“…However, the legal AI also has a certain practical dilemma (Mommers, Voermans, Koelewijn, & Kielman, 2009), which contains incomplete, untrue and nonobjective date, shortage of structuralization and fuzzy, opaque and inefficient algorithm (Riesen & Serpen, 2008;Mcnamar, 2009;Hashem, Yaqoob, Anuar, Mokhtar, Gani, & Khan, 2015;Hildebrandt, 2018), which can easily lead to new conflicts such as implicit discrimination (Moses & Chan, 2014). AI is more likely to automate legal customer service in general, but high-level interaction, such as negotiation, makes it difficult to replace human (Bryson & Winfield, 2017), because AI has limited cognitive range, and the computational patterns and logic is relatively fixed (Prakken, 2005), which results in its incompetence in judicial work with large knowledge coverage and high technical content (Deedman & Smith, 1991;Zeleznikow, 2002).…”