California court questions AI’s role in state bar exam

The use of AI in bar exam question development raises concerns about test validity and transparency in a system that determines the future of aspiring attorneys. California’s Supreme Court has publicly challenged the State Bar to explain its unauthorized use of artificial intelligence to create multiple-choice questions for the February examinations, adding complexity to an already troubled testing situation that included technical failures and a controversial move away from standardized testing models.

The big picture: California’s Supreme Court has demanded the State Bar explain its undisclosed use of AI to develop bar exam questions that affected hundreds of aspiring attorneys.

• The court was not informed before the February exams that the State Bar had permitted its independent psychometrician to use AI for creating test questions.
• This controversy emerges as the State Bar is already petitioning the court to adjust scores for hundreds of candidates who faced multiple technical issues during the exams.

Key details: The State Bar had previously received court approval for an $8.25-million deal with Kaplan to create test questions, but quietly deviated from this plan.

• Of the 171 scored multiple-choice questions, 100 were created by Kaplan, 48 were from a first-year law students’ exam, and 23 were developed using AI by ACS Ventures, the State Bar’s psychometrician.
• The State Bar only revealed this information in a news release nearly two months after the examination took place.

Why this matters: The controversy raises fundamental questions about test validity in professional certification and the appropriate use of AI in high-stakes assessments.

• The bar exam determines whether thousands of aspiring attorneys can practice law in California each year, making question reliability critical.
• The situation also highlights concerns about transparency as California moved away from the National Conference of Bar Examiners’ standardized testing model used by most states.

What the court wants to know: The Supreme Court is seeking comprehensive information about the AI implementation and quality control processes.

• Specifically, the court wants details on how and why AI was used to develop questions, what reliability measures were taken, and whether any AI-generated questions were removed after being determined unreliable.
• The court also requested information about the reliability of the remaining multiple-choice questions used for scoring.