Can we trust AI to detect healthy multilingual English speakers among the cognitively impaired cohort in the UK? An investigation using real-world conversational speech

2026-02-13 · Source: Computation and Language · Field: Technology & Digital — Artificial Intelligence & Machine Learning, Data Science & Analytics, Natural Language Processing · Depth: Advanced, quick

Summary

A study investigated the trustworthiness and bias of AI models in detecting cognitive decline among healthy multilingual English speakers in the UK. Researchers recruited monolingual participants nationally and multilingual speakers from community centers in Sheffield and Bradford, who spoke Somali, Chinese, or South Asian languages and had distinct Yorkshire accents. While Automatic Speech Recognition (ASR) systems showed no significant bias, classification and regression models utilizing acoustic and linguistic features demonstrated bias against multilingual speakers, particularly in memory, fluency, and reading tasks. This bias was exacerbated when models were trained on the DementiaBank dataset, leading to a higher rate of misclassification of healthy multilingual individuals as cognitively impaired. The study found that current AI models are not yet reliable for diagnostic use in these diverse populations.

Key takeaway

For research scientists developing AI tools for cognitive assessment, you should prioritize bias mitigation strategies, especially when working with diverse linguistic and accent groups. Your models must be rigorously tested on multilingual populations, including those with non-native English accents, to prevent misclassification and ensure diagnostic reliability. Do not rely solely on publicly available datasets like DementiaBank without thorough bias analysis and mitigation.

Key insights

AI models for cognitive decline detection exhibit bias against multilingual English speakers, particularly those with specific accents.

Principles

• AI diagnostic tools require bias mitigation.
• Public datasets can introduce or amplify bias.

Method

The study recruited monolingual and multilingual UK speakers, dividing multilinguals by language and accent, then evaluated ASR, classification, and regression models for bias using acoustic and linguistic features.

In practice

• Test AI models on diverse linguistic groups.
• Evaluate models for accent-specific biases.
• Avoid biased datasets for training.

Topics

• AI Bias
• Cognitive Decline Detection
• Multilingual Speech Analysis
• Dementia Datasets

Best for: Research Scientist, AI Researcher, AI Scientist, AI Ethicist

Related on AIssential

• Multilingual Detection of Alzheimer's Disease from Speech: A Cross-Linguistic Transfer Learning Approach — Cross-language transfer learning effectively detects Alzheimer's Disease from speech across multiple languages, achieving strong generalization and rapid inference.
• Creating Multilingual Mental Health Dialogue Datasets: Limits of Persona-Based Localization via Nationality and Language — Simple nationality and language parameter changes in English-centric personas fail to preserve clinical consistency in multilingual mental health dialogues.
• Fair Cognitive Impairment Detection Through Unlearning — Multimodal fusion and gradient reversal unlearning can significantly reduce demographic bias in MCI detection models while improving overall performance.
• Listening Between the Lines: Joint Learning of ASR Embeddings and LLM-Augmented Linguistics for Dementia Detection — Jointly learning ASR embeddings and LLM-augmented linguistics significantly improves dementia detection from speech.
• How Dialect Variation Challenges Natural Language Processing Systems — NLP systems struggle with dialects due to biased training data, leading to technical and social inequities.
• P3B3: A Multi-Turn Conversational Benchmark for Measuring European and Brazilian Portuguese Variety Bias in LLMs — Most LLMs exhibit a strong bias towards Brazilian Portuguese, highlighting the need for balanced multilingual representation.

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by Computation and Language.