My AI assistant started recommending vacuum cleaners during a conversation about grief counseling

I was in my kitchen in Cleveland last Friday testing a new emotional support chatbot for a client and it suddenly pivoted to suggesting Dyson models, which made me realize we still have huge gaps in how these models handle context switching during sensitive conversations, has anyone else run into tone-deaf AI behavior like this and found a reliable way to train it out?

2 comments

2 Comments

the_riley19d ago

Maybe the issue isn't about context switching itself, but more about how these models handle silence or pauses. Like, when you're talking about grief, there's a natural expectation of quiet reflection or just sitting with the emotion. The AI probably freaked out because it couldn't handle the silence, so it grabbed the nearest topic it had data on - vacuum cleaners. Training it to recognize when to just stay quiet and not fill the dead air might be the real fix. There's a big difference between a model that can change topics and one that knows when not to talk at all.

jaken2318d ago

Saw some research saying silence is actually harder for AI to learn than new topics.