Cenário da Simulação
It is Wednesday, 10:00 AM. Diana Rivera has called an emergency review meeting in the main conference room. The prospective client meeting was postponed with a vague excuse. Sarah Park has printed the gerado por IA memo alongside her Westlaw verification showing three fabricadas citations and a conflated statutory analysis. Marcus Chen has been asked to bring his laptop with the AI conversation history. Dr. James Whitfield has prepared a one-page summary of the governança de IA recommendations he submitted two months ago that were never implemented. Everyone in the room knows this could have been a disaster. The question now is what happens next.
Partes Interessadas e Papéis
modules.m3.simulation.stakeholdersSubtitle
Diana Rivera — Sócio-Gerente
Perfil
Championed the AI initiative, now facing the consequences of inadequate implementação. Must balance responsabilização with morale and the firm's continued adoção de IA.
Objetivos
- Determine appropriate responsabilização without scapegoating
- Estabelecer salvaguardas imediatas para prevenir recorrência
- Preserve the firm's adoção de IA momentum while restoring confidence
Restrições
Must report to the seguradora de responsabilidade profissional next week. Knows that overreacting could cause staff to abandon ferramenta de IAs entirely, while underreacting could lead to a real incident.
Informação Exclusiva
Diana has received an anonymous email from another paralegal stating that gerado por IA memos have been sent to clients in at least two other matters without any attorney verification. She has not yet confirmed whether this is true.
Marcus Chen — Paralegal Sênior
Perfil
The paralegal who generated the memo. Experienced, overworked, and deeply embarrassed. Knows the ferramenta de IA was used exactly as most staff had been using it.
Objetivos
- Demonstrate that the fluxo de trabalho failure is systemic, not individual
- Proteger sua posição profissional e continuidade no emprego
- Defender expectativas realistas de carga de trabalho ao lado de novos protocolos de IA
Restrições
Sabe que admitir a natureza sistêmica do problema pode implicar o escritório em responsabilidades mais amplas.
Informação Exclusiva
Marcus has chat logs showing that the ferramenta de IA's own documentation recommends 'always verifying legal citations against authoritative databases' — a warning that was not included in the firm's training materials. He also knows two other paralegals have been doing the same thing.
Sarah Park — Advogada Associada
Perfil
The attorney who assigned the memo and caught the errors. Relieved but aware of her own role in the delegation failure. Under pressure from partners to maximize impulsionado por IA efficiency.
Objetivos
- Estabelecer que a delegação foi razoável, mas que a lacuna de verificação foi uma falha de supervisão
- Pressionar por protocolos claros de atribuição que especifiquem jurisdição, profundidade e requisitos de verificação
- Abordar a cultura de pressão de prazo que incentiva atalhos
Restrições
O sócio para quem reporta disse explicitamente aos associados para 'deixar a IA fazer o trabalho pesado'. Questionar essa abordagem pode prejudicar seu progresso na carreira.
Informação Exclusiva
Sarah discovered during her verification that one of the fabricadas cases was actually a real case name but with completely different facts and holdings. The AI had taken a real citation and attached fabricadas content to it — a more sophisticated form of alucinação than simple fabricação.
Dr. James Whitfield — Garantia de Qualidade Lead
Perfil
The firm's governança de IA specialist who has been advocating for stricter protocols since being hired. Has a comprehensive governança proposal that was submitted but never reviewed.
Objetivos
- Use this incident to implement the governança framework he has been proposing
- Estabelecer um processo formal de revisão e relato de incidentes de IA
- Assegurar orçamento e autoridade para avaliações contínuas de competência em IA
Restrições
Deve apresentar reformas como construtivas em vez de punitivas para manter relações com os sócios.
Informação Exclusiva
James has benchmarked the firm's AI practices against 15 other firms of similar size. Rivera & Goldstein ranks in the bottom quartile for governança de IA maturity. He also discovered that the enterprise AI fornecedor recently updated its termos de serviço to include a liability limitation clause that the firm has not reviewed.
Regras
Duração
45 minutos
Comunicação
Formato de discussão aberta; Diana preside a reunião e gerencia a ordem de fala.
Método de Decisão
The meeting must produce three written outcomes: (1) an responsabilização determination, (2) immediate process changes effective today, and (3) a 30-day action plan. Diana has final decision authority but must secure consensus from at least two other participants.
Fases
Revisão do Incidente (20 minutos)
Diana opens the meeting and asks each participant to present their account of what happened. Marcus walks through his process, including the exact prompt he used. Sarah explains how she caught the errors and what specifically was wrong. James presents his governança assessment. Each person has 4-5 minutes. No cross-examination yet — this phase is about establishing the facts.
Root Cause Analysis and Responsabilização (25 minutes)
Open discussion about what went wrong and who bears responsibility. Participants may reveal exclusive information strategically. Diana must navigate between individual responsabilização and systemic reform. Key tensions: Was this Marcus's fault for not verifying? Sarah's fault for a vague assignment? Diana's fault for championing adoption without governança? The firm's fault for ignoring James's proposals?
Resolução e Reforma (20 minutos)
The group must agree on three deliverables: an responsabilização determination (consequences, if any, for individuals), immediate process changes (what changes today), and a 30-day action plan (what the firm will implement within a month). Each participant advocates for their priorities. Diana must forge consensus and make final decisions.
modules.m3.simulation.simVariationsTitle
- What if the memo had reached the client? Replay the scenario assuming Sarah did not catch the errors and the memo was forwarded to the prospective client, who then retained the firm based on the inflated assessment. How does the responsabilização calculus change?
- E se o rumor anônimo for verdadeiro? Durante a Fase 2, Diana revela que recebeu uma dica anônima sugerindo que este não foi o primeiro incidente de IA — Marcus usou IA de forma inadequada em outro caso há três meses. Como isso muda a dinâmica?
- What if Marcus pushes back? During the responsabilização discussion, Marcus reveals that the partner pressuring associates to maximize AI efficiency once told him directly: 'Just let the AI handle it, that is what we are paying for.' How does this change the responsibility analysis?
Debriefing
modules.m3.simulation.debriefSubtitle
Processo de Decisão
- What specific elements of Marcus's original prompt contributed to the AI producing fabricadas output?
- Como você teria estruturado o prompt de forma diferente? Escreva seu prompt aprimorado e defenda suas escolhas.
- É realista esperar que todo paralegal seja um especialista em engenharia de prompts? Quem é responsável pela competência de IA?
- How do you distinguish between an resultado de IA that is wrong because of a bad prompt versus one that is wrong despite a good prompt?
On Supervisão and Verification
- Should gerado por IA produto do trabalho be subject to different verification standards than human-generated produto do trabalho? Why or why not?
- How do you build verification into fluxo de trabalhos without making ferramenta de IAs slower than doing the research manually?
- Que papel o advogado responsável pela atribuição deve desempenhar ao especificar como a IA deve ser usada?
Padrões Profissionais
- Como uma cultura de pressão de prazo e métricas de eficiência contribui para o uso inadequado de IA?
- What is the right balance between encouraging AI experimentation and enforcing controle de qualidades?
- Should firms measure resultado de IA quality as rigorously as they measure adoção de IA rates?
- Como você cria segurança psicológica para relatar erros relacionados a IA?
Aplicação no Mundo Real
- Have you ever used an ferramenta de IA for research without verifying every citation? What made you trust the output?
- Does your organization have clear protocols for assistido por IA legal research? If not, what would you propose?
- What is the single most important engenharia de prompts habit that would have prevented this incident?
- Name one concrete change you will implement in your own fluxo de trabalho within the next week.
Referências e Fontes
Padrões Profissionais
- ABA Regra Modelo 5.3 — Responsibilities Regarding Nonlawyer Assistants (supervisão of assistido por IA work)
- ABA Opinião Formal 512 (2024) — IA Generativa and the duties of competência, confidencialidade, and supervisão
- Ordem dos Advogados of California Practical Guidance on AI for Lawyers (2024) — Verification and supervisão requirements
Engenharia de Prompts and Governança de IA
- Legal Engenharia de Prompts: Principles for Reliable Assistido por IA Research — Stanford CodeX (2024)
- Diretrizes da AALL para o Uso de IA em Pesquisa Jurídica (2024)
- International Legal Technology Association — Governança de IA Framework for Escritório de Advocacias (2024)
Pronto para Executar Esta Simulação?
This simulation is designed for guided facilitation as part of the Lawra Learning Program. Request a session with role assignments, engenharia de prompts exercises, and expert debriefing for your team.
Comentários
Carregando comentários...