Artificial Intelligence in Psychological Assessment: Advancing Precision and Individualisation in Autism

Abstract

Psychological assessment in mental health has evolved significantly, with standardised tools and clinical observations providing structured and reliable ways to understand cognitive, emotional, and behavioural functioning. Building on these foundations, recent advances in artificial intelligence (AI) offer additional opportunities to complement existing assessment approaches by enabling more objective, performance-based, and child friendly evaluations.

Within this broader landscape, autism spectrum condition represents a particularly meaningful area for the application of AI-enabled assessment. In children, assessment often relies on parent-reported measures, as self-report questionnaires may be developmentally challenging due to limited vocabulary, attention, or engagement. AI-based approaches enable performance-driven and observation based assessment by analysing real-time behavioural data including facial expressions, eye gaze, motor patterns, repetitive movements, and speech characteristics. Analysis of home videos and naturalistic interactions allows for the capture of everyday behaviours that may not be evident in structured clinical environments.

Non-invasive wearable devices can further support continuous monitoring by tracking physiological indicators such as heart rate variability and skin conductance, providing insights into sensory processing, stress regulation, and emotional arousal. Speech-based AI systems can detect changes in language, memory, and communication over time, while interactive AI-enabled toys and digital platforms can support speech development by breaking words into simple, monosyllabic components suitable for young children. Ongoing interdisciplinary research continues to refine AI-based approaches, with growing evidence supporting their feasibility, clinical relevance, and ability to support individualised assessment based on each child’s unique symptom profile.

Ethical and clinical considerations remain central, including bias in training datasets, explainability of AI systems, risks of over-medicalization, and cultural validity particularly in low and middle income contexts. Effective implementation will also require adequate infrastructural resources, clinician training, and adaptability to diverse clinical and community settings. AI should be viewed as a complementary tool that augments rather than replaces clinical judgment, supporting more precise, scalable, and child-centered assessments.