Welcome Stranger to OCC!Login | Register

Robots Should Dumb Down Their Speech to be More Relatable to Humans

Category: General News
Posted: 10:42PM

Robots can, or soon will, make many aspects of our lives a whole lot easier. However, there's just one problem: the robots act too intelligent. That's right, our future robot overlords are too smart for their own good, and would actually benefit from dumbing down their speech when talking to us meatbags. A study by both Cornell and Carnegie Mellon University scientists discovered that humans react better if the robots use filler words and aren't so authoritative in their actions. The study was based on teaching inexperienced bakers how to make cupcakes, with two human instructors leading the way. One instructor was precise in their speech, while the other added in words like "maybe," "um," and various other fillers. Duplicate videos were then made with a robot superimposed over the human instructor, and the subjects of the study took a questionnaire on which was the easier one to follow.

The results? Well the human and robot instructors using the filler words were liked a whole lot more than the clinical ones. It makes sense when you think about it, because wouldn't you rather follow someone who's a little more relatable than technical? After all, Bender from Futurama acts and talks like a human, and he's one of the more popular characters from that show. So, our future robot overlords should take note: if you want to really be our overlords, dumb down your speech.

Source: Cornell University

Register as a member to subscribe comments.
ClayMeow on February 13, 2014 08:58AM
I definitely agree with this study (and not just because I'm a Cornell alum). You see this type of thing in video game AI as well - if an AI is "too good" it feels like it's cheating. It's hard for programmers to capture the stupidity of the human element. lol

This news has comment postings disabled because it is now archived.

© 2001-2018 Overclockers Club ® Privacy Policy
Elapsed: 0.3041441441   (xlweb1)