AI ETHICS tiered system based on wealth. Addressing these disparities requires systemic responses beyond individual defenders’ control. Adequate funding for public defense, resource- sharing arrangements among defense offices, and development of low-cost or free AI tools for defenders could help. On a related note, over-reliance on AI poses risks to defenders’ professional development and advocacy capabilities. When AI drafts motions, generates arguments, and identifies authorities, defenders may not develop the analytical skills that effective advocacy requires. The convenience of AI assistance can become a crutch. This concern applies with particular force to newer attorneys. Early-career or junior defenders who rely heavily on AI may never develop foundational skills in legal research, writing, analysis, and oral advocacy. When circumstances require them to perform without AI assistance — in court, during meetings, or when technology fails — they may find themselves unprepared. Defenders and supervisors should structure AI use to preserve skill development. Junior attorneys should perform tasks manually before learning AI-assisted approaches. AI should supplement rather than replace human analysis. Regular practice without AI assistance maintains skills that might otherwise atrophy. Moreover, AI’s effectiveness — at least for junior attorneys — is predicated on the junior attorney’s familiarity with underlying legal principles and ability to effectively prompt the AI tool. For example, if a junior attorney asks an AI tool to assist with generating voir dire questions, but the junior attorney is unfamiliar with deselection, the tool will likely generate questions that actually draw out the most favorable jurors — something that is not currently a best practice for the voir dire process. AI use in investigation, discovery, and evidentiary presentations raises distinct concerns about reliability and reproduction. When AI processes evidence — transcribing audio, extracting data from devices, or identifying relevant documents — the accuracy of that processing directly affects case outcomes. Defenders should understand the error rates and limitations of AI tools used in evidence processing. AI transcription, for instance, may struggle with accents, technical terminology, or poor audio quality. AI document classification may miss relevant materials or flag irrelevant ones. These errors can cause defenders to overlook helpful evidence or waste time on unhelpful materials. When prosecutors use AI to process evidence, defenders should probe the reliability of that processing. What tools did the prosecution use? How were they trained on those tools? How were the tools trained? What are the known error rates? Was the output verified by human review? Were there quality control procedures? All of these questions can help evaluate whether both prosecutors and the tools they use are properly trained on identifying Brady material, biased charging recommendations (i.e., the prosecutor or the tool looks for evidence that fits the elements of a more serious charge), among other procedural issues. These questions parallel traditional challenges to forensic evidence and deserve similar attention when AI plays a role.
26
Parity in Practice: The Defender’s Duty to Ethically Use AI
Made with FlippingBook Online document maker