jinping’s push to centralize political power and lead the world in artificial intelligence

NOTER_PAGE: (16 . 0.41898527004909986)

significant step toward centralizing political power

NOTER_PAGE: (17 . 0.16366612111292964)

China’s push for AI is an important part of the country’s strategic response to slowing economic growth and is moti- vated by a pervasive belief in nationalist vindication through techno- logical innovation.

NOTER_PAGE: (17 . 0.22504091653027825)

emergence of a robust market for data analysis to make sense of the tens of millions of legal decisions now publicly available.

NOTER_PAGE: (17 . 0.3445171849427169)

most technological expertise resides in China’s technology companies, ra- ther than the courts.

NOTER_PAGE: (19 . 0.16693944353518822)

affinity between centralization of power and a groundswell of political interest in deciding “like cases alike” through algorithms perceived to be scientific, reliable, and impartial

NOTER_PAGE: (19 . 0.37479541734860883)

algorithmic justice cedes decision-making power to the person or persons who write the algorithm

NOTER_PAGE: (19 . 0.44926350245499186)

whose interests are served by the public release of official data

NOTER_PAGE: (19 . 0.5818330605564649)

outsourcing the design of software used for key court functions risks transferring public power to private hands.

NOTER_PAGE: (19 . 0.6448445171849427)

But building a complete repository of judicial data in a far-flung court system has proved a formidable challenge, and missing data remains a serious problem.

NOTER_PAGE: (20 . 0.5098199672667758)

compliance with ju- dicial disclosure requirements has been middling.

NOTER_PAGE: (21 . 0.18003273322422259)

The sheer number of missing documents suggests that the SPC is struggling to implement its own policy

NOTER_PAGE: (21 . 0.337152209492635)

disclosure rates appear largely to be considered a recom- mended target

NOTER_PAGE: (22 . 0.16530278232405893)

courts’ limited capacity to absorb and follow disclosure rules.

NOTER_PAGE: (23 . 0.3584288052373159)

censorship shapes the public record

NOTER_PAGE: (23 . 0.5)

disclosure is simply a low priority for often overworked courts who care more about other, more important performance targets.

NOTER_PAGE: (24 . 0.19639934533551556)

The success of judicial disclosure rests on resources, political will, and centralized control, commodities that are in short supply in many court systems worldwide

NOTER_PAGE: (24 . 0.369885433715221)

Recognizing that the vast amount of data the Chinese state is collecting includes sizable holes also complicates the common under- standing of China as an emergent techno-authoritarian superpower.

NOTER_PAGE: (25 . 0.4369885433715221)

collecting huge amounts of data and deploying it does not make the state omniscient. Indeed, it may create new chal- lenges when data quantity is prioritized over data accuracy.

NOTER_PAGE: (25 . 0.5310965630114567)

Media reports tend to extoll the adoption of artificial intelligence while providing few details, and even those working in courts said to be lead- ers in artificial intelligence often struggle to be specific about how it is changing judicial practice.

NOTER_PAGE: (27 . 0.18003273322422259)

technology that makes things easier for one state agency may create problems for others.

NOTER_PAGE: (29 . 0.2291325695581015)

police, procuratorate, and courts work toward algorithmic analytics in parallel

NOTER_PAGE: (29 . 0.3068739770867431)

Judges and scholars have also expressed skepticism about the usefulness of the software

NOTER_PAGE: (30 . 0.18085106382978725)

the current state-of-the-art software returns a long list of cases that are “similar, but useless.”

NOTER_PAGE: (30 . 0.34288052373158756)

re- quired to enter so much information into the system that “they already know the appropriate sentence” by the time they are done.

NOTER_PAGE: (30 . 0.40671031096563015)

well, maybe that's useful anyway. checklist type of thing

difficulty of arriving at a correct legal solution exclusively from the facts of the case and the text of the law.1

NOTER_PAGE: (31 . 0.3878887070376432)

does the underlying algorithm do a good job selecting “similar” cases?

NOTER_PAGE: (32 . 0.295417348608838)

What about past cases that were wrongly decided?

NOTER_PAGE: (32 . 0.3109656301145663)

Do opinions always accurately report the relevant facts in prior cases?

NOTER_PAGE: (32 . 0.3248772504091653)

call for legal experts to become more deeply involved in software design.

NOTER_PAGE: (32 . 0.40671031096563015)

claiming credit for technological accomplishment proved a more achievable political goal than im- proving court administration.

NOTER_PAGE: (33 . 0.14811783960720132)

no serious efforts to evaluate whether technology is improving court administration

NOTER_PAGE: (33 . 0.24222585924713586)

China can “pass as a high-performance state regardless of the actual results.”

NOTER_PAGE: (33 . 0.39770867430441903)

what matters is the existence of an electronic dashboard showing the records of individual judges, rather than assurance that the records are complete or even the analysis accurate.

NOTER_PAGE: (33 . 0.49918166939443537)

often an assumption that people are fallible while technology is less so.

NOTER_PAGE: (35 . 0.18003273322422259)

cf Franklin

Both the SPC and the market players that repackage court data are focused on celebrating the total number of cases available and, for all that kudos about a new norm of judicial disclosure are deserved, this boosterism obscures the time and effort required to convert atomized data into useful information.

NOTER_PAGE: (36 . 0.18821603927986907)

The irony is that for all the effort Chinese courts have invested in improving access to justice for ordinary people, embracing judicial disclosure may exacerbate the inequalities that dog adjudication worldwide

NOTER_PAGE: (36 . 0.34697217675941083)

Notes

difficulties of making a “techno-tatorship” real when local officials are more incentivized to show action than to produce useful software, and when poor quality data are perva- sive.

centralizing oversight and control of the courts was a political priority that resonated with major themes em- phasized by Communist Party General Secretary Xi Jinping, including global technology primacy.

Court leaders and rank-and-file judges saw the ways in which technology could boost the power and legitimacy of the courts and also ease their workload.

Under pressure from widespread media coverage of wrongly-decided cases, the SPC endorsed judicial transparency as a goal in 2009 and started encouraging lower court efforts to pub- lish judicial decisions.

new rules called on all courts to begin posting most cases online

China Judgments Online also marked a victory for reformers inside the court system who saw disclosure as a way to strengthen public trust, combat corrup- tion, and help courts resist external pressure.

broader push to disclose a broader range of information related to court operations.

“Smart courts” is a capacious slogan that also sometimes encompasses basic techno- logical efforts to make it easier for lawyers and plaintiffs to file paper- work and follow the progress of a case online.

better monitor both judges and society

better inform policy-makers by using big data to “deeply analyze the behavior of court users,” “monitor social contradictions,” and “predict trends in economic and social development.”

software capable of analyzing past cases with similar fact patterns to recommend sentences to judges.

Although the judge would retain discre- tion to disregard the recommended sentence, research at the inter- section of law and psychology suggests that reference points “an- chor” judges’ decisions.

judges are typically pressed for time and reluctant to take responsibil- ity for a decision that strays from the norm.

technology could make their jobs easier and lives bet- ter.

dash- board that sums up the daily activities of courts and judges is an ap- pealing source of information and control.

prospect of a lighter workload and decreased responsibility for decision-making.

saving time is a priority for many judges

allow judges to sidestep responsibility for their decisions

it is hard to imagine a judge getting into trouble for recommending an outcome suggested by court-approved software

A credible claim that any out-of-the- ordinary decisions are likely to be detected could also fend off pressure from politically well-connected litigants

opportunity to position their work as cutting edge

Xi Jinping’s push to centralize political power and lead the world in artificial intelligence

significant step toward centralizing political power

China’s push for AI is an important part of the country’s strategic response to slowing economic growth and is moti- vated by a pervasive belief in nationalist vindication through techno- logical innovation.

emergence of a robust market for data analysis to make sense of the tens of millions of legal decisions now publicly available.

most technological expertise resides in China’s technology companies, ra- ther than the courts.

affinity between centralization of power and a groundswell of political interest in deciding “like cases alike” through algorithms perceived to be scientific, reliable, and impartial

algorithmic justice cedes decision-making power to the person or persons who write the algorithm

whose interests are served by the public release of official data

outsourcing the design of software used for key court functions risks transferring public power to private hands.

But building a complete repository of judicial data in a far-flung court system has proved a formidable challenge, and missing data remains a serious problem.

compliance with ju- dicial disclosure requirements has been middling.

The sheer number of missing documents suggests that the SPC is struggling to implement its own policy

disclosure rates appear largely to be considered a recom- mended target

courts’ limited capacity to absorb and follow disclosure rules.

censorship shapes the public record

disclosure is simply a low priority for often overworked courts who care more about other, more important performance targets.

The success of judicial disclosure rests on resources, political will, and centralized control, commodities that are in short supply in many court systems worldwide

Recognizing that the vast amount of data the Chinese state is collecting includes sizable holes also complicates the common under- standing of China as an emergent techno-authoritarian superpower.

collecting huge amounts of data and deploying it does not make the state omniscient. Indeed, it may create new chal- lenges when data quantity is prioritized over data accuracy.

Media reports tend to extoll the adoption of artificial intelligence while providing few details, and even those working in courts said to be lead- ers in artificial intelligence often struggle to be specific about how it is changing judicial practice.

technology that makes things easier for one state agency may create problems for others.

police, procuratorate, and courts work toward algorithmic analytics in parallel

Judges and scholars have also expressed skepticism about the usefulness of the software

the current state-of-the-art software returns a long list of cases that are “similar, but useless.”

re- quired to enter so much information into the system that “they already know the appropriate sentence” by the time they are done.

difficulty of arriving at a correct legal solution exclusively from the facts of the case and the text of the law.1

does the underlying algorithm do a good job selecting “similar” cases?

What about past cases that were wrongly decided?

Do opinions always accurately report the relevant facts in prior cases?

call for legal experts to become more deeply involved in software design.

claiming credit for technological accomplishment proved a more achievable political goal than im- proving court administration.

no serious efforts to evaluate whether technology is improving court administration

China can “pass as a high-performance state regardless of the actual results.”

what matters is the existence of an electronic dashboard showing the records of individual judges, rather than assurance that the records are complete or even the analysis accurate.

often an assumption that people are fallible while technology is less so.

The irony is that for all the effort Chinese courts have invested in improving access to justice for ordinary people, embracing judicial disclosure may exacerbate the inequalities that dog adjudication worldwide