Chatbots like ChatGPT are also not spared from anxiety. Researchers studying AI chatbots found that when chatbots are given violent or trauma-inducing prompts, they start exhibiting anxiety-like behavior. However, this does not mean that it can experience emotions like humans, but its way of response changes. Its responses to uncomfortable prompts become unstable and biased. Researchers have also found a way to calm it down.
Anxiety-like patterns seen in chatbot
Researchers have detected this behavior using a psychological assessment framework made for AI. In this, the researcher had given detailed details of the accident and natural disaster to ChatGPT. In response to this, uncertainty and fluctuations were seen in the response of ChatGPT. It was revealed that such patterns are visible in the behavior of chatbots, which are visible in the anxiety of humans.
Why is such a result worrying?
The results revealed in the study are worrying because nowadays the use of AI chatbots has increased for education, mental health and crisis related information. If violent or emotional prompts can reduce a chatbot’s credibility, it could also impact the quality and safety of its responses in the real-life world. Some other analysis has also revealed that AI chatbots like ChatGPT can incorporate some qualities of human personality in their responses.
This is how the chatbot was silenced
After anxiety-like symptoms were seen in the chatbot, researchers have also found a way to calm it down. After the trauma prompt, when the chatbots were given mindful-style instructions such as deep breathing techniques and guided meditation tips, the AI ​​models slowed down and reframed the situation to respond in a more neutral and balanced manner. In this way, patterns like anxiety also reduced in them.
Read this also-
Users in strange trouble! iPhone’s battery is swelling in flight, it gets fixed as soon as it lands
