Wed, Sep 27, 2023 page2

• Legal advocates seek clarity on use of AI for rulings

  • By Shelley Shan / Staff reporter
  • • 
    • 
    • 

  The Judicial Yuan should first communicate with the public how artificial intelligence (AI) would assist judges in writing rulings, and what measures it would take to avoid erroneous and biased judgements, before using it in the legal system, legal professionals and legal justice advocates said yesterday.

  Representatives from several groups, including the Taiwan Bar Association and Judicial Reform Foundation, have expressed concern after the judicial branch last month announced that it would begin using generative AI to draft rulings on cases involving fraud, money laundering or unsafe driving once it completes a system test this week.

  The Judicial Yuan also plans to expand the use of AI in drafting rulings on civil lawsuits involving compensation for damage caused by traffic accidents and debt cancelations and liquidation.

  

  Representatives from the Taiwan Bar Association, the Taipei Bar Association and the Taiwan Association for Human Rights hold placards at a news conference in Taipei yesterday.

  Photo: Chen Yu-fu, Taipei Times

  It said the AI system that it would use is “specifically trained to produce drafts for rulings,” rather than being merely a draft generator, adding that the system is “a smart tool that can generate content using corresponding indictments and instantly analyze criminal facts and evidence in them.”

  “The system would transform the way judges compose rulings. However, the Judicial Yuan still has many unanswered questions about this major policy change,” the groups said yesterday.

  The Judicial Yuan last week said the system was trained using a large number of indictments and court rulings, but it did not indicate the time range from which legal documents were gathered for the training, they said.

  The system’s alleged ability to instantly analyze criminal facts and evidence to produce drafts is questionable if it simply generates content for draft rulings from court judgements and indictments and not from relevant files in the cases as well, the groups said.

  The Judicial Yuan also failed to explain how a third-party contractor built the database to train the AI system, and whether the data were stored and managed by the contractor or by the Judicial Yuan, they said.

  The public has the right to know how the contractor avoids problems caused by inadequate or biased information in the database when tweaking the parameters of the AI system, and how the Judicial Yuan and the contractor would settle liability issues whenever there are disputes, they said.

  Although the Judicial Yuan defended the use of AI in the justice system, saying that judges have the ultimate authority to determine the facts, apply the law and sentence in each case, the system might not accurately record changes to the generative content, they said.

  “To dismiss the public’s doubts, the Judicial Yuan should offer further explanations on whether judges still have a final say on the factors leading to their judgements, as well as other issues,” they said.

  Attorneys, prosecutors and other stakeholders should be allowed to try to access the database that is used to train the AI system, they said, adding that this would allow the public to jointly examine the bias and risks that might exist in the database.

  Although AI has been used in other countries to help compile evidence in legal cases, with the quality of the work under strict scrutiny, it has yet to be used in the international community to help judges draft rulings and assist them in making decisions, the groups said.

  A Purdue University study showed that 40 percent of people would not be able to detect errors in AI content that has an accuracy rate of less than 50 percent, a phenomenon that has been identified as “artificial hallucination,” the groups said.

  “The AI system creates draft rulings after judges click on whether defendants entered a plea deal and which laws were used to rule on the case. While the judges issue rulings, the reasons leading to the rulings are generated by AI, which might contain errors that cannot be detected by users. The Judicial Yuan did not explain how such risks could be avoided,” the groups said.

  Meanwhile, technology experts have warned that the extensive use of AI in the US justice system, particularly in helping judges in sentencing based on risks of defendants becoming repeated offenders, might reinforce the prejudice and racial inequality that already exist in the justice system, the groups said.

  Using AI to help judges in their rulings could cause them to overlook cultural factors, situations facing the stakeholders and their intent, which are crucial factors for judges to consider when reaching a decision and cannot be reproduced by AI, the groups said.
• Most Popular

  • 1

    Man convicted of double murder to be executed tonight

  • 2

    TSMC and Nvidia team up on silicon photonics, CEO Jensen Huang says

  • 3

    Hegseth vows to maintain peace in Strait

  • 4

    Nvidia CEO says its advanced packaging technology needs are changing

  • 5

    TikTok could be affecting views of young Taiwanese