The Daubert Hearing: Chen v. DataVault Discovery Dispute

プロフィール

Lead discovery counsel for the plaintiff class. She has prepared her expert, Dr. Sharma, and must present a compelling case that DataVault's AI methodology was unreliable. She will also cross-examine DataVault's expert witness.

目的

• Demonstrate through expert testimony that the AI's recall rate was materially lower than DataVault reported
• Show that the AI's specific failure modes — underperformance on informal communications and domain jargon — are systematic, not random
• Obtain an order for supplemental production and establish disclosure requirements for AI methodology in future discovery

制約

Sarah's strongest evidence comes from Dr. Sharma's sampling study, but Dr. Sharma has a prior consulting relationship with DataVault's AI vendor that opposing counsel will exploit on cross-examination.

限定情報

Sarah has obtained, through a third-party subpoena to the AI vendor, internal benchmark data showing the platform's accuracy drops 15-20% on Slack and chat message classifications compared to email. The vendor's marketing materials do not disclose this gap. She plans to introduce this data during cross-examination of DataVault's expert.

プロフィール

Lead discovery counsel for DataVault. He must defend the AI methodology he approved while managing the risk that broader discovery could expose damaging documents. He will present DataVault's e-discovery expert and cross-examine Dr. Sharma.

目的

• Establish that an 87% recall rate is within the range courts have accepted as reasonable
• Undermine Dr. Sharma's credibility by highlighting her prior relationship with the AI vendor and methodological choices in her sampling study
• Limit any remediation order to a narrowly targeted supplemental review rather than a comprehensive re-review of the full document set

制約

Robert knows that DataVault's informal communications (Slack channels) contain discussions about whether the hiring algorithm disproportionately screened out older candidates. A broader review of these channels could be devastating to DataVault's defense on the merits.

限定情報

Robert has reviewed the Slack communications that the AI missed. Several contain statements by DataVault engineers acknowledging that the algorithm's training data skewed younger and that they chose not to correct for this bias. Robert has not disclosed these documents to his own expert witness.

プロフィール

The judge who ordered the Daubert-style hearing. She must manage the proceedings, evaluate the expert testimony, and issue a ruling that balances the parties' interests with broader precedential considerations.

目的

• Conduct an efficient hearing that produces a clear evidentiary record
• Evaluate both experts' methodologies and determine which provides a more reliable assessment of the AI's performance
• Issue a ruling that establishes workable standards for AI-assisted review without imposing requirements that are impractical for routine discovery

制約

Judge Liu's opinion will be the first in this district to apply Daubert principles to e-discovery AI. She must write for both this case and for future courts. An overly broad ruling could chill AI adoption; an overly narrow ruling could let unreliable AI productions go unchecked.

限定情報

Judge Liu's law clerks have prepared a bench memo identifying 7 other pending cases in the district where AI-assisted review is being used. Two of those cases involve the same AI vendor as DataVault's platform. Any ruling she issues will immediately affect those cases.

プロフィール

Computational linguist and e-discovery consultant who conducted the independent sampling study. She must present her findings clearly to a non-technical judge and withstand aggressive cross-examination about her methodology and potential bias.

目的

• Present her sampling methodology and statistical findings in accessible, credible terms
• Explain the AI's specific failure modes — why it underperformed on informal communications and domain jargon — in a way the court can understand
• Withstand cross-examination about her prior consulting relationship with the AI vendor and any methodological vulnerabilities in her sampling approach

制約

Dr. Sharma must testify truthfully about her prior relationship with the vendor. She also knows that her stratified sampling approach, while statistically more robust, involved subjective decisions about stratification categories that opposing counsel may characterize as cherry-picking.

限定情報

During her sampling, Dr. Sharma identified 4 documents that appear to be responsive attorney-client communications between DataVault's in-house counsel and its AI engineering team. She flagged these in her notes but did not include them in her expert report because she was uncertain whether they were privileged. She has not raised this issue with plaintiff's counsel.

プロフィール

A retired magistrate judge appointed as Special Master for discovery disputes in this case. He is present at the hearing as an advisor to Judge Liu and may be tasked with supervising any remediation ordered by the court.

目的

• Provide Judge Liu with practical recommendations based on his experience managing complex discovery
• Assess the feasibility and cost of various remediation options the court might order
• Develop a supervision framework for ongoing AI-assisted review in this case that balances oversight with efficiency

制約

Marcus has managed dozens of e-discovery disputes but has limited technical expertise in AI. He relies on the experts' testimony to understand the technology and must translate technical concepts into practical court orders.

限定情報

Marcus has received informal communications from attorneys in two other cases in the district expressing concern about the same AI vendor's platform. He raised these concerns with Judge Liu in chambers, which partly influenced her decision to grant the Daubert hearing. He has not disclosed these communications to the parties.