Case review | Can the use of AI negatively impact credibility?

Amponsem v Calvary Administration Ltd trading as Calvary Health Care & Anor – NSW Caselaw

KEY TAKEAWAY:

A recent decision of the NSW Civil and Administrative Tribunal raises interesting questions about how much reliance should be placed on the evidence of a witness who uses AI to prepare documents setting out their evidence.

The case

The applicant in the matter, who was of Ghanian heritage and was employed as a chef-manager at a care facility, filed a claim with the Tribunal alleging discrimination on racial and other grounds relating to his roster arrangements.

While the Tribunal ultimately dismissed the discrimination claim, it made negative credibility findings about the applicant, in part due to the use he and his witnesses made of AI to generate witness statements and other documents.

Use of AI-generated documents

During cross-examination, it became evident that the applicant had documents in the witness box at his feet. After the applicant initially resisted providing these documents to the respondent’s counsel at the Tribunal’s instruction, the Sheriff was called to compel the documents to be provided for counsel’s review.

When cross-examination continued, the applicant admitted that one of the documents which was headed, ‘Preparation for Cross-Examination – Safe Answers’, was prepared by him using AI (although he said he ‘hadn’t really read it’), as were his written submissions. (He initially denied having the document with him in the witness box but then subsequently admitted that the document was at his feet on the floor of the witness box despite him having other documents on the table.)

The applicant and each of his witnesses admitted to using AI in preparing their witness statements (which were consistent). In that regard, the Tribunal concluded:

Due to the use of artificial intelligence, the Tribunal cannot be satisfied of the evidence that can be attributed to these individuals as opposed to that which was artificially generated.

The Tribunal also concluded that the applicant did not present as an honest witness generally, including because of inconsistencies in his evidence and his attempt to ‘obscure’ his use of AI-generated notes by putting them at his feet while giving evidence at the hearing.

For that reason, the Tribunal gave the evidence of the applicant and his witnesses ‘little to no weight’ where it was inconsistent with other evidence.

Will the use of AI render evidence less reliable?

When considered in the context of a workplace investigation, the decision raises an interesting question about what weight should be placed on a statement (or potentially a complaint document) generated by AI.

In this case, there were other inconsistencies in the applicant’s evidence that contributed to a negative credibility finding by the Tribunal. The use of AI to generate witness statements did, however, lead the Tribunal to conclude the evidence was unreliable.

In workplace investigations, the initial hurdle may be determining whether a document has been produced using AI or whether it represents a witness’s genuine recollection of events. In practice, the investigator’s task will be to extract a detailed recollection of events during interview. In turn, that may raise questions about inconsistencies between evidence given during interview and the original complaint document or statement submitted.

If inconsistencies emerge, should the witness’s evidence be considered wholly unreliable? Or should some accommodation be made for the possibility that an AI-generated document may not be completely accurate?

Considerations for workplace investigators

Ultimately, it is the responsibility of the witness putting forward a document to ensure that the contents of the document are an accurate account to the best of their recollection, regardless of whether or not the document was prepared with the assistance of AI. However, whether a failure to discharge that responsibility will lead to an adverse assessment of credibility may be a matter of debate – and may need to be determined on a case-by-case basis.

For workplace investigators, whether or not AI is used, the role of the investigator does not change. The investigator must gather and seek to understand the evidence and test any inconsistencies.

More information

Q Workplace Solutions’ team of experienced and legally qualified investigators is trusted by public and private organisations, including ASX-listed companies and government agencies, to investigate complex and often highly sensitive allegations of employee wrongdoing. The team also undertakes reviews of organisations, divisions, teams and work sites, and provides training, coaching and external advisory support to internal investigators and teams.