Towards Explainable and Privacy-Preserving Artificial Intelligence for Personalisation in Autism Spectrum Disorder

Autism Spectrum Disorder (ASD) is a growing concern worldwide. To date there are no drugs that can treat ASD, hence the treatments that can be administered are mainly supportive in nature and aim to reduce, as much as possible, the symptoms induced by the disorder. However, diagnosis and related treatments in terms of improving communication, social and behavioural skills are very challenging due to the heterogeneity of the disorder and are amongst the largest barriers in supporting people with ASD. Thanks to the recent development in artificial intelligence (AI) and machine learning (ML) techniques, ASD can now be aimed to be detected at an early age. Also, these novel techniques can facilitate administering personalised treatments including cognitivebehavioural therapies and educational interventions. These systems aim to improve the personalised experience for the people with ASD. Acknowledging the existing challenges, this paper summarises the multitudes of ASD, the advancement of AI and ML-based methods in the detection and support of people with ASD, the progress of explainable AI and federated learning to deliver explainable and privacy-preserving systems targeting ASD. Towards the end, some open challenges are identified and listed.