Researchers have explored whether a humanoid robot powered by artificial intelligence can assist scientists in reviewing evidence about drug-impaired driving, potentially offering a new approach to organizing and interpreting toxicology research.
The study, published in the Journal of Intelligent & Robotic Systems, examined collaboration between human researchers and a humanoid robot named Ami.
The research focused on driving under the influence of drugs (DUID), a major concern for road safety and forensic toxicology.
Identifying drugs in drivers requires reliable analytical methods, particularly when investigators must determine which substances were present in biological samples following a traffic incident.
The researchers conducted a systematic review following the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) guidelines.
They examined quantitative studies published between 2007 and 2025 that investigated substances associated with drug-impaired driving.
Studies without confirmatory chemical analysis were excluded to maintain the review's focus on analytically supported findings.
A total of 25 studies met the inclusion criteria.
Human researchers independently reviewed the selected publications, while the same study information was integrated into Ami's knowledge-based system for AI-assisted analysis.
The researchers then compared the information extracted through the two approaches.
According to the published findings, the humanoid-assisted analysis produced results broadly similar to those obtained through the conventional systematic review.
Both approaches identified alcohol, cannabinoids, cocaine, amphetamines, benzodiazepines and opiates among the substances commonly reported in the selected drug-impaired driving studies.
Blood and urine were the most frequently reported biological specimens, while oral fluid was used less often.
The review also identified several laboratory methods used to detect and confirm drugs.
These included immunoassays, gas chromatography–mass spectrometry (GC-MS) and liquid chromatography–mass spectrometry (LC-MS).
The authors described LC-MS as the most sensitive technique among the analytical approaches considered in their review.
An important aspect of the investigation was the robot's use of large language models to extract and interpret information.
The researchers reported that Ami often required additional human prompting to retrieve specific details.
Although they did not identify hallucinations in its answers during the reported comparison, the robot sometimes introduced examples originating outside the 25 studies supplied to its knowledge base.
This observation highlights the need to verify AI-generated information against original scientific sources.
The findings suggest that humanoid-assisted systems may help researchers organize complex toxicology literature more efficiently.
However, the study evaluated research-review assistance rather than the robot's ability to examine forensic specimens, operate laboratory instruments or independently interpret evidence in criminal cases.
The results therefore should not be interpreted as validation of humanoid robots for operational forensic casework.
Further evaluation is needed to establish how reliably such systems perform across larger research collections, different scientific questions and independent testing conditions.
The study illustrates the potential role of artificial intelligence as a supporting research tool while emphasizing the continuing importance of human scientific oversight.