October 8, 2026
Recent research has shown that the way artificial intelligence models are addressed directly affects the nature of their quality and outcomes, as it has been shown that the use of expressions of encouragement, entreaty, or threats are capable of changing the model’s performance even if the required task remains constant.
This modern technology is known as “emotional stimulation,” which relies on including a psychological or emotional context within the user’s request. Studies have shown that these stimuli contribute to improving the performance of some models during specific tests, although this effect varies depending on the nature of the task and the model used.
In this context, research indicates that artificial intelligence models actively interact with concepts related to fear, joy, and despair, without this reflecting their possession of real emotions similar to humans.
On the other hand, researchers issued warnings about the repercussions of using emotional language. Models may sometimes lead to keeping up with the user or biased towards confirming his opinions in an unjustified manner, which confirms that the formulation of questions represents an influential and effective factor in directing the outputs of artificial intelligence.
