“These scales influence whether someone is identified as experiencing distress or depression, so if there’s something incorrect in the translation, that directly affects clinical decisions,” Dr. Wei Teng said.
A Te Whare Wānanga o Waitaha University of Canterbury researcher is testing whether artificial intelligence can bridge language barriers as more Asian New Zealanders report mental health challenges, Medical Xpress reported. Dr. Teng, a scholar of translation studies, examined how effectively ChatGPT translates two standard screening instruments into Mandarin Chinese: the Kessler Psychological Distress Scale (K10) and the Patient Health Questionnaire-9 (PHQ-9).
Mandarin versions of these screening tools already exist. However, Teng said that some current translations feel overly formal or robotic, which can prevent patients from responding openly.
Teng designed structured prompts for ChatGPT, assigning the model a professional role and specifying “caring” and “clinically appropriate” tone requirements. One carefully engineered prompt produced stronger results, indicating that artificial intelligence could improve translation quality when guided precisely.
Inconsistent Results Raise Reliability Concerns
The study uncovered a major flaw regarding output consistency. When colleagues ran the identical prompt through their own free ChatGPT accounts, the resulting translations varied in tone and naturalness.

Large language models generate responses influenced by a user’s previous interactions, meaning outputs change between different users. That variability creates significant hurdles for clinical implementation.
“In mental health, consistency matters,” Teng said. “A screening tool must function the same way each time. Variability casts doubt on reliability and reproducibility.”
Human translation resources remain limited across Aotearoa New Zealand, prompting the search for digital alternatives. Josiah Jennings from the Christchurch-based Asian Community Transformation Trust and registered clinical psychologist Dr. Amy Wang discussed these access gaps with Teng before the research began.
Next Steps for AI Screening Tools
Teng is now collaborating with artificial intelligence researchers to determine whether more deterministic prompting or customized models can resolve the consistency issues. He emphasizes that clinical oversight and interdisciplinary teamwork remain vital before deployment.
“There may be room for AI in this space,” Teng said. “But when people’s well-being is involved, trust and cultural sensitivity must come first.”
Frequently Asked Questions
Why were the K10 and PHQ-9 selected for the study?
Researchers focused on the Kessler Psychological Distress Scale and the Patient Health Questionnaire-9 because they are widely used screening instruments that influence clinical decisions regarding distress and depression.
What specific prompt adjustments did researchers test?
Dr. Wei Teng assigned ChatGPT a professional role with strict tone requirements specifying both “caring” and “clinically appropriate” language.
Dr. Wei Teng holds a PhD in Translation and Interpreting from Auckland University of Technology and a Master of Arts in Applied Linguistics from the University of Adelaide. Before joining the University of Canterbury in 2018, he served as an interpreter and translator for the Republic of China Marine Corps in Kaohsiung, Taiwan, from 2002 to 2004. He received the MBIE Science Whitinga Fellowship in 2021 to advance research on healthcare communication across languages.
Who collaborated on the translation project?
Teng developed the study following conversations with Josiah Jennings from the Asian Community Transformation Trust and clinical psychologist Dr. Amy Wang.
Keep reading