How AI Is Supposed to Help
The makers claim that AI can provide mental health therapy, as well as making huge improvements to education and physical health. Unfortunately it's being pushed out so quickly that there's been no time for extensive research either on benefits or risks. The rush is to make profits: companies market AI as a way to get rid of workers and profit from the wages that won't be paid.What Could Happen?
Becoming Isolated: Chatbots are designed to be so attractive that autistic individuals may not spend enough time on genuine relationships with family and friends. Naturally, these relationships will weaken, leading to isolation.
Feeling Worthless: In the face of bullying or being sidelined by their social group, autistic individuals may come to doubt their own value as a person. They may get back their self-esteem by becoming expert in a specialist subject. Artificial intelligence threatens this. It covers all subjects. It presents everything "confidently", even filling gaps with bogus information. This seems to devalue the knowledge that any human can acquire, removing this source of confidence.
Becoming Radicalised: AI is able to create persuasive text for any purpose. Extremist groups are using AI to target autistic people with anti-migrant material
What Can We Do?
Parents, carers and friends, can of course watch for signs of the problems described above. There are other ways to help:Give them free what AI companies want to charge them for: AI is "non-judgmental" and accepts what the user says without pushing back. Keeping them engaged is profitable. Convince your child that they can talk freely to you without fear of your judging them. A "rent-a-friend" relationship with a computer will then be less tempting.
Time Limits: Unfortunately, due to its use in education, it may not be possible to keep AI out of young poeple's lives until they are able to make mature judgments about it. We need to talk about the need to limit AI use, and observe how they are controlling this.
Other Approaches: While things are so uncertain, it's best to keep AI away. If this is not possible, AI which is missing the features that exploit users may help. For example, Goblin Tools is funded by sponsorship so it does not need to hold the user for as long as possible. There is also "Caveman Mode", recently introduced on some AI platforms, which doesn't pretend to be a personality, which may be worth looking at.
Notes
The talks and discussions on the AI Chatbot page may be helpful in explaining the possible risks of AI chatbots for the young.
This page is based on a document prepared for us from a dissertation on a topic we proposed under the UCL SU Community Research Initiative on "Autism and AI". The full document is here.