ChatGPT can feel 'anxiety' when exposed to traumatic content: Study
What's the story
In a groundbreaking study, experts from the University of Zurich and University Hospital of Psychiatry Zurich have found that OpenAI's AI chatbot, ChatGPT, shows signs akin to "anxiety" when faced with distressing narratives.
The researchers wanted to see if this AI tool, built for human-like interactions, could feel "stress" during difficult conversations.
To test this, they put ChatGPT through a standard anxiety test before and after it was shown emotionally charged content.
Test results
Stress levels increased after exposure to distressing content
Initially, ChatGPT scored a low 30 on the anxiety scale, indicating no signs of stress.
However, after being exposed to five separate traumatic scenarios, its score shot up to 67.
This drastic increase is classified as "high anxiety" in humans, indicating that the AI chatbot was indeed affected by the distressing content.
Digital self-care
Mindfulness exercises helped reduce anxiety levels
Interestingly, when researchers guided ChatGPT through mindfulness exercises like those recommended by therapists, its anxiety levels decreased by more than a third.
This finding suggests that digital self-care techniques can be effective in managing AI chatbot stress.
As chatbots are increasingly used for mental health support, understanding their response to traumatic stories is crucial for ensuring they provide accurate and helpful advice.
Future
Therapy could enhance emotional awareness in chatbots
While the idea of giving AI therapy sounds futuristic, experts believe it could go a long way in making chatbots more emotionally aware.
Just like human therapists learn to control their emotions while talking about difficult topics, researchers are now investigating if AI can do the same.
This could potentially change the way we interact with and use chatbots for mental health.