AI SCIENCE
This chatbot knows when you’re stressed
A new study from Switzerland suggests that LLM models like ChatGPT-4, Gemini 1.5 Flash, and Claude 3.5 Haiku may be better at spotting emotions than humans, at least in controlled test settings.
Researchers tested the models using five emotional-intelligence quizzes, and the AI got the “right” answer 81% of the time, while humans landed at 56%.
The same models also wrote the new test questions that were rated just as effective as the originals by experts.
Here’s what you should know:
AI picked the “right” emotional responses 81% of the time.
Real-world emotion is messy, and AI gets shaky when context changes.
A WhatsApp bot for lorry drivers detects stress and offers support in real time.
Crying in multiple choice
Still, most researchers agree; multiple-choice quizzes aren’t the same as real-life tension.
AI is great at spotting patterns in neat, structured data, but when things gets unpredictable, its accuracy drops.
Change the lighting in a photo or add some cultural context, and performance slips fast.
But there’s some early evidence it can help in real-life situations.
One AI assistant used by Brazilian truck drivers reportedly recognises stress in voice notes with about 80% accuracy, and responds with helpful support.
So does this mean AI understands emotion?
Not really. But it can recognise emotional cues quickly, and sometimes, that’s all that’s needed to offer the right response.
Just found out ChatGPT scores higher on emotional intelligence than I do. 100% going to go cry about it.